Research On Workpiece Recognition Technology Based On Binocular Vision

Binocular vision can gain the spatial position information of the target workpiece,which is beneficial to enhance the accuracy of workpiece recognition and the precision rate of workpiece positioning.This article aims to set up a workpiece recognition frame based on binocular vision.Through camera calibration,image de-noising,image matching and so forth,and then the spatial position information of the target workpiece can be obtained to realize the recognition and positioning of the workpiece.This article mainly studies the key technologies of image de-noising,matching and recognition positioning of the procedure.First of all,aiming at the classical wavelet thresholding de-noising,the hard threshold function will give rise to image fault after de-noising,and the soft threshold has the issue of constant deviation.In this article,a new progressive semisoft threshold function is put forward,and the analysis of continuity and deviation are carried out.It is proved that this function can take into account the advantages of soft and hard threshold.At the same time,a new adaptive threshold formula is provided,The effectiveness of the improved wavelet threshold function is verified by calculating the de-noised image mean square error,signal-noise ratio and other indicators.Secondly,the shortcomings that the SURF algorithm relies heavily on the Hessian matrix to identify feature points,and only performs unidirectional Euclidean distance matching when performing feature point matching.In the first step,bidirectional Euclidean distance is adopted for matching,which can eliminate portion mismatching points quickly and accurately.Then RANSAC method is introduced to assume the mathematical model of accurate matching features,and eliminate the "exterior points" that do not conform to the mathematical model,so as to improve the matching precision and lay a solid foundation for the next step of three-dimensional information acquisition of workpiece.Simulation results verify the effectiveness of this article's algorithm and eliminate portion SURF algorithm mismatching points.Based on the image matching results,optical parallax was obtained from the pixel values of the target workpiece on the left and right images,and the location information of the target workpiece was ultimately calculated.The simulation results show that the distance error of the method in this article is small,which is appropriate for the recognition and positioning of workpiece in industrial manufacture.