| Molding production is the most important and most advantageous way of processing for industrial products. Using machine vision technology for molding production real-time monitoring and product quality control has its important significance. Anomaly recognition of molding production based on visual inspection was presented.The research works include three key technologies and an application system: mold foreign substances inspection technology, molding production oriented surface defects detection technology, molding production oriented shape inspection technology, and hierarchical modular dual-mode visual inspection system. There are four innovations: technology of images scattergrams analysis, method of Top-Hat transform sequence analysis, technology of adaptive edge detection based on small scale wavelet transform is proposed, method of shape analysis using wavelet multi-resolution.The organization structure of full text is as follows:In chapter 1, the technology of molding production anomaly recognition base on visual inspection is described. The research and application of machine vision detection at home and abroad is reviewed. The significance and shortage of research on mold production monitoring is pointed out. The research content and architecture of the dissertation is introduced.In chapter 2, the two difficulties of image change detection is studied. In order to make the algorithm illumination invariant, the regression relation between the pair of compared images was derived by analyzing the probability distribution of the corresponding pixel gray in the two images. For the purpose of solving the problem of the geometric deviation between two images, MRA (Multi-resolution analysis) was used before inspection. A pair of mould monitor images is compared and recognized through this technology and realize protecting molds automatically.In chapter 3, Iterative open operation and Top-Hat transformation of the signal are computed using size driven structure elements and the O-TH (Open-TopHat) decomposition tree is obtained. Objects analysis spectrum is established whose amplitude is the energy of the O-TH tree leaf nodes. Then, the optimum size of the structure element and the optimum segment threshold are obtained. On this basis, the preliminary segment result is patched by comparing the elements in the sequence with the Top-Hat transform result which uses the optimum structure element, and the targets in the signal are detected completely. Scratches, weld lines, silver streaks and craze crack are detected by this technology and realize inspecting defects on cavity and product surface automatically.In chapter 4, the characteristics of object edge under complex background is studied. The impact of wavelet scale on the transform result is analyzed. Edge localization techniques based on small scale wavelet transform and integrity edge linking is presented, and the contour of the objects is extracted completely. The shape analysis technology based on MRA is proposed, which make the transformation starting-point invariant. The objects' shape is identified under the course scale and the defects of objects' contour is detected accurately under the fine scale. The shape of molds or products are captured and analyzed whether there are warping, damage, molding flash or other abnormalities happened by this technology and achieve their good shape quality.In chapter 5, the development and application verification of the prototype system for molding production anomaly inspection is described. The constraints and key elements of design for hardware and software platform is elaborated, and the reliability, generality and scalability of the machine vision system is realized. The methods of dissertation are verified by application including foreign substances, surface defects and shape deformation through independent prototype system.In chapter 6, the research results of this dissertation are summarized and the ways for further research are pointed out. |
PDF Full Text Request