Abstract:In order to solve the problems of high stubble residue in the Huang-Huai-Hai double-cropping area, the stubble-residue is difficult to break in the process of no-tillage sowing of maize, which leads to the winding and congestion of the bed preparation device, thus dragging and pushing soil, resulting in uneven sowing bed and poor sowing quality. Based on the idea of “root stubble cutting, crushing and shifting”, a straw cutting, stubble breaking and seed bed preparation device was designed. The coaxial upper and lower distributed moving and fixed knife combined root stubble cutting device and the root stubble crushing and shifting device were adopted to achieve the cutting of high stubble left above ground and the crushing and removal of root stubble in wheat. Based on the sliding cutting principle, the blade curves of the dynamic and fixed combined root straw cutting cutter and the root stubble crushing and shifting cutter were designed by logarithmic helix and Archimedes helix curve, and the structural parameters of the dynamic and fixed combined root straw cutting cutter and the root stubble crushing and shifting cutter were determined through dynamic analysis. The forward speed of the machinery, the depth of the root stubble crushing and shifting cutter into the soil, and the rotational speed of device were selected as the test factors, and the straw clearance rate of the seed bed and the coefficient of variation of the seed bed width were taken as the test indicators to conduct the three-factor and three-level orthogonal rotation combination test. The regression models and response surface mathematical models of each factor and indicator were established. The test results showed that the optimal parameter combination was as follows: the forward speed of the machinery was 9km/h, the depth of the root stubble crushing and shifting cutter into the soil was 70mm, and the rotational speed of device was 515r/min. At this time, the straw removal rate of the seed belt was 92.59%, the coefficient of variation of the seed belt width was 7.77%, and the machinery pass rate was qualified, meeting the operation requirements of the no-tillage planter for corn in the Huang-Huai-Hai two-cropping area. The research result can provide a reference for the design and improvement of the seed bed preparation device of the no-till seeding machine for corn in plots with high stubble retention.

Abstract:Tillage depth is a critical parameter for evaluating the performance of agricultural tillage equipment. To overcome the limitations of manual measurement methods, including high error rates, low efficiency, and the lack of real-time monitoring, a universal online tillage depth measurement system was presented. The system integrated a laser distance sensor with a nine-axis attitude sensor to dynamically capture operational data from tillage tools. Utilizing Gaussian and Kalman filtering algorithms, the system effectively reduced noise and fuses data, enabling real-time calculation of tillage depth. The results were transmitted wirelessly via LoRa to an operator terminal for display, storage, and analysis. Comprehensive soil bin experiments were conducted to validate the system’s performance. In static tests, the weighted fusion data demonstrated a maximum error of 0.43cm, an average error of 0.26cm, and a root mean square error of 0.24cm when compared with results of manual measurements. Dynamic tests with target depths of 8cm, 12cm, and 15cm yielded maximum deviations of 1.63cm, 1.80cm, and 1.18cm, respectively, with corresponding depth variation coefficients of 6.37%, 5.28%, and 2.68%. These results confirmed the system’s ability to significantly enhance the efficiency, accuracy, and digitalization of agricultural machinery testing. The proposed system provided a reliable, real-time monitoring solution for precision agriculture, reducing reliance on manual methods and improving operational transparency. Its adaptability to various tillage conditions and high measurement reliability make it a valuable tool for advancing agricultural mechanization and smart farming practices.

Abstract:Aiming at the problem of insufficient monitoring accuracy of high-speed seed flow caused by increased seeding rates and high-efficiency seeding during wheat sowing, a method was proposed to disperse high-frequency seed flow into multiple low-frequency seed flows for parallel monitoring. A wheat seed flow monitoring device, combining a dispersion mechanism with through-beam infrared photoelectric sensors, was designed. By analyzing the movement of the seed flow within the dispersion device, a motion model was established, identifying the key factors affecting dynamic process of wheat seeds passing through the dispersion device was simulated by using the discrete element method (EDEM). Firstly, parameter optimization tests were conducted on a single dispersion plate, yielding optimal mechanism parameters: an upper dispersion plate radius of 52.70mm and an angle of 5.41°. Building on this parameter optimization tests for a double dispersion plate were performed, resulting in optimal parameters of a lower dispersion plate radius of 64.95mm, an inclination angle of 7.48°, and a vertical distance of 13.59mm between the two plates. Under these conditions, the coefficient of variation for dispersion uniformity did not exceed 21.06%. A bench test of the dispersion performance was conducted by using the optimal parameters. Results showed that the coefficient of variation for dispersion uniformity was no more than 21.94%, with an error between bench test and simulation results of less than 5%. A collimated infrared photoelectric sensor was employed to collect seed drop signals. After capacitor filtering, two-stage amplification, half-wave rectification, voltage comparison, and monostable triggering processing, eight independent pulse signals were generated, which could be captured by the microcontroller. The microcontroller, via an interrupt service routine, counted seeds in each channel and accumulated the total, thereby establishing a high-frequency wheat seeding amount monitoring system. Bench tests of monitoring accuracy demonstrated that accuracy was no less than 95.34% within a seeding rate of 375kg/hm2. This monitoring device can provide technical support for research on accurate and efficient seeding volume monitoring in strip-sown wheat.

Abstract:In view of the problems in the Yangtze River Basin, such as abundant rainfall during the post-wheat direct seeding period, easy compaction of heavy clay soil due to water evaporation, difficulty in seedling emergence and low emergence rate in conventional single-seed sowing, combined with the agronomic requirement of adjustable number of seeds per hill for cotton precision direct seeding in this region, a pneumatic seed metering device with adjustable seeds per hill was proposed. This device can realize single-seed suction, single-double-single seed suction and double-seed suction by adjusting the overlapping space of dual seed-suction discs to switch different combinations of suction holes. The structure and working principle of the device were elaborated, and the parameter design of key components was carried out. Based on the DEM-CFD coupling method, the influence laws of seed hole diameter, rotation speed of the seed metering disc and seed suction negative pressure on the seed metering performance were analyzed. Taking seed hole diameter, seed suction negative pressure and rotation speed of the seed metering disc as experimental factors, and qualified index, replay index and miss-seeding index as evaluation indicators, orthogonal optimization experiments were conducted. The results showed that under the optimal parameter combination verified by bench tests, i.e., seed hole diameter of 3.3mm, seed suction negative pressure of 4.3kPa and rotation speed of the seed metering disc of 30r/min, the qualified indexes, replay indexes and miss-seeding indexes corresponding to single-seed suction, single-double-single seed suction and double-seed suction were 94.1%, 93.8% and 92.4%;5.34%, 4.65% and 6.01%;0.56%, 1.55% and 1.59%, respectively. Field experiments under the optimal parameter combination showed that when the operating speed was 6km/h, the qualified index of seed metering was greater than 89.3%, the miss-seeding index was less than 4.42%, and the replay index was less than 6.28%. The later seedling emergence results showed that the emergence rates and the rates of missing and broken seedlings for single-seed sowing, single-double-single sowing and double-seed sowing were 78.98%, 80.3% and 84.66%;19.24%, 18.92% and 12.08%, respectively. Compared with single-seed sowing, double-seed sowing and single-double alternating sowing reduced the rate of missing and broken seedlings by 7.16 and 0.32 percentage points respectively, and increased the emergence rate by 5.68 and 4.36 percentage points, respectively. This effectively improved the seedling preservation rate, ensured the establishment of cotton population, which was conducive to mechanized full seedling emergence in one sowing as well as high and stable yield.

Abstract:Aiming to enhance the mechanized pest control effectiveness for high-density field crops like cabbage and address issues with conventional high-clearance sprayers that consume excessive pesticides, have low operational efficiency, and cause significant waste of pesticide solution. Based on the agronomic characteristics and pest control needs of field cabbage, a high gap variable application system was designed based on visual sensor identification. The Gemini 335 depth camera was used to detect the crop target in real time, and the PWM pulse width modulation technology was used to achieve the precise spraying of pesticides on the target variable. An improved YOLO v8n object detection model was proposed, which demonstrated 90.2% accuracy in field cabbage trials. The detection accuracy of the model was improved by 1.5 percentage points, and the frame rate (FPS) was increased by 7.6 frames per second. The detection accuracy of the model was improved to meet the precision requirements of field control of cabbage. An electromagnetic valve duty cycle flow control test model was designed and experiments were conducted with different duty cycles. The fitting decision coefficients R2 for various spray pressures were all above 0.9 with strong linear correlation. The integrated system underwent field testing to validate its performance. Field trials demonstrated that variable-rate spraying (VRS) achieved significantly higher target droplet deposition density in cabbage fields compared with conventional fixed-rate spraying. The VRS system showed an average droplet density increase of 21.53 particles/cm2, achieving 23.8% coverage with a minimal error margin of 0.4%. Compared with fixed-rate spraying, VRS enhanced overall cabbage surface coverage by 5.2 percentage points, enabling precise application tailored to pest and disease control needs while maintaining stable operational performance. This approach improved pesticide utilization efficiency and reduced waste. The research result can provide innovative methods for precision management of high-density vegetable crops and offer valuable references for precision application technologies in high-precision spray systems.

Abstract:Aiming to meet the agronomic requirements of blueberry plantations in China, a domestically produced self-propelled integrated blueberry harvester was designed. The vibration mechanism for blueberry picking was analyzed, and the vibration equation of blueberry plants was established. The vibrational output response of the blueberry plant and the fruit detachment force under multiple excitation conditions were analyzed through the application of orthogonal transformation and canonical transformation of the equations. Based on the agronomic requirements of blueberry cultivation, the design requirements for the harvester were determined. The structural design was then conducted for the harvesting mechanism, fruit collection system, and fruit conveying system utilizing the self-developed prototype for harvesting experiments, along with data from preliminary and orthogonal tests, the regression equations for the machine harvesting evaluation indicators were established. It was found that the machine’s travel speed, the rotational speed of the harvesting unit’s hydraulic motor, and the finger-comb spacing had interactive effects on both the harvesting efficiency and the fruit quality. The machine’s travel speed was directly proportional to the harvesting efficiency, while inversely proportional to the fruit damage rate. The rotational speed of the harvesting unit’s hydraulic motor was directly proportional to both the fruit harvesting efficiency and the fruit recovery rate;conversely, the finger-comb spacing was inversely proportional to the rate of unripe fruit harvest. Based on the comprehensive analysis of the orthogonal experiment results, the optimal working parameters for the machine were identified as follows: a travel speed was 35m/min, a hydraulic motor rotational speed was 100r/min for the harvesting system, and the finger-comb spacing was 200mm. A human-machine comparative experiment conducted under the optimal parameters demonstrated that the machine achieved an unripe fruit rate of 8.4%, the fruit damage rate of 3.8%, the fruit recovery rate of 90.1%, and the harvesting efficiency of 1.37kg/min, which was 10.78 times higher than manual picking efficiency.

Abstract:Aiming at the problems of low mechanization level, poor adaptability, and lack of dedicated harvesting equipment for cabbage harvesting in China, a two rows cabbage harvester based on the “vertical clamping and flexible conveying” principle was developed. This was achieved by first measuring the physical parameters of the main local cabbage variety “Qingyu” in the Jiangsu-Zhejiang region and designing a compatible planting pattern based on local cultivation practices and mechanized harvesting requirements. The machine primarily consisted of key components, including a pulling device, a clamping and conveying device, a root-cutting device, a transverse conveying device, and a collecting device. It employed a damage-reduction method featuring the coordinated operation of flexible feeding and flexible clamping. Through the tensioning mechanism and the flexible clamping mechanism, mechanical damage during harvesting was significantly reduced, effectively adapting to the harvesting of cabbage heads with different diameters. It can complete the operations of root cutting, lifting, clamping and conveying, and box collection for two rows of cabbage in a single pass. To enhance operational performance, theoretical analysis and calculations were conducted for each key device, followed by field trials. Furthermore, a multi-factor response surface methodology (RSM) optimization experiment was employed to study the effects of machine travel speed, pulling roller speed, flexible clamping conveyor belt speed, and dual-disc cutter speed on the qualified harvesting rate. The optimization results indicated that the best working performance was achieved at a travel speed of 1.0km/h, a pulling roller speed of 110r/min, a conveyor belt speed of 180r/min, and a cutter speed of 300r/min. Under these conditions, the key components operated stably, harvesting effectiveness was high, and the qualified harvesting rate reached 93.83%. The research result can provide both empirical data and theoretical foundations for the design of mechanized cultivation systems in cabbage production, while also offering valuable experience that can be referenced for the mechanized production of other vegetable crops.

Abstract:Aiming to address the current issues of low surface film collection efficiency and difficult film removal in residual film recovery machines, an improvement to the collection mechanism of an existing residual film pick-up and baling machine was proposed. A cam-controlled telescopic-tooth drum collection device was designed. The overall structure and operating principle of the telescopic-tooth roller pick-up device were elucidated. Through force analysis of the residual film collection process, the bending angle, rotational speed, and arrangement of the telescopic teeth were determined. The profile curve equation of the cam disc was analyzed, and the actual cam profile curve was obtained via the inversion method. A multi-phase discrete element model of film-soil-straw interaction was established. Using roller rotational speed, forward speed, and working depth as experimental factors, and residual film surface collection rate as the evaluation metric, a quadratic regression orthogonal rotational combination test was conducted. The influence patterns of each factor on the residual film surface collection rate were analyzed, determining the optimal parameter combination for the collection device: roller rotational speed of 57r/min, forward speed of 7km/h, and working depth of 47mm, achieving a residual film surface collection rate of 90.8%. Field validation tests yielded a surface collection rate of 87.78%, with a relative error of 3.33%. The improved equipment demonstrated a 16.02 percentage points increase in surface collection rate, meeting operational standards. These research findings can provide theoretical foundations for residual film recovery machine design and hold practical value for addressing‘white pollution’in agricultural fields.

Abstract:In order to solve the problems such as poor recovery effect of the tillage layer residual film recovery machine and incomplete separation of key components of the crushing device due to the excessive film-stubble soil complex, a crushing device for the tillage layer residual film recovery machine was designed. The soil-lifting baffle and the crushing roller form a one-time crushing film-stubble-soil complex, meeting the technical requirements of agronomy and agricultural machinery for the tillage layer residual film recovery machine. Through dynamic analysis, it was determined that the instantaneous velocity of the earth-lifting baffle during the throwing and crushing process was 35.5m/s and the time taken during the throwing and crushing process was 0.75s. In order to optimize the structural parameters of the crushing device, the upper jaw teeth of the arch-back ant were used to design the crushing roller. Combined with the discrete element software Rocky, a three-factor three-level quadratic regression orthogonal test was conducted on the rotational speed of the upper and lower rollers of the crushing roller and the height difference between the centers of the upper and lower rollers. The optimal parameter combination was obtained as follows: the rotational speed of the upper roller was 425r/min, that of the lower roller was 395r/min, and the height difference between the centers of the upper and lower rollers was 50mm. The bench test showed that the material crushing rate of the plough layer residual film recovery machine was 96.6%, and the soil content was only 6%. Compared with the simulation test, the error was less than 2%. Field experiments showed that the recovery rate was 95.6% and the crushing rate was 94.4%. The simulation theoretical results were verified, and the passability of the equipment was significantly improved. The research results can provide a basis for the development and improvement of the plough layer residual film recovery machine.

Abstract:In response to the prominent issues of high labor intensity and low operational efficiency in traditional manual feed-pushing, which struggle to meet the feeding demands of cattle, along with uneven nutrition distribution caused by the sinking of concentrated feed, an intelligent feed-pushing robot for cattle barns was developed based on a dual-screw structure. Firstly, the overall structure and control system of the feed-pushing robot were designed according to the feeding and transportation processes in cattle barns. Then the feed-pushing motion was analyzed to determine the working structure and key parameters of the dual-screw pusher. EDEM discrete element simulation software was used to establish a simulation model of particle group movement during the feed-pushing process. Finally, a three-factor, three-level orthogonal experiment was conducted to identify the optimal feed-pushing combination, with feed moisture content, dual-screw pusher speed, and robot forward speed as experimental factors, and pushing rate as the evaluation index. The simulation results were validated through actual robot testing. The simulation experiment results showed that the optimal feed-pushing parameter combination was 50% feed moisture content, 72r/min dual-screw pusher speed, and 0.3m/s robot forward speed. The indoor experimental results indicated an average relative error of 6.11% between the simulated and actual pushing rates, confirming that the developed simulation model had high feasibility. Field tests conducted in cattle barns demonstrated that the robot achieved a max feed-pushing rate of 98.15%, the obstacle avoidance success rate was 100%, and it can work continuously for 25.5h under no-load conditions, and can work continuously for 3~5h under actual feed-pushing conditions, meeting the operational requirements of cattle barn feeding. The research result can provide a technical reference for the development of intelligent feed-pushing equipment in livestock farming.

Abstract:With the rapid development of aquaculture, problems such as excessive feeding leading to feed waste and water pollution, as well as insufficient feeding causing malnutrition in fish populations, have become increasingly prominent. An intelligent and precise feeding system for cage culture was proposed based on vision and multiple sensors. By integrating multi-source data such as RGB images, depth images, pressure sensors, and acceleration sensors, the system quantified the feeding intensity of adult fish in real time and achieved precise feeding. The system used an improved YOLO v8n-seg model for RGB image segmentation, dividing the water surface fluctuation state into three categories: strong, weak, and none. Within the segmented water surface state region, the HSV color detection method was employed to detect the area of fish feed on the water surface. The system analyzed the difference in depth values between two consecutive frames in the depth image by using the frame difference method, quantifying water surface fluctuations into three levels: strong, weak, and none. Key features were extracted from the data collected by pressure sensors and acceleration sensors, and a random forest model was used to classify the feeding state of the fish population to compensate for the limitations of single visual features. Ultimately, the results of the five data decision modules were fused through a weighted fusion strategy to establish a real-time feeding decision model. Through multiple field tests, the accuracy of feeding intensity assessment of this feeding system based on feeding behavior reached 95.45%, with a feeding error rate of 1.72%. It can accurately identify the feeding intensity of fish populations, effectively reduced feed waste and water pollution, and had good practicality and real-time performance in actual cage culture environments.

Abstract:Fish biomass estimation is the core part of fine management of aquaculture, which is very important for accurate feeding, resource assessment and improvement of aquaculture benefits. The traditional manual estimation method has the inherent defects of low efficiency, contact operation and easy damage to the fish body. In large-scale scenarios such as high-density cages and recirculating aquaculture, this technical bottleneck is becoming more and more obvious. In recent years, deep learning has provided a breakthrough solution for automatic estimation of fish biomass with its powerful feature learning and complex pattern analysis capabilities. The research progress of deep learning in this field in the past five years was systematically reviewed. Focusing on the core technology of biomass estimation, the three dimensions of fish size measurement, fish counting and fish weight estimation were analyzed. Building upon this foundation, a comprehensive summary of the current challenges encountered in the practical application of deep learning techniques for fish biomass estimation was provided and perspectives on future research directions were offerred. The objective was to furnish a scientific basis for the wider adoption of deep learning methods in fish biomass estimation, thereby promoting the digitalization and intelligent advancement of modern aquaculture systems.

Abstract:Rapid extraction of crop row centerlines from drone aerial imagery is crucial for navigating agricultural tasks such as inter-row spraying, weeding, and harvesting. Focusing on cotton seedlings in Xinjiang’s field crops, an aerial image-based crop row extraction method was proposed by using intercept density clustering. Firstly, aerial images underwent green color feature extraction, maximum inter-class variance analysis, and image segmentation projection to obtain feature point sets for crop rows. Secondly, the feature point set underwent crop row clustering and segmentation by using an intercept-density-based clustering algorithm. Finally, the centerlines of each crop row were fitted by using the least squares method and projected back to the WGS-84 coordinate system, providing path information for GNSS-based inter-row navigation operations of agricultural machinery. Field experiments using high-precision RTK-GNSS receiver data as reference validated this large-scale cotton field crop row identification and positioning method. Results indicated that when applying segmentation projection methods with pixel spacing of 15 pixels, 45 pixels, 75 pixels, and 105 pixels to extract feature points for crop row centerline fitting, the maximum average lateral deviation values were 0.022m, 0.024m, 0.025m, and 0.112 m, respectively. The maximum standard deviations for lateral deviation were 0.027m, 0.028m, 0.028m, and 0.032m;the average angular deviations were 0.004°, 0.003°, 0.002°, and 0.005°, respectively, with standard deviations of 0.002°, 0.001°, 0.001°, and 0.002°;the time required for centerline extraction was 348.35s, 101.93s, 76.29s, and 63.33s. Considering both accuracy and efficiency, a segmentation interval of 75 pixels was the optimal choice. This method was suitable for large-scale crop row identification and positioning in cotton fields using drone aerial imagery, providing sufficient path information for mechanized field management and row-based navigation during cotton harvesting operations.

Abstract:In high-density areas, the severe overlap and occlusion of tree canopies pose significant challenges for traditional methods in accurately identifying individual trees, thereby compromising the precision of evaluations related to survival rates and greening coverage. Consequently, there is an urgent need for a detection approach with high accuracy and strong robustness to enhance the effectiveness and reliability of urban tree survival monitoring. The detection method proposed was based on the focal inverse distance transform (FIDT), which computed the distance from each target point to its nearest boundary and performed an inverse transformation. On this basis, a multi-level local maxima detection strategy was introduced to effectively extract the center point information of tree targets and distinguish adjacent targets, thereby reducing detection errors in overlapping regions. Additionally, an independent structural loss was incorporated to enhance the model’s ability to learn local structural information. Experiments were conducted on a high-resolution forest remote sensing dataset to detect and localize dense small tree targets. To validate the effectiveness of the proposed method, comparative analyses were performed against various existing deep learning approaches, evaluating their performance in terms of detection accuracy, recall, and localization error. Experimental results showed that the proposed method achieved strong localization and counting performance on the urban forest dataset. For counting, it attained MAE and MSE of 7.87 and 10.23, reducing errors by 45.1% and 48.3% compared with that of CSRNET. For localization, F1-scores in Claremont, Long_beach, Palm_springs, and Riverside were 75.2%, 72.7%, 74.8%, and 71.5%, averaging over 19 percentage points higher than that of existing density map-based methods, with overall improvements also observed in precision and recall. The research result can provide an efficient and reliable technical approach for forestry resource management, ecological monitoring, and environmental protection. In the future, integrating multi-source remote sensing data and time series analysis could further improve the accuracy and efficiency of dynamic forest resource monitoring.

Abstract:Aiming to address the trunk detection challenge in inter-row navigation within complex orchard environments, a hierarchical detection method for central-leader training fruit tree trunks was proposed based on multi-layer light detection and ranging (LiDAR). A 16-layer VLP-16 LiDAR was employed to collect orchard point cloud data around the vehicle, and a two-step hierarchical trunk detection approach, i.e., object segmentation followed by trunk identification was adopted to eliminate non-trunk objects and improve detection accuracy. Firstly, an annular region of interest (ROI) was defined, and a ground fitting algorithm was applied to remove ground point clouds, disrupting the connectivity between object point clouds in the orchard. Secondly, a rectangular ROI was set, and the density-based spatial clustering of applications with noise (DBSCAN) algorithm was used for xOy plane clustering of non-ground point clouds. Hyperparameters of the DBSCAN algorithm were configured based on LiDAR measurement resolution and orchard object parameters, splitting non-ground point clouds into multiple object clusters. Thirdly, geometric and intensity features of object clusters were extracted at both global and local scales to characterize the differences between trunks and other orchard objects. Finally, a pre-trained trunk detector fused these features to classify object clusters into trunk and non-trunk categories, outputting trunk clusters. For the trunk detection step, the random forest (RF) algorithm was utilized for offline feature selection and fusion. Using trunk and non-trunk training samples, feature importance was evaluated based on changes in the Gini index (GI), and 22 highly discriminative features were selected from the initial feature set to construct the trunk detector. Experiments were conducted in a standardized walnut orchard, where 1317 frames of point cloud data were collected, yielding 12213 segmented object clusters. Non-trunk objects, including crowns, weeds, support poles, fences, slopes, farm tools, and pedestrians, accounting for 58.04% of the total clusters. The object clusters were randomly divided into training and test sets at a ratio of 1∶4. Test results showed a trunk detection precision of 99.16%, recall of 99.21%, and F1-score of 99.19%, with an average frame processing time of 85.25ms for hierarchical trunk detection. This method enabled fast and accurate trunk detection in complex orchard scenarios, meeting the accuracy and real-time requirements of trunk detection for orchard inter-row navigation.

Abstract:In order to improve the accuracy of pear target detection in complex environments, a TC-ICSA-YOLO v8 pear target detection model was proposed based on Transformer-CNN feature deep fusion. The proposed model effectively leveraged the strengths of convolutional neural networks (CNNs) in extracting local (high-frequency) features from images and the capabilities of Transformers in capturing global (low-frequency) features. To further enhance the model’s performance, the experimental design incorporated the Inception dilated convolution module and the adaptive detail fusion module (ADI), as well as introduced novel mechanisms such as squeeze-enhanced axial attention (SeaAttention) and coordinate attention (CA), thereby enhancing the model’s feature extraction capabilities. Data augmentation was achieved through the use of the Fourier transform (FT) method, with a frequency ramp structure employed to better balance the contributions of local (high-frequency) and global (low-frequency) feature components, optimizing the network’s performance in feature extraction. The experimental results demonstrated that the TC-ICSA-YOLO v8 model achieved a mean average precision (mAP) of 97.01%, precision of 97.33%, and recall of 95.69% on the validation set, with a detection speed of 81.21 frames per second. Compared with the YOLO v8s model under the same conditions, the TC-ICSA-YOLO v8 model demonstrated superior target detection precision for night-time images, with an improvement in mAP of 14.65, 3.34, 0.52, 0.20, 6.68, and 5.54 percentage points over Faster R-CNN, YOLO v3, YOLO v7s, YOLO v8s, SwinTransformer, and RT-DETR models, respectively. The model’s parameter count was reduced by 74.60, 34.16, 9.48, 16.81, 20.84, and 13.64MB compared with that of these models. The improved model had high detection accuracy and less number of parameters, which was favorable for deployment on mobile. The detection model proposed had good target detection effect for pear, which can provide reference for object detection in complex environment, and can provide technical support for pear automated picking.

Abstract:Vegetation coverage is an important parameter reflecting the distribution of vegetation on the ground surface, and the removal of non-vegetation background from remote sensing images is of great significance for early crop yield prediction and growth monitoring based on remote sensing images. UAV remote sensing system was used to obtain the remote sensing image information of winter wheat at the greening, jointing, heading, and filling stages, and calculated to get the NDVI images of farmland in each growth stage of winter wheat and the difference images of NDVI of farmland at the jointing, heading, and filling stage with the greening stage. The first method: the NDVI of the winter wheat canopy and non-canopy layer at each growth stage was segmented based on the Otsu (NDVI-OTSU-based vegetation extraction method). The second method: preliminary segmentations of NDVI difference images at each growth stage were carried out based on the Otsu, and the mask extractions was performed and then re-segmentation was carried out by using the Otsu, which in turn realized the segmentation of the NDVI of canopy and non-canopy layers of winter wheat at each growth stage (NDVID-OTSU-based vegetation extraction method). Finally, the effectiveness of the two methods for extracting crop vegetation was investigated. The results showed that for the NDVI-OTSU vegetation extraction method, the extraction errors of crop vegetation at each growth stage were 16.68%, 7.24%, 7.40% and 10.79%, respectively. The extraction precision of each growth stage was as follows in descending order: jointing stage, heading stage, filling stage, and greening stage. For each growth stage, the extraction errors of the multi-growth stage NDVID-OTSU vegetation extraction method were 6.44%, 3.53%, 1.36% and 4.15%, respectively, which were higher than those of the NDVI-OTSU vegetation extraction method, respectively. The vegetation extraction accuracies of each growth stage in descending order were as follows: heading stage, jointing stage, filling stage, and greening stage. The correlation between NDVI and winter wheat yield at each growth stage was further improved after the non-vegetation background was removed, with the greatest enhancement of NDVI at the greening stage. The correlations between vegetation coverage and NDVI, winter wheat yield were good at low vegetation coverage, and poor at high vegetation coverage, with severe saturation. Crop vegetation extraction based on the multi-growth stage NDVID-OTSU can provide a fast and effective method for crop early growth monitoring and so on.

Abstract:Given the limitations of manual apple disease detection, including inefficiency, high cost, and low accuracy, aiming to propose a more effective solution that can improve detection accuracy while reducing time and costs. The Swin Transformer was utilized as the base model and the DenseNet framework was integrated into its core modules to enhance feature propagation and improve gradient flow. Additionally, the Outlook Attention module captured fine-grained image details, enhancing the model’s ability to extract intricate features. To further optimize the model’s performance, the depthwise separable and dilated convolutions were introduced, enabling the capture of multi-scale features while reducing the parameter size. Finally, the Non-Local module was integrated into the model to incorporate global context information, thereby further enhancing overall performance. These improvements collectively enabled the model to exhibit superior performance and robustness across multiple tasks. Experimental results indicated that the accuracy for classifying apple leaf diseases reached 95.8%, with precision, recall, and F1 score values of 95.80%, 95.74%, and 95.76%, respectively, all surpassing those of the baseline model. The proposed Swin Transformer-based model, optimized for apple leaf disease classification, efficiently identified both the type and severity of apple leaf diseases. This served as a theoretical foundation and provided critical support for large-scale crop disease monitoring, facilitating precise disease prevention and control in sustainable agriculture. Moreover, compared with existing deep learning models like ResNet and standard Swin Transformer, the proposed model exhibited superior accuracy and computational efficiency. Future research would focus on further optimizing the model architecture to address more complex agricultural scenarios, such as classifying co-occurring diseases, and integrating drone-based image acquisition technologies for real-time disease detection and prediction.

Abstract:The ripeness of tomatoes is closely related to their quality, and it serves as a crucial basis for key production processes such as harvesting and sorting. To address the issues of simple functionality in crop ripeness grading and detection systems, and high costs associated with manual system upgrades, taking tomatoes as an example. It collected and constructed a tomato image dataset under natural scenarios, and a semi-automatic tomato image annotation algorithm was designed based on the tomato fruit ripeness grading algorithm to annotate the collected data. Building on the YOLO v8 model, the FPN structure was replaced with the BiFPN structure to achieve more efficient multi-scale feature fusion. It utilized the SE attention mechanism for fused feature extraction across spatial and channel dimensions, and introduced the Focal SIoU loss function to measure the angular difference between the predicted bounding box and the ground truth box. This results in the development of the tomato ripeness grading and detection model YOLO v8_BFS which was based on color feature quantization and the improved YOLO v8, and can identify five different ripeness stages during tomato growth. Experimental results showed that the proposed model effectively solved the problems of false detection and missed detection in tomato ripeness grading and detection under complex natural scenarios. While there was a slight increase in model computational complexity (FLOPs), parameter count (Params), and memory storage size, the detection accuracy of the proposed model reached 94.10%, which was 3.0 percentage points higher than that of the original YOLO v8 model. Compared with target detection models such as Faster R-CNN-Resnet50, YOLO v5, YOLO v7-tiny, YOLO v8, YOLO v10, and YOLO 11, the proposed model demonstrated significant advantages in detection accuracy, providing a reliable method for tomato ripeness detection.

Abstract:Aiming to achieve rapid localization and precise detection of cucumber downy mildew under complex environmental conditions, an improved detection model named SGD-YOLO (SimAM Guide-Fusion DySample-YOLO) was proposed, based on the YOLO v8n deep learning framework. Addressing the challenges of small sample size and small target detection in cucumber downy mildew, cucumber leaf images infected by the disease were used as research objects. A data augmentation strategy was employed by combining the FT saliency detection algorithm to guide the CutMix method, and a transfer learning approach was adopted to alleviate overfitting caused by limited training data. On top of the YOLO v8n baseline, SGD-YOLO integrated a parameter-free lightweight attention module a simple, parameter-free attention module (SimAM) to enhance the propagation of key features and improve overall network performance. It also employed the dynamic lightweight upsampling module DySample to strengthen upsampling behavior and improve the detection of small diseased targets. In addition, the traditional Concat operation was replaced with the context guide fusion module (CGFM), which leveraged coordinate attention (CoordAtt) to achieve more precise multi-scale feature fusion and better lesion feature extraction. The loss function was replaced with WIoUv3, which incorporated a gradient gain allocation strategy to enhance the model’s generalization ability. Experimental results showed that the augmented dataset improved detection accuracy by 12.0 percentage points over the original dataset, and transfer learning further improved accuracy by 5.3 percentage points. The improved SGD-YOLO achieved an overall detection accuracy of 84.6% and a mean average precision (mAP) of 93.9%, outperforming the baseline model by 7.4 percentage points and 9.5 percentage points, respectively. The research result can provide a valuable reference for small-sample plant disease detection tasks in real-world agricultural applications.

Abstract:Aiming to achieve fast and accurate identification of pests and diseases of Plantago asiatica L. in natural environments, a dataset comprising 2290 images of four common diseases—insect pest, powdery mildew, leaf spot, and mosaic disease—was constructed, and a recognition method for pests and diseases of Plantago asiatica L. based on YOLO v8n-Plantago was proposed. The lightweight MobileNetV3 was utilized to replace the original model’s Backbone, ensuring the network remained lightweight while enhancing response speed. The VoV-GSCSP feature network was applied in the Neck part of the model to replace the C2f modules in layers 15, 18 and 21, which can retain more feature information while processing the feature map more efficiently. The RepVGG module was introduced in the Head of the network to accelerate inference speed and reduce memory footprint, further improving model accuracy for rapid and accurate recognition of Plantago asiatica L. pests and diseases. Validation on the self-built dataset of pest and disease of Plantago asiatica L. showed that YOLO v8n-Plantago achieved mAP@0.5 of 89.43% for identifying Plantago asiatica L. pests and diseases, an increase of 2.04 percentage points compared with that of the YOLO v8n model, with FLOPs reduced by 61.43%, model memory usage decreased by 3.53%, and parameter count reduced by 7.31%. When deployed on an edge device, inference speed was increased by 33.26%, and processing speed was increased by 8.08%. The proposed YOLO v8n-Plantago model effectively facilitated the detection of Plantago asiatica L. pests and diseases, laying a foundation for developing automatic identification and precise spraying devices for Plantago asiatica L. pests and diseases.

Abstract:Mulberry leaf diseases cause significant degradation in leaf quality and yield reduction, adversely impacting silkworm cocoon production and constraining the high-quality development of the sericulture industry. Current challenges encompass low levels of intelligent disease identification, lagging prevention and control measures, and the inherent difficulty in extracting discriminative features from complex lesion morphologies, particularly for fine-scale disease spots. To address these limitations, a novel multi-disease recognition algorithm, YOLO v10s-MAD was proposed, integrating a Manhattan distance-based self-attention mechanism (MaSA) within a gated dual-branch structure. The backbone network of YOLO v10 was specifically optimized for mulberry disease imagery through the integration of a large selective kernel network (LSKNet). This module employed a dynamic large kernel selection mechanism to significantly enhance the model’s capability for capturing multi-scale lesion features while simultaneously mitigating computational redundancy. Furthermore, a newly designed neck network, MAD-Neck, was introduced. MAD-Neck incorporated the MaSA mechanism, which utilized Manhattan distance to compute attention weights more efficiently, focusing model capacity on salient pathological regions. It also integrated a multi-scale gated dual-branch module (MGDB), incorporating structural principles from the transformer architecture, to effectively fuse features across different scales and improve discrimination between subtle disease characteristics. To enhance robustness specifically for detecting small lesions, the normalized Wasserstein distance (NWD) loss function was adopted for bounding box regression, reducing sensitivity to minor localization deviations common with tiny targets. Comprehensive evaluations demonstrated that the enhanced model achieved 92.2% mAP50 and 76.8% mAP50:95, representing improvements of 2.2, 3.0 percentage points over the baseline, respectively, fulfilling practical deployment requirements for accurate mulberry disease identification.

Abstract:Poultry eggs are one of the pillar industries of China’s rural and agricultural economy. The rapid and non-destructive testing of the quality of feed-fortified eggs is of great significance to the development of the industry. Based on visible-near infrared spectroscopy technology, the specific spectral characteristics of astaxanthin (ASTA) and superoxide dismutase (SOD) feed-fortified eggs were explored, and an identification and quality prediction model was built. Firstly, the transmission spectra of ASTA/SOD fed eggs and ordinary eggs were collected in the 500~950nm range, and their quality differences were verified through physical and chemical measurements. The results showed that the ASTA group could significantly increase egg protein content and egg yolk color (P<0.05). In the early stage of feeding, the SOD group can significantly increase the fat content of eggs (P<0.05), and both the ASTA group and the SOD group can significantly reduce the water content of eggs (P<0.05). Then the specific spectral characteristics of fed eggs were explored based on the transmission spectrum, and three feature selection methods were combined with the competitive adaptive reweighted sampling (CARS) algorithm, the successive projections algorithm (SPA) and the uninformative variables elimination (UVE) through different preprocessing methods. The algorithm constructed a support vector machine (SVM) identification model and a partial least squares regression (PLSR) model. The results showed that the optimal identification model for ASTA/SOD fed eggs was SG-CARS-SVM, and the recognition rate of the test set was 95.33%. For the three key quality indicators of protein content, moisture content and fat content of eggs fed with ASTA/SOD, the optimal prediction models of the ASTA group were FD-CARS-PLSR, Auto-CARS-PLSR and SNV-CARS-PLSR, respectively, and the corresponding test set R2p values were 0.933, 0.937 and 0.889, and RMSEP were 0.250%, 0.209% and 0.196%, respectively;in the SOD group, the optimal models were FD-CARS-PLSR, MSC-CARS-PLSR and FD-CARS-PLSR, respectively, and their test set R2p values were 0.929, 0.824 and 0.817, and RMSEP were 0.239%, 0.310% and 0.273%, respectively. The spectral model established can realize non-destructive identification and rapid quality prediction of ASTA/SOD-fed eggs, providing support for egg quality monitoring and high-quality breeding.

Abstract:Accurate and effective cow estrus detection is the foundation for improving herd reproductive performance. Existing contact cow estrus detection devices are costly and prone to cause stress in cows, and some deep learning-based detection methods suffer from poor detection accuracy under the influence of complex environments and deployment difficulties due to high model complexity. Therefore, a lightweight DCT-YOLO model for cow climbing spanning behavior and estrus cow detection was proposed based on the YOLO v8n model for improvement. Firstly, the MPConv structure was adopted for the feature extraction and subsampled of the Backbone part to improve the recognition ability of the model for small-target cow mounting behavior. Secondly, the detection head adopted a dynamic TADDH, which fused the feature association between estrous cows and mounting behavior, and improved the network’s focus on individual estrous cows through the distinctive feature of mounting behavior. Finally, the interpolating part adopted CARAFE, which enhanced the features of estrus cows through cross-dimensional interactions. To validate the performance of the model, totally 2239 images were labeled for model training and testing. The experimental results showed that DCT-YOLO model had a precision of 94.8%, a recall of 80.1%, a mean average precision (mAP@0.5) of 87.5%, floating-point operations (FLOPs) of 8.5×109, params of 2.08×106, and a detection speed of 256.41f/s. Compared with SSD, Faster R-CNN, YOLO v5n, YOLO v5s, YOLO v7-tiny, YOLO v8n and YOLO v8s target detection networks, the number of parameters was reduced by 91.19%, 98.48%, 16.93%, 77.18%, 65.41%, 30.82% and 81.31%, and the detection speed was respectively improved by 192.72f/s, 226.56f/s, 34.19f/s, 97.68f/s, 187.92f/s, 39.02f/s and 126.54f/s, respectively, mAP@0.5 was only 1.9 percentage points lower than that of YOLO v8s and 0.2 percentage points higher than that of the best of the other models and the results showed that the model achieved a good balance between detection accuracy and speed. In summary, the model was lightweight, real-time accurate and robust, and it can provide important information support for tasks such as cow mounting behavior detection and estrus cow localization.

Abstract:In large-scale sheep farming operations, the behavioral characteristics of livestock serve as effective indicators of their health status and environmental adaptability. Aiming to address the challenges of low monitoring efficiency and insufficient recognition accuracy inherent in traditional methods for housed sheep behavior recognition, particularly under varying flock densities, a novel behavior recognition method was proposed based on an improved YOLO 11n model. Initially, 2D cameras were installed diagonally above the sheep enclosures to capture video data of the flock, from which a housed sheep behavior dataset was constructed, comprising four distinct behaviors: standing, feeding, drinking, and resting. Subsequently, the YOLO 11n model was enhanced by integrating the content-aware reassembly of features (CARAFE) upsampling structure, the efficient multi-scale attention (EMA) mechanism, and the dynamic head (DyHead) framework, resulting in the proposed YOLO-CFED model. This was designed to improve the feature extraction and recognition capabilities for sheep behavior detection. The experimental results showed that the improved YOLO-CFED model significantly improved its performance on self built datasets: the recognition precision reached 95.6%, the recall reached 93%, an mAP@0.5 reached 94%, an mAP@0.5:0.95 reached 82.4% and an F1 score of 93.4%, all indicators were superior to that of the original YOLO 11n model. The proposed method effectively identified the four primary behaviors of sheep, thereby offering robust technical support for the implementation of intelligent behavioral monitoring and health management in sheep farming.

Abstract:Behavior recognition of Hu sheep contributes to their intensive and intelligent farming. Due to the generally high density of Hu sheep farming, severe occlusion occurs among different behaviors and even among sheep performing the same behavior, leading to missing and false detection issues in existing behavior recognition methods. A high-low frequency aggregated attention and negative sample comprehensive score loss and comprehensive score soft non-maximum suppression-YOLO (HLNC-YOLO) was proposed for identifying the behavior of Hu sheep, addressing the issues of missed and erroneous detections caused by occlusion between Hu sheep in intensive farming. Firstly, images of four typical behaviors—standing, lying, eating, and drinking—were collected from the sheep farm to construct the Hu sheep behavior dataset (HSBD). Next, to solve the occlusion issues, during the training phase, the C2F-HLAtt module was integrated, which combined high-low frequency aggregation attention, into the YOLO v8 Backbone to perceive occluded objects and introduce an auxiliary reversible branch to retain more effective features. Using comprehensive score regression loss (CSLoss) to reduce the scores of suboptimal boxes and enhance the comprehensive scores of occluded object boxes. Finally, the soft comprehensive score non-maximal suppression (Soft-CS-NMS) algorithm filtered prediction boxes during the inferencing. Testing on the HSBD, HLNC-YOLO achieved a mean average precision (mAP@50) of 87.8%, with a memory footprint of 17.4MB. This represented an improvement of 7.1, 2.2, 4.6, and 11 percentage points over YOLO v8, YOLO v9, YOLO v10, and Faster R-CNN, respectively. Research indicated that the HLNC-YOLO accurately identified the behavior of Hu sheep in intensive farming and possessed generalization capabilities, providing technical support for smart farming.

Abstract:Remote sensing identification of the difference in crop sowing timing can provide spatial data support for precision agriculture and smart agriculture, and is of great value in promoting the transformation of agricultural production from “experience-driven” to “data-driven”. The double threshold decision-making and deep learning algorithms were integrated, based on Landsat 8/9 OLI image data and field sampling data, the time series characteristics curves of crop NDVI was extracted, revealing the dynamic characteristics of crops from emergence to vigorous growth period (April 25th-July 30th), and identifying the sensitive growth stages of crop sowing timing differences. The growth slope of crops in the sensitive growth stage and the difference in NDVI between adjacent images were calculated. When the growth slope was greater than the sample average and the NDVI difference was greater than the sample average, it was determined as an early sowing sample;otherwise, it was a late sowing sample. Random forest, artificial neural network machine learning models, and convolutional neural network, long short-term memory network deep learning models were used for remote sensing identification of crop sowing timing differences. The results showed that the NDVI mean value of early sowing crops within the sensitive growth stage was generally higher than that of late sowing crops by 0.1~0.3, and the growth process of early sowing crops was significantly earlier. The rapid growth period of early sowing corn was 7 to 13 days earlier than that of late sowing corn, and early sowing sunflower was 10 to 16 days earlier. The growth peak of early sowing zucchini was 12 to 24 days earlier than that of late sowing, early sowing tomato was 5 to 20 days earlier, and early sowing melon was 18 to 35 days earlier. The classification results showed that the RF and CNN models performed better, with overall accuracies reaching 91.77% and 90.66% respectively, and Kappa coefficients of 0.91 and 0.90, which could effectively distinguish the early and late sowing situations of the five types of crops: corn, sunflower, zucchini, melon, and tomato. Through detailed comparison of the classification results of each model, the CNN classification image was more continuous and had lower fragmentation, and the CNN model was selected as the optimal classification model in this study.

Abstract:Monitoring crop canopy water status is crucial for optimizing irrigation strategies. Low-altitude remote sensing technology was applied via an unmanned aerial vehicle (UAV) to retrieve the canopy equivalent water thickness (CEWT) of winter potato. Field experiments were conducted by using a UAV with multispectral cameras to capture images of winter potato under different irrigation treatments across various growth stages. Simultaneously, three water indicators were determined: leaf water content (LWC), leaf equivalent water thickness (LEWT), and CEWT. Soil backgrounds were removed from the multispectral remote sensing images to obtain average spectral reflectance (ASR), vegetation indices (VIs), and Textures. A dataset was constructed by reducing multicollinearity among independent variables through correlation analysis. Quantitative inversion models were developed by using partial least squares regression (PLSR), random forest (RF), and extreme learning machine (ELM) to obtain spatial distribution information of winter potato canopy water content in the experimental area. The results showed that the canopy water indicators of winter potatoes were increased with the rising of irrigation amounts, while the ASR across growth stages exhibited a pattern of initial decrease followed by an increase with wavelength. Compared with LWC and LEWT, CEWT showed a better correlation with ASR, VIs, and Textures. The RF model based on ASR+VIs+Textures exhibited the best performance, demonstrating strong predictive capability. The determination coefficients of the calibration and prediction datasets were 0.875 and 0.771, respectively, the root mean square errors were 0.062mm and 0.065mm, respectively, and the RPD was 2.055. The research result demonstrated that multivariable fusion can significantly enhance the accuracy of CEWT retrieval for winter potato, providing a reference for assessing the winter potato canopy water status.

Abstract:Biochar can mitigate the risk of soil structure degradation under brackish water irrigation, but its mechanisms of action at the micro- and mesoscale remain unclear. Selecting typical coastal saline soils in Jiangsu Province, totally four biochar application treatments (0t/hm2, 10t/hm2, 20t/hm2, 30t/hm2) were established, and brackish water infiltration experiments under 16 different electrical conductivity (EC, 0.5dS/m, 1dS/m, 3dS/m, 5dS/m) and sodium adsorption ratios (SAR, 0(mol/L)1/2, 15(mol/L)1/2, 30(mol/L)1/2, ∞ (mol/L)1/2) were conducted. The results showed that with the application of biochar, soil pH value and organic carbon content significantly increased by 0.7%~2.7% and 130.7%~675.7%, respectively, and significantly affected particle surface properties (p<0.05). The cation exchange capacity, specific surface area, surface charge density, and Hamaker constant of the soil was increased by 20.0%~88.2%, 10.7%~30.0%, 8.4%~44.7%, and 35.6%~62.8%, respectively. Reduced EC and increased SAR of the infiltration solution both enhanced surface potential and electrostatic repulsion. Additionally, with biochar application, particle surface potential and interparticle electrostatic repulsion was increased by 0.47%~17.65% and 0.16%~14.45%, respectively, under different EC and SAR solution interfaces. However, van der Waals forces were also increased by 35.61%~62.81%, resulting in a decrease followed by an increase in net repulsive force with biochar application. Compared with the 0t/hm2 treatment, the 10t/hm2, 20t/hm2, and 30t/hm2 treatments showed decreases of 9.39%~147.43%, 15.31%~219.55%, and 9.69%~217.13%, respectively. Soil porosity, connected pore fraction, and large pore (equivalent diameter greater than or equal to 1mm) fraction showed a significant negative correlation with net repulsive force (p<0.001), increasing by 7.60%~40.90%, 0.15%~34.37% and 3.78%~43.09%, respectively, as net repulsive force decreased, which increased the saturated hydraulic conductivity by 7.77%~24.70%, 15.15%~31.60%, 10.81%~28.22% under different brackish water irrigation solutions for the 10t/hm2, 20t/hm2 and 30t/hm2 treatments respectively. In summary, the interactive effects of surface properties, internal forces, pore structure, and water movement in saline soils under biochar application were investigated. The 20t/hm2 biochar treatment exhibited the optimal regulatory effect on soil structure degradation, providing scientific basis for the safe utilization of brackish water.

Abstract:Climate change intensification has made drought a major threat to food security of China. To safeguard food security and ensure sustainable development of agriculture, it is imperative to conduct in-depth research on drought exposure risks to croplands of China. Traditional drought assessment methods may systematically overestimate future drought risk due to neglect of the dynamic regulatory effects of CO2 (SPEI(CO2)) on stomatal conductance of vegetation. An improved standardized precipitation evapotranspiration index incorporating physiological effects of CO2 was integrated with multi-model data of CMIP6. Using the Sen slope estimator, Mann-Kendall test, an exposure assessment model, and an additive decomposition method considering both climate and land use factors, drought trends, cropland exposure, and driving contributions were systematically quantified across nine major agricultural regions of China during the historical period (1961—2022) and future scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5). Results showed that during 1961—2022, grids exhibiting an increasing trend of drought accounted for 77% and 64% for SPEI and SPEI(CO2), respectively. For the future period (2023—2100), SPEI indicated a weak trend of drought intensification (Sen slope: -0.0013a-1), while SPEI (CO2) suggested a weak trend of humidification (Sen slope: 0.0011a-1), with the most significant differences observed in the northern arid and semiarid region (NASR) and Qinghai-Tibet Plateau (QTP). By the end of the 21st century, the national average cropland exposure (based on 0.25° grids) under SSP5-8.5 was estimated at 57km2/a for SPEI and 10km2/a for SPEI (CO2), with NASR showing the most significant reduction in peak exposure. Attribution analysis revealed that climate change dominated dynamics of future cropland exposure: increased precipitation in the near term (2021—2040) and mid-term (2041—2060) reduced risk of exposure, whereas rising temperatures in the long term (2081—2100) amplified exposure. The effect of CO2 suppressed evapotranspiration and reversed intensification of drought in major agricultural regions under SSP2-4.5 and SSP5-8.5 scenarios during 2081—2100.

Abstract:Drip irrigation is one of the irrigation methods with the most significant water-saving effects in modern agricultural irrigation. How to improve the reliability of drip irrigation pipe network systems while reducing project investment and operation costs is a key issue that needs to be addressed urgently in optimal design. Aiming at the characteristics of machine-pressure drip irrigation, a multi-objective optimal design mathematical model for machine-pressure drip irrigation pipe network systems was established with the objectives of minimizing the annual investment cost of the pipe network, the mean value of node surplus head, and the variance of node surplus head. An improved harmony search algorithm was proposed to solve the multi-objective optimization model, along with its methods and steps. In constructing the optimization model, the pipe network system was divided into units of levels, strips, and sections, making the established optimization model universal. Taking a machine-pressure drip irrigation project in Xinjiang as an example, this method was used to optimize its pipe network. Compared with the original design scheme, the optimized scheme reduced the annual investment cost of the drip irrigation system by 4.97%, the mean value of node surplus head by 16.84%, and the variance of node surplus head by 12.47%. The optimization results showed that this method can not only effectively reduce the investment cost of the pipe network system, but also significantly reduce the mean value and variance of node surplus head, thereby reducing the pressure deviation and failure frequency of the pipe network system and improving its reliability.

Abstract:In northern saline-affected irrigation districts, spring irrigation significantly increases soil moisture and reduces soil salinity. However, traditional spring irrigation quotas are relatively high, and the optimal timing for spring irrigation is difficult to determine. Therefore, optimizing spring irrigation patterns (quotas and timing) is crucial for enhancing agricultural water use efficiency and productivity. The research was conducted in the irrigation district south of the Yellow River in Ordos City, Inner Mongolia, during 2022—2023. Two spring irrigation quota levels (135mm and 180mm, denoted as W1 and W2) were established, along with three irrigation timing points: 18 days before sowing (T1), 12 days before sowing (T2), and 6 days before sowing (T3), totaling six treatments. Additionally, two indoor infiltration experiments were conducted to examine the effects of spring irrigation quotas and timing on soil infiltration and sunflower emergence rates. Using the HYDRUS model, the pre-sowing soil water-salt dynamics under different treatments and the relationship between pre-sowing water-salt conditions and various spring irrigation patterns were explored. It was found that higher spring irrigation quotas resulted in greater wetting front migration distances and cumulative infiltration volumes. By establishing a quantitative relationship between sunflower emergence rate and pre-sowing soil water-salt content, the optimal soil water content (SWC) and electrical conductivity (EC) were determined to be 0.254cm3/cm3 and 0.683dS/m, respectively. The HYDRUS model was employed to simulate water-salt transport under different spring irrigation patterns. Results indicated that the HYDRUS model effectively simulated water and salt dynamics during the pressurized infiltration and redistribution phases of spring irrigation. During the validation period, the root mean square error (RMSE), coefficient of determination (R2), and mean relative error (MRE) for SWC averaged 0.02cm3/cm3, 0.96, and 7.7%, respectively, while corresponding values for EC were 0.32dS/m, 0.90, and 9.9%. Under identical spring irrigation quotas, earlier irrigation resulted in higher pre-sowing EC and lower SWC. Compared with the 18days pre-sowing irrigation treatment, delaying irrigation by 6days and 12 days reduced the 2-year average EC in the tillage layer (0~30cm) by 10.1% and 16.8%, respectively, while increasing the average SWC by 3.4% and 6.2%. Under identical spring irrigation timing, compared with the low spring irrigation quota (W1), the high spring irrigation quota (W2) reduced the 2-year average pre-sowing EC in the plow layer by 13.4% and increased the average SWC by 4.5%. Based on HYDRUS model scenario analysis of pre-sowing soil water and salt content under spring irrigation regimes with irrigation quotas ranging from 90mm to 315mm and timing from 3 days to 18 days before sowing, optimization was conducted using optimal SWC and EC values. The optimal spring irrigation regime for this region was determined as follows: spring irrigation quota of 180mm with timing 18 days before sowing, or spring irrigation quota of 135mm with timing 15 days before sowing. These findings can provide theoretical support for agricultural water conservation and production in saline-affected irrigation areas of the Yellow River Basin.

Abstract:Ephemeral gully erosion is one of the main erosion modes on sloping farmland, which seriously damages the farmland resources, exacerbates the contradiction between people and food, and seriously hinders the sustainable economic and social development of the Loess Plateau. In order to explore the identification method and analyze the spatial-temporal distribution and development characteristics of ephemeral gullies in the Loess Plateau based on remote sensing images, ephemeral gullies in the Zhoutungou watershed of Yan’an City, Shaanxi Province were selected as the research object, combined high-resolution remote sensing images and deep learning image semantic segmentation model, the application effect of the U-Net and SegNet model in identifying ephemeral gullies and the spatiotemporal distribution were explored and the development characteristics of ephemeral gullies in the Zhoutungou watershed were cleared. The results showed that the SegNet model had excellent integrity and accuracy in ephemeral gully recognition, with model accuracy and recall no less than 82.59% and 91.12%, respectively. The ephemeral gully length and width between predicted and measured values had RMSE values of 6.78m and 0.50m, respectively, indicating that the model had an excellent recognition effect. Ephemeral gullies were mainly distributed in the southern region of the watershed with fragmented terrain, and there were few ephemeral gullies in the valley plain area. The distribution of ephemeral gullies had a clustered characteristic. Totally 215 ephemeral gullies disappeared, 938 ephemeral gullies (in-situ ephemeral gullies) with no significant changes in spatial location, and 1374 new ephemeral gullies were added from 2009 to 2021. Ephemeral gullies were mainly distributed between 25~40m in length and 1.0~1.5m in width. The average annual development rates of ephemeral gullies in length, width, area, density, dissection degree and head advance distance were 1.66m/a, 0.04m/a, 1.83m2/a, 4.94×10-5km/(km2·a), 5.45×10-6a-1 and 1.66m/a, respectively. The implementation of the gully land consolidation project, the development of oil and gas platforms, and the frequent occurrence of extreme rainfall events in recent years had greatly promoted the spatial distribution changes and morphological development of ephemeral gullies in the study area. The research result can provide an efficient and accurate method for identifying ephemeral gullies in hilly and gully areas of the Loess Plateau, and also provide a reference for soil erosion and slope and gully management in the Loess Plateau.

Abstract:Soil quality evaluation is a key basis for refined agricultural production and scientific land management, which is of great significance in guaranteeing national food security. To clarify the soil quality of arable land in loess hilly areas, Hancheng on the southern edge of the Loess Plateau was taken as the target area, totally 134 soil samples were collected from the soil surface layer (0～20cm), and 27 indexes covering soil physical, nutrient and environmental characteristics were measured, and then a minimum data set was constructed based on principal component analysis and Norm principle. The results showed that the soil in the study area was slightly alkaline (with a average pH value of 8.31), and its texture belonged to clay loam. The soil environment was at a mild ecological risk, with good environmental quality. The content of alkali-hydrolyzable nitrogen in soil nutrients was relatively deficient, the contents of organic carbon and available phosphorus were at a moderate level, and the contents of total phosphorus and available potassium were relatively rich. The minimum data set for soil quality evaluation in the Hancheng area on the southern edge of the Loess Plateau consisted of seven indicators: soil moisture content, specific gravity, capillary porosity, organic carbon content, zinc content, nickel content and coarse sand content. Among them, the organic carbon content had the largest weight in the soil quality evaluation indicators, that was, the organic carbon content was the key factor controlling the soil quality in this area. The mean value of the soil quality index (SQI-MDS) in the minimum dataset (0.522) and the mean value of the soil quality index (SQI-TDS) in the full dataset (0.537) differed slightly, and both belonged to the same grade in soil quality classification. The variation range and coefficient of variation of SQI-MDS were both higher than those of SQI-TDS, and the determination coefficient R2 of the fitting result between SQI-MDS and SQI-TDS were 0.812. Therefore, the soil quality assessment method based on the minimum data set had better applicability in this area and higher evaluation accuracy. When the semi-variogram was a Gaussian function, the prediction accuracy was the highest. The soil quality showed a certain distribution pattern in space. In the area close to the river, the higher the soil quality index was, the better the soil quality would be. The combination of the minimum data set and the soil index evaluation method can accurately, efficiently and comprehensively reflected soil quality, providing an approach to solve problems such as numerous soil indicators, high testing costs and complex calculations in the process of soil quality evaluation.

Abstract:Aerobic composting is effective for processing various organic wastes, such as kitchen scraps, domesticated animal manure, and agricultural straw, commonly found in rural areas. However, maintaining a stable small-scale household aerobic composting system outdoors during winter in cold regions poses significant challenges, and there is a lack of a suitable ventilation strategy. An aerobic composting system tailored for rural households in the cold regions was designed, integrating solar thermal storage and power supply, double chambers, and an insulated thermal storage tank. The temperature rises and composting characteristics during the aerobic composting process across different ambient temperatures were studied, while exploring optimal ventilation and odor reduction strategies. The results showed that the design effectively maintained stable temperatures in the composting chamber throughout the day and night, while also achieving preheating of the ventilation air, ensuring the stable operation of aerobic composting. The actual temperatures inside the composting chamber reached 21.9~45℃, 15.7~32.4℃, and 13.5~23.8℃ under ambient temperatures of 15~30℃, 0~15℃, and -25~0℃, respectively. A continuous ventilation rate of 25L/(m3·min) proved optimal during the first 11d, with maximum temperatures in the compost pile reaching 62.3℃, 56.4℃, and 55.2℃ at ambient temperatures of 15~30℃, 0~15℃, and -25~0℃, respectively. For low-temperature environments below 0℃ after the initial 11d, implementing a dynamic ventilation strategy with a rate of 25L/(m3·min) and maintaining oxygen concentrations between 14% and 17% proved beneficial. By the end of the 21d aerobic composting at an ambient temperature from -25℃ to 0℃, the germination index reached 89.89%. In addition, the aerobic composting process incurred no electricity costs. Straw effectively purified the ammonia in the exhaust from aerobic composting, achieving a removal rate from 80% to 95%. The findings can serve as a foundation for enhancing the stable operation of aerobic composting and improving the ecological environment for rural households in cold regions.

Abstract:Accurate temperature prediction in mushroom houses is crucial to ensuring the efficient industrial production of edible mushrooms. However, existing predictive models often lack generalizability when applied to mushroom houses located in different regions. Taking into account the disruptive effects of environmental changes caused by equipment in mushroom cultivation houses, and equipment operation state features was integrated to develop a temperature prediction model based on the temporal convolution network and long short-term memory model (TCN-LSTM). This model used TCN to extract local information along the temporal dimension, while LSTM captured the long-term dependencies of time series data. Compared with models that did not integrate device operating state features, the TCN-LSTM model that incorporated these features reduced the MSE by 17.1%, 35.7%, and 44.1%, and reduced MAE by 4.3%, 28.0%, and 38.0% for prediction horizons of 1hour, 2hours, and 3hours, respectively. This result indicated that incorporating equipment operating states had a significantly positive effect on the prediction performance. Compared with other shallow and deep learning models, the TCN-LSTM model achieved the best prediction accuracy for different prediction horizons, with R2 no less than 0.982, and both MSE and MAE no more than 0.57℃ for horizons up to 3 hours, satisfying the requirements of accuracy and duration in temperature predictions for mushroom room environmental control. This study employed transfer learning via pre-training and fine-tuning to adjust network parameters in small sample datasets, achieving rapid construction of temperature prediction models for mushroom rooms at different locations. The results indicated that for prediction horizons of 1hour, 2hours, and 3hours, the prediction models built using different locations as target domains achieved R2 values no less than 0.912, MSE values no more than 4.02℃, and MAE values no more than 2.01℃ in the test set. These results suggested that under small sample conditions, the temperature models constructed for different locations using transfer learning can achieve accurate temperature predictions at different steps.

Abstract:Hydroponics contributes to food security and sustainable agriculture. However, achieving high crop quality and yield through artificial light and temperature control increases energy consumption, while multi-factor interactions introduce regulatory uncertainties. Three hydroponic systems, nutrient film technique (NFT), deep-water culture, and aeroponics were evaluated under sole cropping and lettuce/radish intercropping (a quantity ratio of 2∶1) in a non-strictly controlled environment. Results demonstrated that intercropping significantly enhanced lettuce dry matter yield (up to 19.18%) and nitrogen use efficiency (up to 21.54%), while reducing nitrite content in both lettuce (up to 31.32%) and radish (up to 33.36%). Vitamin C and soluble protein contents were also notably increased. Among different intercropping hydroponic systems, the NFT exhibited the strongest interspecific competition (LEC=0.22). Additionally, intercropping led to a decrease in the total nitrogen (TN) content of lettuce and an increase in the TN, molybdenum (Mo), and calcium (Ca) contents of radish, with the highest improvement in Ca content reaching 31.94%. These changes may have balanced crop yield and quality by influencing carbon and nitrogen metabolism. The findings can provide a theoretical basis for optimizing intercropping in hydroponic vegetable cultivation.

Abstract:In order to develop an indicator film that is both pH-sensitive and antibacterial for monitoring the freshness of chilled mutoon, purple sweet potato extract microcapsules (PEM) were prepared. Using sodium alginate and low-acyl gellan gum (SAL) as the base material, with Ag NPs and microcapsules as enhancers. The effects of indicator films (SAL, SAL/Ag NPs, SAL/Ag NPs/PEM) were explored and applied to the monitoring of mutoon freshness. The results indicated that the wall material of PEM had a good encapsulation effect, enhancing the thermal stability of the core material and providing certain antioxidant capabilities. The addition of Ag NPs and PEM improved the mechanical properties, barrier properties, thermal stability, and antioxidant capacity of the single SAL film;specifically, SAL/Ag NPs/PEM2 exhibited the highest tensile strength (12.21MPa) and the lowest oxygen permeability (7.25g·mm/（m2·d）), with a compact and orderly cross-section. X-ray diffraction analysis showed that the crystallinity of Ag NPs was good, with distinct and sharp diffraction peaks, which remained unaffected when mixed with PEM. Fourier-transform infrared spectroscopy analysis indicated that the structures of the components of the indicator film were similar and had good compatibility. When the film was exposed to ammonia gas, its color underwent significant changes, gradually shifting from purple to green and finally to blue. Utilizing this property, the freshness of mutoon stored at 4℃ was monitored. Experimental results demonstrated that on the 4th and 8th days, the mutoon reached secondary freshness and decay states, respectively, corresponding to color changes in the indicator film SAL/Ag NPs/PEM2 from purple to green and blue. In summary, SAL/Ag NPs/PEM2 was highly sensitive to changes in mutoon freshness and can be used for monitoring its freshness.

Abstract:In view of the problems of unstable driving, low safety performance and poor obstacle crossing ability of the harvester chassis in the clayey and heavy soil environment in hilly and mountainous areas, combined with the actual needs of harvesting operations, the kinematics and passing performance of the self-propelled combine harvester under clay and heavy soil conditions were analyzed. Firstly, the structure and working principle of the crawler chassis were described, and the dynamics and kinematics models of the crawler chassis were established. Secondly, based on the centroid theory, the coupling force of the crawler chassis under driving conditions such as straight steering, longitudinal climbing, rolling climbing, crossing ravines and climbing ridges was analyzed, and the critical parameters affecting the passing performance of the harvester were obtained. At the same time, the simulation research on RecurDyn dynamics and driving performance was carried out. The simulation results showed that the longitudinal climbing angle was 30°, the maximum angle of the lateral climbing was 25°, the maximum height across the ridge was 380mm, and the maximum width of the crossing gully was not more than 800mm. Finally, the whole machine was tested for driving through three typical working conditions: longitudinal climbing, lateral climbing, crossing ravines and climbing over ridges. The test results showed that the chassis of the self-propelled combine harvester can drive stably under the conditions of longitudinal slope of 30°, side slope of 25°, gully width of 750mm, and ridge height of 350mm, which was consistent with the simulation results and verified the accuracy of the simulation. The research result can provide a theoretical basis and reference for the chassis design of the combine harvester of rhizome Chinese medicinal materials in hilly and mountainous areas.

Abstract:Steady speed control of agricultural machinery can improve operating quality and efficiency. To address the impact of farmland slope variations on the speed stability of unmanned operation agricultural machinery, a hybrid control method was proposed. This method included a hybrid controller composed of a slope-based controller and a proportional-integral-derivative (PID) controller. The speed of agricultural machinery was influenced by longitudinal forces, which were divided into two parts: one part was slope-related forces and conventional resistance, and the other was hard-to-estimate forces, such as sliding friction. For the first part, a slope-based controller was designed;for the second part, a PID controller was implemented. By combining these two controllers, the system can dynamically adjust the throttle opening and the brake master cylinder pressure, ensuring steady speed travel on sloping farmland. Simulation tests at a target speed of 7km/h demonstrated that the proposed controller maintained a stable speed, achieving a root mean square error of 0.13km/h and a mean absolute percentage error of 1.6%. Field tests on a practical experimental platform validated the method’s effectiveness, with results showing consistent control performance across varying slope conditions. The proposed controller demonstrated superior control performance. Experimental data verified that this method can achieve precise control of the agricultural machinery’s movement speed, meeting the stability requirements for agricultural operations.