Research And Implementation Of The System Of Glass Bottles On-line Detection Based On Machine Vision

Glass bottles account for a large proportion in the packaging industry, but they wouldoften crack in the manufacturing process. The traditional manual inspection is of lowproductivity and high cost, so the research and development of automatic detectiontechnology of glass bottles is very necessary. As an emerging automatic detection technology,the machine vision technology, of high speed and precision, has been not only widely used inforeign countries but also booming developed in domestic. This paper has researched thesystem of glass bottles on-line detection based on machine vision.This paper describes the important and related technology of machine vision. Accordingto the technical specifications of the system of glass bottles detection, we design themechanical and electrical solutions, select the system hardware, and build the framework ofthe system software and point out that the core of the system software is the module of imageprocessing. This paper has focused on studying the control module of the system and thedetection algorithm of bottleneck. And we successfully develop an online detection system ofglass bottles, and develop a set of software of image processing of bottles.According to the overall design of the system, we analyse the control mechanism andpoint out that the key part of the control is the elimination module. And we select the PLC tocomplete low-level control of the system, and gave details of the mechanical and electricalagencies of the elimination mechanism. We achieve the software algorithms of the controlpart, and it works well. It is of two steps, the bottles positioning and bottles’ defect detecting,to detect the defect of the bottles. First we introduce the Hough transform algorithm, analysethe limitations of traditional Hough transform in engineering applications, and then introducethe improved Hough transform based on the vertical nature to positioning the bottle. Then wediscuss the defect detection algorithm of bottleneck. We analyse the characteristics of theimage subtraction and radial projection method, propose the tangential differential methodsuitable for the present system. The analysis showed that compared to the former twoalgorithms, the algorithm is better able to detect the existence of defects. Finally, we introducethe hardware and software implementation of the system, and the applications in the industrialfield.