A neural network approach for defect identification and classification on leather fabric

In this paper, an automated vision system is presented to detect and classi fy surface defects on leather fabric. Visual defects in a gray-level image are located through thresholding and morphological processing, and their ge ometric information is immediately reported. Three input feature sets are p roposed and tested to find the best set to characterize five types of defec ts: lines, holes, stains, wears, and knots. Two multilayered perceptron mod els with one and two hidden layers are tested for the classification of def ects. If multiple line defects are identified on a given image as a result of classification, a line combination test is conducted to check if they ar e parts of larger line defects. Experimental results on 140 defect samples show that two-layered perceptrons are better than three-layered perceptrons for this problem. The classification results of this neural network approa ch are compared with those of a decision tree approach. The comparison show s that the neural network classifier provides better classification accurac y despite longer training times.