Study On Workpiece Recognition, Based On Machine Vision

Workpiece recognition is an indispensable part of sorting,grabbing and other follow-up operations in industrial production line.Aiming at the disadvantages of traditional artificial recognition,machine vision is applied to the automatic recognition of the target workpiece,to solve the problem of different workpiece classification.In this paper,image processing,feature extraction algorithms and software programming techniques are studied.Firstly,we take common workpieces as research objects and preprocess their images.An algorithm using median filter combined with morphological opening-closing filter is applied to image denoising,and achieves a better edge-preservingsmoothing effect.In order to make some of the details of the images easier to identify,we adopt the method of top-hat and black-hat operation cascade to enhance the workpiece image contrast.These measures lay the foundation for the subsequent feature extraction and image matching.We analyze main workpiece recognition methods,and compare the advantages as well as disadvantages of various algorithms.In the process of corner detection,we introduce scale space and corner purification to improve traditional Fast corner detection algorithm,we describe the corners with SURF descriptor.We choose KNN algorithm as similarity measurement,LSH index in high-dimensional vector space to search,to realize the initial match of the feature points;the RANSAC algorithm is used to eliminate the false matching.The matching accuracy is increased obviously compared with the original SURF detection algorithm.Finally,we designed a workpiece recognition system based on machine vision,it contains five modules: image acquisition,image preprocessing,workpiece feature extraction,corner matching,result output and management.The application of machine vision technology in automatic workpiece recognition can greatly improve the production efficiency and liberate the labor force.Meanwhile,the improved algorithm can not only reduce computation cost for workpiece image matching,exclude pseudo-corner points and improve speed of recognition,but also achieve better matching performance in the respects of noise interference,rotation,zoom scaling.