自主导航农业车辆的全景视觉多运动目标识别跟踪

Multiple Moving Objects Tracking Based on Panoramic Vision for Autonomous Navigation of Agricultural Vehicle

摘要:

为提高自主导航农业车辆导航路径的准确性和行驶作业的安全性，提出自主导航农业车辆的全景视觉多运动目标识别跟踪方案。该方案采用全景视觉进行无盲区的多运动障碍目标的检测，并解决了多运动目标跟踪中遮挡重叠的问题。首先系统将多目相机采集的图像拼接成全景图像，采用分段图像的改进核函数算法对运动目标进行快速自动检测跟踪；其次采用基于路径预测的粒子滤波算法进行多运动目标跟踪并解决遮挡重叠的问题。通过试验表明：采用改进的核函数目标快速跟踪算法，与传统核函数跟踪算法相比，减少系统内存消耗66.8%，加快运算速度35.63%；采用基于路径预测的粒子滤波多目标跟踪算法，在多运动目标遮挡重叠的情况下，平均提高运动目标跟踪成功率39.5个百分点，算法平均耗时0.78s。

关键词:农业车辆 自主导航 全景视觉 多运动目标 核函数 粒子滤波

Abstract:

In order to improve the accuracy of the navigation path and satisfy the safety of driving for autonomous navigation of agricultural vehicles, a method of detecting and tracking multiple moving objects was proposed based on panoramic vision. Panoramic vision possessed the advantages of non blind area detection and the improved algorithm solved the problem of the overlap in multiple moving objects tracking. Firstly, multi-vision images were acquired to stitch panoramic images, the improved kernel function algorithm based on segmented image was used to detect and track the moving object automatically and rapidly. Secondly, the particle filter algorithm based on path prediction was used to track multiple moving objects and solved the overlap problem. Compared with the traditional kernel function algorithm, experiments showed that the memory consumption was reduced by 66.8% and the algorithm speed was increased by 35.63%. Multiple moving objects detection using the particle filter algorithm based on path prediction could take averagely 0.78s to detect moving obstacles, and the success rate of moving objects tracking was increased by 39.5 percentage points under the condition of overlap in multiple moving objects.

Key words:Agricultural vehicle Autonomous navigation Panoramic vision Multiple moving objects Kernel function Particle filter algorithm