基于单目视觉行间机器人导航线提取方法

A method of extracting navigation line of inter row robot based on monocular vision

针对农业移动机器人视觉导航路径提取受光照影响大,实时性差的问题,提出了一种利用农作物行特征辅助进行的视觉导航方案,实现机器人的田间自主导航。选用单目相机获取农田的环境信息,首先针对农田光照不均的情况提出了YCr Cb亮度均衡法,然后针对实时性差的问题提出了OTSU(大津法)融合遗传退火算法的图像二值分割算法,并利用Hough直线变换检测,实现农田导航路径的快速提取。视觉导航实验结果表明,所提出的光照平衡算法较直方图均衡法实时性提升25.7%,优化后的OTSU法二值化速度提升了26.5%,在双边缺植情况下,导航线平均偏移量为2.3°,符合农田导航的实用性和实时性要求。

In order to solve the problem that the extraction of visual navigation path of agricultural mobile robot is greatly affected by light and has poor real-time performance, a visual navigation scheme assisted by crop row features is proposed to realize the robot’s autonomous navigation in the field. A monocular camera is selected to obtain the environmental information of farmland. First, YCrCb brightness equalization method is proposed for uneven illumination of farmland. Then, aiming at the problem of poor real-time performance,an image binary segmentation algorithm based on OTSU(Otsu method) and genetic annealing algorithm is proposed, and Hough line transformation detection is used to achieve rapid extraction of farmland navigation path. The visual navigation experiment results show that the real-time performance of the proposed light balance algorithm is improved by 25.7% compared with the histogram balancing method, the binarization speed of the optimized OTSU method is improved by 26.5%, In the case of bilateral lack of planting, the average deviation of the navigation line is 2.3 °, which conforms to the practicality and real-time of farmland navigation.