Research On Real-time Quality Detection System Of Multi-color Printing Based On Machine Vision

By studing on the advantages of machine vision applied to industry detection and wavelet transform applied to image processing, the real-time printing quality detection system is developed aimed at detecting the printing quality.Firstly, this thesis studies the image capture of printing quality detection system which is focused on liner CCD, and introduces the principles and the applications of hardware used to capture images. In order to avoid the influence of error factors such as noise and geometry distortion, a method of image de-noising based on wavelet threshold shrinkage is proposed after analyzing various de-noising methods. This method can de-noise efficiently and reserve the edge information as much as possible. After de-noising, the image shoud be registered for the image matching.Secondly, this thesis compares several traditional matching algotithms and analyzes their respective merits and demerits. Considered both accuracy and efficiency for matching, a new matching method is presented, which combines the double-template matching and wavelet multi-resolution layer matching. This method can efficiently avoid mismatching and greatly reduce the calculation, so the efficiency and the accuracy of printing quality detection system can be satisfied.Finally, the thesis puts forward an overprinting detection algorithm based on image information. The image captured by liner CCD should be pre-processed, color-seperated and image-matched, then the overprinting error can be computed by the image-matching data. As the experiment shown, the overprinting system has the possibility and superiority of the real-time printing quality detection system, so it can satisfy the anticipated goals. It supports more theories and experiment data to further research of the real-time printing quality detection system.