Design of Point Cloud Acquisition System for Farmland Environment Based on LiDAR

Abstract:

Accurate perception in farmland environment is the premise to realize the obstacle avoiding in autonomous navigation for agricultural machinery. Stable and reliable environmental data acquisition system is the necessary condition for accurate perception. A point cloud acquisition system was designed for farmland environment based on LiDAR, which can realize the stable and reliable acquisition of farmland environment point cloud and the position and attitude of agricultural machinery. A multisensor data acquisition software was designed, which can achieve accurate and consistent global point cloud data acquisition. The system was composed of point cloud data acquisition module, vehicle position and posture acquisition module and data fusion module with tractor as mobile carrier. Among them, point cloud data acquisition module can acquire the point cloud data of surrounding environment and solve the problem of close blind area; vehicle position and posture acquisition module can acquire realtime agricultural machinery position and posture information; data fusion module can receive and integrate the environmental point cloud data and vehicle position and posture data, and then obtain the point cloud data after compensation. The system realized online collection of sensor data, time synchronization, spatial registration, data realtime display and storage. Point cloud acquisition experiments under farmland environment were carried out. The results showed that the acquisition system had good outdoor working stability. The online typical frame loss rate was not larger than 1%, while the offline typical frame loss rate was not larger than 0.47%, which can meet the requirements of farmland point cloud data acquisition. In order to analyze the data quality of point cloud collected by the system, the ground point clouds were filtered by straightpass filtering respectively by using the point cloud after compensation and the original point cloud. The results showed that the point cloud after compensation contained only a small amount of ground point cloud after filtering, which can be used as reliable data for obstacle avoidance in autonomous navigation of agricultural machinery.