Research On Rice Seedling Belt Recognition System Based On Lidar

With the increasing development of smart agriculture,precision agriculture has become one of the research hotspots of modern agriculture.Among them,the automatic navigation technology of agricultural machinery is the key link to realize precision agriculture.In the automatic navigation process of agricultural machinery,the identification and positioning of crop seedling belts have a great influence on the navigation effect and operation accuracy of agricultural machinery.Use the seedling belt recognition technology to plan the driving path in the drivable area,that realize the autonomous low-damage seedling operation of field management machinery,which is of great significance to promote the development of precision agriculture.Therefore,this paper uses the high gap field management machine as the experimental platform to design a seedling zone identification system based on lidar and carry out the research on rice seedling zone identification.The main research contents are as follows:(1)A three-dimensional point cloud information acquisition device is designed based on two-dimensional lidar.Based on the two-dimensional lidar,this research designed a threedimensional point cloud information acquisition device consisting of a lidar,a threedimensional rotating pan/tilt,an inertial measurement module and a control module.Established a three-dimensional rotating pan-tilt movement model,and realized the conversion of point cloud data from Cartesian coordinates to three-dimensional rectangular coordinates.A method for automatic correction of the three-dimensional PTZ is proposed,which solves the problem of the angle deviation between the collecting device and the driving direction of the agricultural machinery during the operation.The 3D point cloud information acquisition device expands the range of lidar scanning and recognition while controlling low cost,and controls the lidar to perform pitch,roll and rotation actions,which can realize 3D data collection of the measured object.(2)In order to solve the problems of laser radar scanning offset and improve the accuracy of trajectory recognition,this research proposes an extremum point clustering seedling zone recognition algorithm based on the offset compensation model.In this paper,rice is used as the identification object of the seedling zone,and the point cloud line characteristics of the rice are obtained by lidar,and the rice seedling belt model is established.Using extreme value detection algorithm to obtain the rice seedling zone clustering elements,and use the offset compensation model combined with the improved K-means clustering algorithm to perform offset correction clustering.Finally,the least square method is used for linear fitting to realize the trajectory identification of the travelable area of the rice seedling belt.Through static field experiments,the maximum lateral deviation of the X-axis is 16 mm,and the maximum deviation of the median point is 6 mm.(3)Developed a rice seedling band recognition system based on lidar.In order to realize the automatic navigation and driving of low-damage seedlings of field management machinery and improve the accuracy of seedling belt track recognition.The three-dimensional point cloud information acquisition device is used to obtain the point cloud data of the rice seedling belt,and the trajectory parameters of the rice seedling belt are obtained by using the extreme point clustering seedling belt identification algorithm of the offset compensation model.It can provide theoretical parameters for the straight-line,low-injury autonomous driving of the high gap field management machine.At the same time,in order to realize the visualization of data collection and human-computer interaction,the Miao Belt Recognition and Monitoring System was designed.(4)Completed the field test of the rice seedling band identification system based on lidar.Using the high ground clearance field management machine as the experimental platform,the seedling belt recognition system was mounted on it,and the field static data collection and the Tanaka dynamic linear navigation path tracking experiment were carried out.Before entering the field,keep the field management machine still,and obtain a maximum lateral deviation of30 mm through seedling belt identification,and a maximum deviation of 13 mm at the midpoint.After entering the field,in the dynamic linear path tracking test,the maximum lateral deviation was 33 mm,the average error was 18.5 mm,the standard deviation was 2.02 mm,and the seedling damage rate was less than 0.8%.The results prove that the seedling belt recognition system can meet the needs of low-injury seedling automatic or assisted driving operations.