基于特征波长的接触式作物叶绿素检测系统

Contact-based Crop Chlorophyll Detection System Based on Feature Wavelengths

摘要:

基于叶绿素分子在可见光和近红外光谱区间（波段400~1 000 nm）对光的吸收和反射的敏感特性，设计了一种接触式作物叶绿素检测系统，来实现作物叶绿素含量的无损、快速和准确检测。使用高光谱相机采集玉米叶部397～1003 nm反射光谱，并使用分光光度法萃取叶片叶绿素含量真值，开展叶绿素敏感响应波长筛选。经蒙特卡洛无信息变量消除(MCUVE)算法在10～50个特征波长内进行变量筛选，发现采用30个特征波长时具有最优的叶绿素含量检测能力，同时通过连续投影(SPA)算法进行特征波长筛选，2种算法共得到7个重合特征波长，又通过对波段和叶绿素含量进行相关性分析，剔除低相关性波段，最终得到6个特征波长。根据筛选出的特征波长对接触式图像传感器波段进行选型，设备的硬件主要包括传感器图像采集、主控制器、电源等模块，实现作物叶部近红外和可见光反射光谱数据采集、处理、显示和存储功能。开展传感器性能测试和田间应用测试，通过分析获取的多光谱图像的反射率构建叶绿素含量偏最小二乘(PLS)检测模型，验证集决定系数为 0.705;通过分析各植被指数与叶绿素含量的相关性，选取了相关性较高的归一化红边植被指数(NDRE)、绿红差值植被指数(GMR)和地面叶绿素指数(MTCI)进一步组合建模，检测模型精度提高到0.713，最终将模型嵌入系统实现了田间叶绿素含量快速检测，为作物长势分析提供了技术支持。

Abstract:

Based on the sensitive characteristics of chlorophyll molecules to light absorption and reflection in the visible and near-infrared spectral range （400~1 000 nm）, a crop chlorophyll detector based on a contact image sensor can be designed to achieve non-destructive, rapid, and accurate detection of crop chlorophyll content. Firstly, a hyperspectral camera was used to collect the reflection spectrum of cornleaves in the range of 397～1003 nm, and the true value of leaf chlorophyll content was extracted by using spectrophotometry. Nextly, the screening of chlorophyll - sensitive response wavelengths was carried out. The Monte Carlo uninformative variable elimination（MC - UVE）algorithm was used to screen variables within the range of 10 to 50 feature wavelengths, and it was found that using 30 feature wavelengths provided the optimal detection capability for chlorophyll content. Simultaneously, the successive projections algorithm（SPA）was employed for feature wavelength screening. The two algorithms yielded a total of seven overlapping feature wavelengths. Further, through correlation analysis between the bands and chlorophyll content, low-correlation band was eliminated, ultimately resulting in six feature wavelengths. The selected feature wavelengths were used to choose the bands for the contact image sensor. The hardware of the device mainly included sensor image acquisition, main controller, display, and other modules, which realized the functions of near - infrared and visible light reflection spectrum data acquisition, processing, display, and storage of crop leaves. Sensor performance tests and field application tests were conducted. By analyzing the reflectivity of the obtained multispectral images, a partial least squares detection model for chlorophyll content was constructed，with a coefficient of determination for the validation set of 0.697. By analyzing the correlation between various vegetation indices and chlorophyll content, the normalized difference red edge（NDRE）, green minus red（GMR）, and normalized difference red edge（MTCI）vegetation indices with higher correlation were selected for further combined modeling, improving the detection model accuracy to 0.706. The model was embedded into the system, ultimately achieving rapid detection of chlorophyll content in the field and providing technical support for crop growth analysis.