Application of perceptron in recognizing surface defect information on steel sheet

A novel method was presented to separate defective information from steel sheet surface images by anatomizing the grey values of steel sheet surface images, which uses a classifier composed of some simple perceptrons. The method directly preprocesses image data in camera, and only provides defective data out to be further processed, so the amount of data in the system which must be further processed is reduced dramatically. Furthermore, the method was implemented in FPGA, and experiment results show that after using this method, the system hardware structure is simple, as well as the real-time performance is improved greatly.