| Orchard plant protection automation and intelligent has become the core trend of modern orchard development.As for the characteristics of hilly and mountainous orchards with uneven landforms and roads,the crawler-type sprayer system is constructed,and the objective is to solve the problem of location and canopy information perception in unmanned operation of crawler sprayer.Therefore,this paper uses multi-line Lidar and Inertial Measurement Unit(IMU).Simultaneous Localization and Mapping(SLAM)and tree canopy information extraction scheme are studied.The main research contents include:(1)Construction of crawler-type sprayer system.According to the hilly orchard environment,the 3D structure and software system framework of the sprayer are designed,and the main components of the sprayer are selected.Based on this,Lidar and IMU are used as core environment sensing devices,and key sensor devices are selected,error calibration and sensor information synchronization are carried out,which provides the basis for sensors fusion.(2)Laser Inertial SLAM based on graph optimization.In order to solve the problem of occlusion of Global Navigation Satellite System(GNSS)signals in hilly and mountainous orchards,laser inertial tightly coupled SLAM technology is used to realize the estimation of location when GNSS signal fluctuates and construct the point cloud map at the same time.Finally,public data sets are used for testing.The results show that the proposed algorithm can obtain real-time pose information at a fast adult walking speed and construct a point cloud map.The constructed point cloud map clearly and accurately reflects the environment structure.(3)Multi-angle tree crown information extraction based on Lidar.In order to obtain real-time crown information,a multi-angle crown information extraction scheme is proposed.Firstly,the adaptive clustering method is used to detect the canopy target.Secondly,the improved alpha-shape combined with adaptive slicing method is used for3 D reconstruction of canopy point clouds at different growth stages.Based on this,a target tracking algorithm based on linear Kalman filter and NN algorithm is designed to track tree crown targets.Finally,the weight value of the tracked key position of the canopy volume is selected to get the final canopy volume value.In order to verify the effectiveness of the proposed algorithm,the 3D point cloud reconstruction method and the tree crown target tracking algorithm are simulated respectively.The simulation results show that compared with voxel method and platform body-convex hull method,the 3D point cloud reconstruction method in this paper can be better applied to the canopy with different crown density.The maximum total root mean square error rate of the multitarget tracking algorithm designed in this paper is less than 5.14%.(4)Hilly orchard field test verification.In order to verify the effectiveness and feasibility of the proposed algorithm,the laser inertial SLAM algorithm and the overall scheme of tree crown information extraction are tested and verified respectively in modern hilly orchard.The experimental results show that the laser-inertial SLAM algorithm of crawler sprayer has a good performance in orchard row and field,and can obtain accurate real-time pose and build a globally consistent map.In the orchard interrow scene,the average trajectory deviation is 0.24 m.In orchard field scene,the average trajectory deviation is 0.18 m.Meet the orchard real-time positioning requirements.The tree crown detection algorithm of the sprayer has a correct detection rate of 86.28% and can track the tree crown target in real time.Compared with the manual measurement results,the relative error rate of the obtained canopy volume ranges from-7.95% to+16.02%,realizing the real-time and accurate measurement of tree canopy volume. |
PDF Full Text Request