| With the development of computer vision,the role of visual robots in the production process is increasingly important,and the positioning and grasping of robots are widely used.In the industrial production field,it is inevitable to locate and grab weakly textured objects.At present,robot vision technology does not solve the problem of positioning and grabbing weakly textured objects well.Regarding the problem of weak texture target localization,this paper studies the binocular vision and robot positioning system of speckle projection.The positioning system inherits the advantages of the binocular robot and realizes the reconstruction and positioning of weak texture targets.The main research contents of this article are as follows:(1)The binocular vision and robot positioning system model of speckle projection is established.The model completes the three-dimensional reconstruction of weakly textured objects by projecting speckle.After the coordinate transformation of the hand-eye relationship,the robot can realize the three-dimensional positioning of the target.According to the principle of three-dimensional positioning,the test platform of the positioning system was constructed,and the corresponding selection of its hardware was completed.The measurement calibration model of the robotic binocular vision system was established.In order to avoid the problem of high calibration error caused by the low quality of the collected single image,the Zhang's calibration method was used to calibrate the binocular camera model twice in order to improve the binocular camera Calibration accuracy.The method of calibrating the hand-eye relationship of the manipulator is studied,and the two-step method of Tsai is used to complete the calibration of the hand-eye relationship.(2)The method and principle of edge detection are studied,and several typical edge detection methods are used to evaluate the edge detection effect of the target.Sobel edge detection method with better effect and higher positioning accuracy is selected as the target edge detection algorithm.In this paper,the collected spherical or bowl-shaped target images are fitted to the edges after edge detection in order to achieve accurate positioning of the target,and an improved fitting circle algorithm based on Hough transform is proposed.By using the improved circle fitting algorithm to fit the edge contour to the circle,the results show that the improved algorithm has better and more accurate fitting results,which makes the positioning accuracy higher.(3)The three-dimensional reconstruction of the weak texture target projecting speckle is studied,and the advantages and disadvantages of the global and local digital speckle algorithms are compared.The local digital speckle algorithm is adopted for the measured object.How to set the speckle parameters and the division of the size of the local area can better reflect the surface information of the object,and the rule between the number and size of the speckle and the projection effect is obtained.Aiming at the projected speckle target,the matching algorithm is introduced,and the local stereo matching algorithm flow of secondary guided filtering based on LOG operator weighting is designed.This algorithm reduces the influence of noise and halo introduced by speckle projection,improves the matching accuracy,and finally reconstructs the measured object.Experiments show that the object reconstructed by this method has good effect.After the three-dimensional information of the object is obtained by the camera,after the hand-eye conversion,the robot can locate the threedimensional target.(4)Use Visual Studio2015 + Open CV develop binocular vision and robot positioning system for speckle projection,and at the same time,combine the graphical user interface application development framework Qt to design man-machine interface.Positioning experiments were carried out on the two models to verify the feasibility of the positioning system and the positioning accuracy of the robot. |
PDF Full Text Request