| 摘要: |
| 摘要：【目的】在喷杆式喷雾机田间作业环境下，研究喷雾机喷杆的不同位置和不同高度对障碍物检测、识别和避障能力，以期为农业机械在大田作业中障碍物的检测和农机智能化控制提供新的思路。【方法】文章采用三维固态激光雷达，利用卡尔曼滤波算法进行北斗和激光雷达采集数据的融合，以确定最终喷雾机主体的坐标系位置，利用体素化网格法和欧洲聚类法对相邻采样周期内雷达采集的数据进行过滤，聚类和分割，以确定障碍物的绝对位置，在此基础上，通过相对坐标转换获得障碍物的相对位置和高度状态。【结果】试验结果表明：整个系统可以胜任喷杆式喷雾机的机械臂障碍物检测，该系统对障碍物的检测结果略有波动，但波动范围不大，这表明系统检测相对稳定。横向和纵向平均偏差率分别为3.93%和1.26%。20 m处垂直地面高度方向，平均偏差率为0.09%。【结论】障碍物检测系统能够有效进行静态障碍物的检测与识别，并且通过检测结果能够获得障碍物的位置和高度信息，从而指导机械臂进行有效避障。但是系统对动态障碍物不能进行有效检测与识别，也不能检测障碍物的硬度和厚度等信息。 |
| 关键词: 激光，障碍物检测，障碍物识别，避障 |
| DOI： |
| 分类号: |
| 基金项目:吉林省重点技术攻关项目 |
|
| Design and performance test of obstacle detection system for boom manipulator |
|
Zhang Mingyu, Qi Ruifeng, Wang Qiang, Liu Feng
|
|
Jilin Provincial Agricultural Machinery Research Institute
|
| Abstract: |
| Abstract: [Objective]Under the field operation by boom sprayer, the ability of obstacle detection, identification and obstacle avoidance in different positions and heights of lance boom were studied to provide new ideas for agricultural machinery in field operation and intelligent control.[Method] In this paper, three dimensional solid-state lidar and Calman filtering algorithm are used to fuse the data collected from Beidou and lidar to determine the position of the coordinates of the final sprayer body. By using the voxel grid method and the European clustering method, the radar data collected in the adjacent sampling period are filtered, clustered and segmented to determine the absolute position of the obstacles. The relative position and height of obstacles are obtained by relative coordinate transformation.[Results] The test results show that the whole system can be used to detect the obstacles of the boom of sprayer. The detection result of the system is slightly fluctuant, but the fluctuation range is not large, which indicates that the system detection is relatively stable. The horizontal and vertical average deviation rates were 3.93% and 1.26%, respectively. The average deviation rate is 0.09% in the vertical direction of 20 m. [Conclusion] The obstacle detection system can effectively detect and identify static obstacles, and obtain the position and height information of obstacles through the detection results, so as to guide the manipulator to effectively avoid obstacles. But the system can't detect and recognize the dynamic obstacles effectively, and can't detect the information of the hardness and thickness of the obstacles. |
| Key words: laser, obstacle detection, obstacle recognition, obstacle avoidance |