Algorithm Research of Vision Survey for Pattern Processed Trace and System Realization
|School||Guangdong University of Technology|
|Course||Measuring Technology and Instruments|
|Keywords||vision survey contour extraction boundary tracing vectorization|
In the industrial production, some machining movement is asked to execute the specific path request, such as in the welding, embroiders, the carving and so on. Using the visual analysis data and control system localization for enhancing the efficiency, the precision and the automaticity, the computer vision track technology has the opportunity greatly. While computer vision survey applies to quilting process, we can transform collected pictue to machined code, or do some picture collection, image processing and tracing in real-time to pattern processed trace for dynamic correct processed trace.This research originates from Guangdong Provincial natural science fund project in 2005 named "Research on the algorithm of dynamic position intelligent compensation based vision survey". Focus on the pre-real-time control processing stage of the entire flexible material processing, this paper has studied the vision survey algorithm for pattern processed trace, formulated and analysed on the key points and basic theory, also has some experiments. That is including image processing theory, design of vision survey system, image collection, image pretreatment, automatic contour extraction and vectorization.Purpose of image pretreatment is to erase the unrelated imformation, enhance the detectability of useful imformation and reduced data. Hance it’s necessary to use image smoothing and enhancing for pretreatment. This paper uses self-adaptive filtering, for providing a better environment initialization of image segmentation and outline extraction.Edge detecting is one of important image analyse, also the key point in this research. This paper introduces the basic theory of threshold segmentation, mathematical morphology and edge detection, analyses and contrast test the typical edge detecting algorithm. Combining with moment-preserving thresholding and multiple-dimensioned morphologic edge gradient operator, utilizes the self-adaptive contour extraction to extract the pattern contour.Image vectorization is the basic of pattern processed trace programming. An auto-adapting adjust approach based on direction module is proposed to track image edges, which optimized paths searched, and an optimaization algorithm for fitting beeline and curve by minimum distance method and straightway.During the paper research, not only has realized the establishment of system organization, management software based on Visual C++6.0, as well as corresponding processing algorithm, but also uses some methods and achievement of this research, has developed "Reproduced image system based on the visual survey" (production of the ninth Guangdong Province"challenge cup"), which can effectively realize image gathering, some special methods of image processing, the pattern image outline transforms the processing code, then reappear the pattern on the working platform.