| In recent years,autonomous navigation technology for orchards has developed rapidly.Many experts and scholars have carried out in-depth research on autonomous navigation technology for densely planted and sparsely planted orchards,but the research on navigation technology for trellis cultivation mode is insufficient.Structured scaffolding orchard,proposed a multi-sensor combination heading fusion positioning method,fast line-up and target perception method,and carried out autonomous navigation robot system development and navigation performance test based on this,providing for the realization of autonomous navigation of the scaffolding orchard Equipment and technical support.The main research content and work of this thesis include the following aspects:(1)The environment of the scaffolding orchard in southern Jiangsu was investigated,and the characteristics of the scaffolding orchard scene and the special requirements of autonomous navigation were summarized.Based on this,the complex scene perception method and autonomous navigation system of the scaffolding orchard based on the combination of 2D-3D information were designed.Design the upper and lower computer navigation hardware platform based on NVIDIA Jetson TX2-STM32 microcontroller and the distributed autonomous navigation software system based on ROS,and complete the sensor selection,calibration and coordinate system 1,combined with the crawler chassis to complete the multi-sensor combined autonomous navigation system Development.(2)Carry out autonomous positioning research in the scaffold orchard scene according to the autonomous navigation scheme.In order to obtain the navigation target in the lidar point cloud,the obstacle noise and outlier noise are eliminated through DBSCAN clustering,and the distance threshold eliminates the interference points of the non-operating line to obtain navigation Target point,but the test found that there is a problem of sharply reduced positioning accuracy under large pose deviations.Aiming at this problem,a positioning method of electronic compass and lidar heading information fusion is designed and related experiments are designed to verify.The results show that the average heading information fusion The lateral error rate is 3.3%,and the average heading error rate is 6%,which meets the positioning needs of autonomous navigation.(3)Through the analysis of the navigation path planning under the scaffolding,a threesegment navigation path planning method is designed for the beginning of the line-between the lines-the end of the line: Aiming at the problem of the line beginning of the line,it is proposed to classify and trigger the vehicle body pose state according to the different thresholds of the pose deviation dual index Corresponding on-line trajectory program,and achieve the goal of fast online with the solved optimal on-line angle;the center line of the left and right tree rows is the navigation path between the robot rows and the identification method of the fruit tree row tail.(4)In order to ensure the safety and stability of autonomous navigation,research on target perception in scaffolding scenes is carried out.The target perception system under the scaffold includes the recognition of obstacles and the extraction of contour information.In this paper,a depth camera is used to obtain 3D containing color information and depth information.Point cloud data is filtered through straight-through filtering and statistical filtering to filter out complex background,scaffolding surface and outlier noise.The region growth segmentation algorithm separates ground and above-ground types.Above-ground types include grape stems,pillars and large obstacles,Use the OBB target encircling algorithm to obtain the target contour size,realize the distinction between the navigation target and the obstacle target according to the different aspect ratio,complete the recognition of the obstacle target and the contour extraction and design related test verification.(5)According to the software and hardware platform design related verification tests,the recognition accuracy of the line tail line was verified in a simulated scaffold orchard environment.The results showed that the average accuracy rate of the line tail line recognition algorithm was96.67 when the heading deviation was 0 and-π/6.% And 63.33%,the line-end line recognition algorithm can achieve better recognition results when the heading deviation is small.The performance test of the fast-on-line method through the autonomous navigation system shows that at a constant speed of 0.3m/s,an initial lateral deviation of 1.4m and an initial heading deviation condition of-π/4,-π/18,0,π/18,Under the condition of π/4rad,the online time is 6.11,7.15,7.46,7.74,8.9s,and the online distance is 1.357,1.367,1.387,1.383,1.403 m.The accuracy test of interline autonomous navigation is carried out by the autonomous navigation system.The results show that at a constant speed of 0.3m/s,the average lateral deviation of the inter-line autonomous navigation system is between 2.73-3.14 cm,and the average standard deviation is 3.36 cm. |
PDF Full Text Request