| 英文摘要 | In recent years, machine vision technology has attracted both the academia and industry's great interest and attention due to its fruitful applications in some technological fields, and these fields are expanding rapidly. Inspecting methods based on machine vision improve the efficiency of production and raise the level of automation. Therefore, in the modern automatic production processes, machine vision systems are extensively adopted in many fields, such as condition control,product inspection and quality monitoring. However, there are lots of challenges in these vision-based inspecting systems which have not got well resolved yet,such as how to guarantee the error-free performance and strong robustness while satisfying the real-time requirements of application environment. Especially,comparing with abroad advanced technology, domestic research and development of inspecting systems based on machine vision which has independent property rights have fallen behind. So researches on the key technologies of vision-based applications have great significance in promoting our own technology and long term economic development. This thesis is focused on some key techniques of a typical visual system based on machine vision -presswork's variable contents on-line inspecting, including automatic rapid barcode localization in complex scenes, improving the real-time recognition ability of low quality barcode images, fast digit recognition with font independence ,etc. The main contributions of this thesis include the following issues: 1.Global threshold methods cannot achieve good performance when the lighting is uneven, whereas local threshold methods are sensitive to noise. To address this issue, we propose a new adaptive binarization algorithm-refined Bernsen binarization based on information entropy. The threshold obtained by this algorithm not only has adaptability, but also has noise immunity to a certain extent. 2.The main shortcomings of the current vision-based barcode localization when the images acquisition is less desirable, for example, the background is too complex, or the lighting is uneven, include low efficiency and lack of accuracy, since under those conditions, the contrast of barcode regions is not so obvious or the edges between black and white strips are too vague to be determined. To tackle this problem, we present an efficient algorithm based on contour for barcode localization. Experimental results showed that our algorithm... |
修改评论