Finding and quantitative evaluation of minute flaws on metal surface using hairline

In this research, a method to detect minute flaws on metal parts is proposed to remove the defective parts before assembling in a factory. The input gray-scale images of metal parts are used directly to recognize the flaw without any image conversion to shorten the recognition time. The recognition problem to find flaws and detect its position on the metal parts is converted here to another problem to search for the maximum peak and the variables giving the peak. Then the recognition problem can be treated as optimization problem, and this conversion allow us to utilize the high performances of GA in the optimization. The effectiveness and problems of proposed method are studied on standing points of the recognition speed and the quantitative recognition ability. Based on the analysis, we furthermore improved our system to increase the detection rate of the flaws, that is, the lighting direction is changed to find the best lighting condition that can emphasize the contrast between the metal surface and the bruise by using the reflection character of the hairline on the metal resulted by polishing process.