Glass Detection System, Based On Digital Image Processing

As the production scale of glass bottle is enlarged, the speed in production line increased and more and more strict demands required on quality, the traditional inspection method for glass bottles can no longer meet today's requirements. With the development of computer hardware and software as well as the improvement of digital image theory, the online detecting method based on digital image processing techniques, with such characteristics as rapid speed, high precision and non-touch, has become feasible and thus become the preference for glass bottle manufacturers.First, some fundamental knowledge about the background of inspection system as well as its significance and present research in China and abroad is reviewed. Second, several aspects of inspection system such as collectivity project, technology index and software and hardware setting are discussed. Then we make some study and analysis on algorithm and software design in detail. In terms of algorithm, we analyze in depth digital image processing techniques related to glass bottle inspection, putting emphasis on the discussion of their application on bottle mouth crack and size inspection as well as the analysis of the advantages and disadvantages of different kinds of algorithm. As for bottle mouth crack inspection, we put forward a positioning algorithm based on the displacement sensor, with which the collected image could be positioned timely and exactly. We also develop a rapid circle track inspection algorithm to realize the judgment of the crack. As for bottle mouth size inspection, we adopt fifteen-direction boundary chain track algorithm and Robert boundary inspection algorithm to detect size images. Through a great amount of experiment, we prove the inspection algorithm to be effective and feasible as far as crack and size are concerned.In terms of software design, we use Visual C++6.0 to explore a set of software for testing and research, e.g. glass bottle inspection system. This system successfully completes the testing and research for all algorithms, thus establishing a solid foundation for further study. During the software exploration, we use COM control to beautify the software interface, enhance its function, and improve its operation speed. Here there are three methods improved: a) a macro definition, which is a multi-file method, is raised in setting up the dynamic link library. b) ADO database accessing technique is adopted, in which class is encapsulated into DLL to make the operation easy. c) API function is utilized to realize series communication, which makes the exchange between computer and control system convenient. While accomplishing main research topic, we study the transverse direction topics image mosaic. It briefly explained application and the vital role of the Image mosaics in each domain and studied a fast algorithm for image mosaics. First, we make a particular mode of projection of images, with coordinates on a graph transformed into those on a column plan, to maintain the spatial restraint relations of actual scenery. Then we obtain the horizontal offset and vertical offset with restricted phase correlation, using the fast Fourier algorithm to increase the processing speed of algorithm. After that, we establish a seam-line in two aspects of color intensity and geometry structure according to the best seam-line criterion. Finally, after decomposing the images, we deal with image mosaics under different resolutions so as to eliminate exposure differences. The whole process could be realized with the coding of Visual C++.