基于机器视觉的机耕船障碍检测及避障策略研究  

Research on obstacle detection and obstacle avoidance strategy of tractor boat based on machine vision

作　　者：易飞 商锦萍 李静 YI Fei;SHANG Jinping;LI Jing(Yiwei Power Co.,Ltd.,Jingmen 448000,Hubei,China;Wuhan College of Bioengineering,Wuhan 430415,Hubei,China;Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming 650500,Yunnan,China)

机构地区：[1]亿纬动力有限公司,湖北省荆门市448000 [2]武汉生物工程学院,湖北省武汉市430415 [3]昆明理工大学机电工程学院,云南省昆明市650500

出　　处：《农业装备与车辆工程》2023年第4期157-161,181,共6页Agricultural Equipment & Vehicle Engineering

摘　　要：随着农村劳动力转移城市,传统的劳作也逐渐被先进的劳动生产方式替代,机耕船的制造厂家在考虑如何将机耕船进行高智能性、高工作效率的升级,实现无人驾驶。对机耕船障碍检测和避障策略进行了研究分析,选用基于OpenCV的BM和SGBM算法来实现双目立体匹配测距,通过实验对比两种算法的实际测距效果,BM和SGBM算法测距数据误差分别控制在0.867%和1.23%以内;通过构建样本库训练分类器,得到稳定的障碍物识别模型;选取人工势场法作为研究的基础,提出了一种改进的人工势场法,使机耕船顺利跳出局部极小值点,顺利避开了障碍物抵达目标终点。With the transfer of rural labor force to the city,the traditional labor is gradually replaced by advanced labor production mode.The manufacturer of machine tiller is considering how to upgrade the machine tiller with high intelligence and high work efficiency to achieve unmanned driving.The obstacle detection and obstacle avoidance strategies of the machine tiller are studied and analyzed.BM and SGBM algorithms based on OpenCV are selected to realize binocular stereo matching ranging.The actual ranging effects of the two algorithms are compared through experiments.The ranging data errors of BM and SGBM algorithms are controlled at 0.867%and 1.23%respectively.A stable obstacle recognition model is obtained by constructing a sample base to train the classifier.The artificial potential field method is selected as the basis of the research,and an improved artificial potential field method is proposed,so that the machine cultivator can jump out of the local minimum point smoothly,avoid obstacles smoothly and reach the target end point.

关 键 词：双目立体视觉 图像处理 立体匹配 人工势场法 OPENCV 

分 类 号：S222[农业科学—农业机械化工程] TP391.4[农业科学—农业工程]