| Apples planted in the natural environment need to be picked in the mature stage,and the workload is heavy.Under the background of the current shortage of agricultural labor force,it is urgent to develop intelligent agricultural picking equipment.Therefore,the study of apple picking robot is of great significance.At present,the technical bottlenecks restricting the performance improvement of apple picking robot mainly focus on image information acquisition,apple target recognition and tracking,positioning and control in complex environment.Accordingly,the research of this paper focuses on image fog removal,oscillating apple tracking recognition and servo control in the process of robot picking,and the main research achievements are as follows:(1)Aiming at the problem that it is difficult for apple picking robot to obtain image information under haze conditions,a fog removal algorithm based on boundary constraint and guided filtering is proposed.Firstly,the transmittance of the image is solved by combining boundary constraints with contextual regularization.Then,the transmittance is optimized by guided filtering.Then the color of the image is corrected by gamma changes.Finally,the dark channel theory model is used to restore foggy images.Compared with the He algorithm and Meng algorithm,it is shown that the algorithm in this paper has higher overall quality of fog-free images,and has a good effect on defogging images affected by light,so it is applicable to a wider range of scenes.(2)In order to track and recognize apple accurately and efficiently,this paper proposes a tracking and recognition method combining SURF feature affine transformation on the basis of normalized cross-correlation template matching.Firstly,the OTSU threshold method based on R-G color features was used to segment the initial image and identify the target fruit(i.e.,obtain the template image).Secondly,SURF feature extraction and matching are carried out between the template image and the next frame image,and eliminate the mismatched supporting points.Affine transformation is used to correct the template image and predict the target position.Finally,NCC matching is used to search for the best fruit matching location within 1.2times of the predicted location.The test results show that without template correction and apple position prediction,the running time of each frame is 0.59 s and 5.17 s respectively,while the running time of the proposed method is 0.41 s.Meanwhile,the tracking error is reduced.This method can significantly improve the tracking speed and efficiency,and also provides a reference for the tracking recognition of other oscillating fruits.(3)For eliminating the errors caused by locating and manipulator motion,this paper designs and implements a micro-control method of apple picking robot based on visual servo.Firstly,the camera calibration is carried out,and the visual servo control system is dopted IBVS approach by using the eye-in-hand control system.The jacobian matrix of the robot eye-hand relationship model is derived.Finally,the control theory was simulated and verified,and the results showed that the end-effector can be moved exactly to the desired position,ensuring the apple picking robot to complete the picking operation smoothly. |
PDF Full Text Request