Research on inspection and classification of leather surface defects based on neural network and decision tree

Surface defects of leather have great influence on the quality of leather products. A method is proposed in this paper to detect and classify the leather surface defects automatically and solve the problems such as misjudgment and high cost, etc. caused by artificial method. Here Feed-forward Neural Network (FNN) combining decision tree is adopted to select optimal attributes and classify the defects, which avoid the disadvantages of neural network like long processing time, “black-box” and that of decision tree like large calculation of construction and pruning and difficult to find the best. The effectiveness of the method is verified by experiments. The method proposed can also be used in other quality control relative to surface defects.