Design Of Obstacle Avoidance System Of Orchard Lawn Mower Based On Binocular Vision Technology

With the steady development of agricultural industry structure,the production scale of fruit industry continues to expand.The orchard overgrown with weeds as the fruit trees grew As an important machine for cutting grass in orchards,the lawn mower needs to avoid obstacles such as rocks and branches of different sizes and shapes on the ground.In addition,the row spacing of fruit trees is small and the ground is rugged.Traditional lawn mowers cannot operate in this environment.Therefore,this paper designs a system that combines the obstacle depth information calculated by binocular vision system and makes the lawn mower execute different obstacle avoidance strategies in different distance areas.Obstacle detection is the primary condition for the mower to avoid obstacles successfully.In order to obtain abundant obstacle information,this paper adopts binocular stereo vision technology to detect the obstacle in real time through camera calibration,image preprocessing,stereo matching and depth measurement,and accurately calculate the relative distance between the obstacle target and the lawn mower,so as to improve the safety of the operation of the lawn mower.By analyzing the research status of binocular vision technology and obstacle avoidance algorithm at home and abroad,the overall technical route of the orchard mower obstacle detection system was designed.The camera calibration was completed by ZED binocular camera,and the binocular stereo vision experiment platform was built.In the application environment of this paper and taking into account the lens distortion,Zhang Zhengyou calibration method was adopted,and the precise camera internal and external parameters were obtained through nonlinear optimization.The acquired obstacle image was preprocessed,and the different filtering denoising algorithm and segmentation were performed.The comparative analysis of the algorithm selected the Gaussian filtering,histogram equalization and K-means clustering algorithm suitable for processing the orchard environment image to segment the obstacles.The segmented image was stereo-matched,and the classic SIFT and SURF stereo matching were compared and analyzed.The algorithm selected the SURF matching algorithm which takes into account the accuracy and speed according to the characteristics of obstacle recognition and location in the orchard environment.At the same time,the RANSAC algorithm was used to remove the mismatched points to obtain exact matching.Using the disparity map and camera parameters to measure the obstacle depth,the obstacle detection and relative distance acquisition task of the lawn mower was completed.Aiming at the problem of unreachable target and local minimum existing in traditional artificial potential field method,the path planning of lawn mower was completed.In order to verify the real-time and stability of the obstacle detection and distance measurement algorithms in this paper,four sets of depth measurement experiments were carried out on different orchard obstacles at different distances.When the measurement distance is less than 6000 mm,the measurement error is within 550 mm,and the measurement accuracy is high,the average positioning time is 2.985 s.It shows that the binocular vision system can meet the needs of most mowing operations in the orchard environment.The improved artificial potential field method was used to simulate the obstacle avoidance in different situations.The simulation results show that the algorithm has high real-time and security.