无人农机障碍物检测系统设计  

Design of Obstacle Detection System for Unmanned Agricultural Machinery

作　　者：张淼 李晋阳 赵湛[1] Zhang Miao;Li Jinyang;Zhao Zhan(Key Laboratory of Modern Agriculture Equipment and Techenology,Ministry of Education,Jiangsu University,Zhenjiang 212013,China)

机构地区：[1]江苏大学现代农业装备与技术教育部重点实验室,江苏镇江212013

出　　处：《农机化研究》2025年第6期99-105,共7页Journal of Agricultural Mechanization Research

基　　金：江苏省农业科技自主创新基金项目[CX(22)3090];江苏省重点研发计划(现代农业)项目(BE2018324);镇江市重点研发计划(现代农业)项目(NY2022008)。

摘　　要：设计了基于RTK-GNSS、VLP-16激光雷达和姿态传感器的障碍物检测与定位系统,具有田间障碍物检测、插秧机位姿获取等功能。系统以久保田2ZGQ-6D(NSPU-68CMD)型号插秧机为实验载体,硬件上由车体位姿采集模块、点云数据采集模块和系统供电模块组成,软件上由坐标转换模块和障碍物检测模块组成。其中,车体位姿采集模块可获取插秧机作业田块的经纬度坐标和车体姿态信息;坐标转换模块可根据车体姿态和经纬度坐标选定插秧机的作业区域,建立插秧机的作业田块坐标系;点云数据采集模块可获取插秧机作业环境的点云信息;障碍物检测模块可根据环境点云信息,通过直通滤波、高斯滤波、地面点云分割、欧式聚类、OBB包围盒拟合和障碍物状态估计计算障碍物的定位坐标,判别障碍物的运动状态;系统供电模块实现障碍物检测系统的稳定供电。在农田环境进行障碍物检测试验,结果表明:系统对障碍物的定位略有波动,障碍物横向移动时,最小偏差为3.0 cm,最大偏差为14.3 cm,平均偏差为6.8 cm;纵向移动最小偏差为1.5 cm,最大偏差为16.0 cm,平均偏差为7.9 cm;在设定的障碍物检测区域内,对静态障碍物和动态障碍物的检测准确率可达到90%。系统能够对田间障碍物进行有效识别和准确定位,但无法对障碍物的类别进行判断。The obstacle detection and positioning system based on RTK-GNSS,VLP-16 lidar and gesture sensors was be designed,which can realize the function of field obstacles detection and seedlings.The system was based on Kubota 2ZGQ-6D(NSPU-68CMD)type seedlings as the experimental carrier.The hardware was composed of the car body position collection module,point cloud data collection module and system power supply module.The software was composed of coordinate conversion modules and obstacles detection modules.Among them,the vehicle pose acquisition module can obtain the longitude and latitude coordinates of the rice transplanter operation field and the vehicle pose information;The coordinate conversion module can select the operating area of the rice transplanter based on the vehicle posture and latitude and longitude coordinates and establish the coordinate system of the rice transplanter's operating field;The point cloud data collection module can obtain point cloud information of the working environment of the rice transplanter;The obstacle detection module can calculate the positioning coordinates of obstacles and distinguish their motion status based on the environmental point cloud information through direct filtering,Gaussian filtering,ground point cloud segmentation,Euclidean clustering,OBB bounding box fitting,and obstacle state estimation;The system power supply module realizes stable power supply for obstacle detection system.The obstacle detection experiment was conducted in a farmland environment,and the results showed that the system's positioning of obstacles fluctuated slightly.When the obstacle moved horizontally,the minimum deviation was 3.0 cm,the maximum deviation was 14.3 cm,the average deviation was 6.8 cm,the minimum deviation for vertical movement was 1.5 cm,the maximum deviation was 16.0 cm,and the average deviation was 7.9 cm;Within the designated obstacle detection area,the detection accuracy for static and dynamic obstacles can reach 90%.This system can effectively identify and accurately locat

关 键 词：无人农机 障碍物检测 RTK-GNSS 激光雷达 欧式聚类 

分 类 号：S252[农业科学—农业机械化工程]