The Research For Detecting The Medicinal Glass-Bottle Defects Based On Computer Vision

At present, the quality of medicinal bottle mainly relies on artificial inspection. It will increase the factory's labor cost and management cost, while the accuracy and speed of detection does not consistent with the production requirements. As for the inspection of a medicinal glass-bottle defects, the following aspects were studied and discussed in this thesis.Firstly, against the research for detecting the medicinal glass-bottle defects based on computer vision, at the first, we need the images of high quality. Start with the selection of acquisition equipment and design illumination apparatus. At the same time, the structure of the inspecting system was introduced.Secondly, machine vision technology will be combined with image processing technology during the course of a glass bottle defect detection and identification process. Discussed in detail the measured object image preprocessing process, we proved the viable algorithms. In the edge detection, curve fitting algorithm was used to improve traditional pixel detection methods and was attained the sub-pixel accuracy.Thirdly, the medicinal glass-bottle defects of the mouth and the body are taken the different inspection algorithms. Fast boundary tracking algorithm of binary-image was researched. As for the mouth of the bottle detection key technology is to locate the center of circle. After comparison a variety of methods identifying the center of circle, we adopted the improved method of four-point vertical chord-cut and proposed double circle algorithm to inspect the mouth of bottle defects. We apply fast boundary tracking algorithm to extract the region of cracks of bottle, and calculate geometric parameters to judge defect types. Experiments testify that kinds of algorithms achieved good results.