Image Fusion Of The Surface Defects Feature Of Strip Steel Based On Different Image Collecting Modes

Currently, the surface quality of strip steel has been a focused issue for steel enterprise. In order to acquire the steel products of good surface quality, first of all, the enterprise must obtain the surface defects of strip steel, and then, control the production of strip steel. In fact, whether the surface quality of strip steel could be raised, not only lies on the surface defects classification, but also the collecting of the most all-sided and accurate defects features before classifying.The difference of collecting mode for the surface defects could make a great affect on the feature of the defects collected. The collecting mode with sole CCD sensor still is used for the current inspection system of the rolled steel surface defects. However, the feature of the defects acquired from this mode could not usually fully represent and express the actual defects, and in many cases just a part of them. So, before correctly classifying the defects, it is very indispensable to completely and accurately collect the feature of the defects.Based on the above problem, this dissertation puts forward a new idea which is using multi-sensor to collect the representative defects images of strip steel surface based on different collecting modes under the condition of laboratory, and then applying two different methods for image fusion, which are the simple image fusion and the image fusion based on wavelet transform. To integrate and process the features from different defects images. At meanwhile, the system software for image fusion is developed to simulate the results of image fusion. At last, a subjective performance evaluation system of image fusion is founded to validate the superiority of image fusion method based on wavelet transform.The idea from this dissertation to integrate and process the features from different surface defects images of strip steel solves some problems on defects losing with sole CCD sensor collecting defects image, especially the feature losing due to the difference of the collecting modes under bright-field and dark-field. By this new idea this paper could acquire more complete, and more accurate defects feature, and then provide data information to the latter identifying and classifying of defects.