Research On Fruit Recognition And Location Picking Of Kumquat Based On Binocular Vision

At present,most of the mechanical arm and binocular vision are applied in industry.In order to further improve the application field of mechanical arm and vision technology,this topic applies the mechanical arm and machine vision technology to the agricultural automatic picking equipment,so as to realize the automatic picking of agricultural products and liberate the labor force.Therefore,this paper studies the fruit recognition and picking of Kumquat based on binocular vision.The main research contents and conclusions are as follows(1)The binocular vision system and camera parameter calibration are studied.Based on the principle of binocular vision imaging,the binocular stereo vision system of Kumquat picking arm is built.By analyzing the camera calibration method,the binocular stereo vision system built by Zhang Zhengyou calibration method is selected,and the camera calibration experiment is carried out by using MATLAB software.The internal and external parameters of the binocular stereo camera are accurately obtained,and the calibration results are analyzed.(2)The fruit recognition algorithm of Kumquat was studied.Firstly,the research of image preprocessing is carried out,and several common filtering methods are compared through experiments.Finally,median filtering is used to filter and denoise the image according to the experimental results.Secondly,the segmentation method of Kumquat image is studied,and three different segmentation methods of Kumquat image are compared through experiments,and K-means clustering segmentation algorithm is selected as the segmentation algorithm of Kumquat image Law.By comparing the common edge detection algorithms,Canny algorithm is selected to identify and locate the fruit of small kumquat without occlusion.Then the least square ellipse fitting method is proposed to calculate the center of picking point of small kumquat fruit.The experimental results show that the contour fitted by the least square ellipse is consistent with the real fruit contour The average recognition error is 1.38%.Finally,a convex hull detection method based on volume wrapping algorithm is proposed to extract the fruit contour of Kumquat under occlusion;then a method to retain the real contour and remove the false contour segment is proposed,and the center and radius of the target fruit are obtained by using the random three-point circle determination method on the real contour;the experimental results show that the method of random three-point circle determination on the real contour of Kumquat can locate the kumquat The average positioning error of orange is 1.0 mm～2.5 mm.(3)The spatial location of kumquat fruit was studied.Firstly,the purpose of stereo correction principle of image is analyzed,and then the Bouguet algorithm is used for stereo correction experiment.Secondly,stereo matching is proposed by using the centroid of kumquat fruit as the unique constraint to obtain the information of target matching point pairs.On this basis,the centroid of kumquat fruit is matched,and finally the centroid matching location of Kumquat is tested The experimental results show that the error of centroid matching in x-axis direction is-0.6～2.0 mm,in Y-axis direction is-1.0～1.8 mm,and in z-axis direction is-1.5～1.5 mm.(4)The kinematics analysis of picking arm is studied.In order to understand the reliability and stability of Kumquat picking arm,the operation process of Kumquat picking arm was simulated and designed.Firstly,the D-H parameter method is used to establish the mathematical model of the small kumquat picking arm,construct the forward and inverse kinematics equations,and solve the kinematics with the analytical method;then the kinematics simulation experiment of the picking arm is carried out by using the rotbotics toolbox of MATLAB software to verify the correctness of the forward and inverse kinematics solution;finally,the operation process of the small kumquat picking arm is analyzed The simulation of trajectory planning is carried out in this paper.The simulation results show that the velocity,acceleration,angular velocity and angular acceleration curves of each joint are smooth.(5)The system platform of Kumquat picking robot was built and the picking experiment was carried out.The system is mainly composed of binocular camera and Esther manipulator.Before the experiment,according to the theoretical analysis,the eye in hand vision system mode is adopted,and then the hand eye calibration is completed to obtain the transformation matrix from the camera coordinate system to the end effector coordinate system of the small kumquat picking arm.Then,the picking and grabbing experiments are carried out on this basis,and the effectiveness of the method is verified by the experimental results.