Application Research Of Binocular Vision In Workpiece Recognition And Positioning System

In recent years,the requirements for industrial automation production have been continuously improved,more and more visual technologies have been applied to the industrial field.In the vision system,the binocular vision system can not only acquire more scene information,but also quickly and easily sense the depth of 3D.The application advantage is obvious.Since the workpiece identification and positioning plays a very important role in the modern industrial production process,this paper designs a set of workpiece identification and positioning system using binocular vision,which is the key to camera calibration,workpiece detection and recognition,stereo matching and positioning.The technology is researched to realize automatic identification and positioning of workpieces,thereby further improving industrial automation.Firstly,based on a large number of related literatures,some of the system's skills are presented;through combining the imaging principle of the camera,establishing a nonlinear model,considering the accuracy and distortion of the parameters,the main research is Zhang calibration method.The corner points are extracted,the parameters and external parameters of the binocular camera are obtained,and the relative spatial positional relationship is obtained,thereby paving the way for positioning.Secondly,combining the difficulties of binocular vision technology,the main research is not only about feature extraction,but also aiming at recognition algorithms.By researching the image filtering and enhancement related algorithms,meanwhile,selecting the appropriate method for image preprocessing based on the characteristics of the scene image,an improved SURF-DAISY algorithm is proposed.This method combines the standard SURF algorithm with the improved enhanced rotation.The denatured DAISY operator performs feature matching.On this basis,the KD-tree priority search strategy is used to achieve fast matching,and the RANSAC algorithm is purified to achieve accurately recognition of the workpiece target.The experimental results show that the proposed algorithm can obtain more matching point pairs under the conditions of workpiece size scaling,position rotation,illumination difference and noise,really has a good matching effect.Compared with the standard SURF algorithm,the improved algorithm is more accurate and the time spent is greatly reduced,indicating that the algorithm is applied to the fast robust performance in workpiece recognition.Then,combined with the binocular stereo camera that has been calibrated,the method ofcalculating actual position and attitude is designed for the target object of this experiment.The approximate target area is determined for the identified workpiece target.Aiming at the industrial application environment,the problem of poor timeliness of Otsu algorithm in the process of workpiece segmentation is improved.Otsu is improved based on the improving particle swarm optimization algorithm to extract the target contour and centroid.In the system,due to the two-camera's not parrel,the algorithm of Bouguet can help to solve this problem,to regulate the polar line.The NCC optimization algorithm is used to realize fast stereo matching of the centroids in the left and right views,obtain parallax,and combine the three-dimensional reconstruction principle to obtain the position and posture of the workpiece target in space.Finally,according to the principle of 3D reconstruction,the CCD binocular camera and PC were used to design and build a binocular stereo vision system test verification platform.The accuracy of the calibration method was verified by analyzing the calibration results.At the same time,the MFC host computer is programmed to obtain a visual interface to identify different workpiece targets.The parallax method is used to solve the three-dimensional coordinate and attitude information of the workpiece,and the experimental data obtained is compared with the actual measured data.The universal applicability and efficiency of stereo matching and positioning algorithms.