Image Matching And Location Of Common Workpiece Based On Binocular Stereo Vision

With the continuous development and progress of intelligent recognition algorithms and machine learning technologies,machine vision technology has attracted people's attention.The image matching and positioning technology based on binocular stereo vision is one of the directions for the development and application of machine vision technology.At present,binocular stereo vision technology has been applied in such popular areas as workpiece recognition,vehicle detection,and virtual reality.Compared with monocular vision,binocular vision has obvious advantages,that is,it can flexibly acquire stereoscopic information of target objects in various environments.In this paper,the stereo camera matching of the binocular camera and the basic edge features are deeply studied.The main technologies involved are: the calibration of the binocular camera,the identification and positioning of the target workpiece,and the threedimensional reconstruction.First of all,realize the camera calibration.Because the traditional calibration method can not realize the stability and high precision at the same time,this paper proposes an algorithm based on the combination of Hough and chaotic particle swarm.The experimental results show that the algorithm has good stability and high calibration accuracy.The accuracy is much higher than that of other traditional calibration algorithms.Then,determine the stereo matching method for the workpiece.An optimization method of shape context is proposed.This method essentially uses the information of the histogram to obtain candidate matching points by optimizing the shape context.The rough matching is performed first,then weighted and then processed with the shape context.The similarity measure completes a fine match.Finally,using RANSAC algorithm to eliminate the mismatched pairs that affect the matching accuracy.Finally,the experimental verification of the shape context optimization algorithm is carried out,and it is concluded that the method does improve the accuracy of the target image matching.Finally,a parallel binocular vision model is used to perform mathematical analysis on the key grab points of the target workpiece,and the 3D model of the grab points is reconstructed.Then obtain the angle between the long axis of the workpiece and the horizontal axis of the image,and use this angle to determine the orientation of the workpiece.Finally,the workpiece is captured.Experiments show that this method has certain feasibility.