Research On Spherical Fruit Recognition And Location Based On TOF Depth Information

With the research and development of robot technology,the research of fruit and vegetable picking robot has gradually become a research hotspot of domestic scholars,and the fruit recognition and positioning algorithm is the core of the picking robot.At present,the traditional fruit recognition and location algorithms are two-dimensional image algorithms based on color information,which are easily affected by the lighting environment,and have some limitations in dealing with overlapping,occlusion and near color background fruit recognition.Therefore,in order to solve the above problems,this thesis proposes an algorithm for fruit recognition and location from the perspective of fruit morphology by using depth images.This algorithm uses the depth camera to obtain the depth image of the fruit tree,calculates the gradient vector of each pixel point through the depth image,regards the gradient vector as the motion vector field,and calculates the divergence of the vector field.According to the principle of maximum divergence,the divergence center point is searched from the vector field.Then,the fruit center point is screened out from the divergence center point by using the difference between the fruit and leaf contour images,and the fruit boundary point is searched by using the eight direction search method with the fruit center point as the starting point,and the fruit image in the closed area formed by connecting the fruit boundary points in turn is introduced into the point cloud.Finally,according to the point cloud of the fruit image,the fitting circle of the target fruit is obtained by RANSAC algorithm,and then the size and three-dimensional spatial position of the fruit are obtained.The results show that the average correct recognition rate of the algorithm is84.8%,and the average recognition time of each fruit is 1.3s.The algorithm does not make use of the color features required in the traditional algorithm,but only uses the depth information in the depth image to identify and locate fruits,which can effectively overcome the drawbacks of the traditional algorithm affected by color,illumination and other factors.Because the algorithm does not need the color image information,it can not only realize the recognition and location of green fruits,but also realize the normal operation of the harvesting robot in the night environment.It provides technical support for fruits recognition and location algorithm research in complex environment.