Research Of Fabric Defect Detection Based On Golden Image Subtraction

In textile industry, fabric defect detection is one of the main solutions to fabric quality assurance. Traditional human inspection has low detection speed and is labor-intensive. Besides, the detection accuracy is affected by the experience and fatigue of the human inspector. Thus it is devoid of consistency and reliability. So, research and development of automatic fabric defect detection system has great significance for textile industry development. The main subject of this study is researching defect detection algorithm and designing on-line automatic defect detection system based on machine vision. The detail procedure is described as follows.(1) According to patterned fabric defective problem, a fabric defect detection algorithm based on Gabor filter and Golden Image Subtraction is put forward in this chapter. Stage of training is to use artificial experiments or genetic algorithm for choosing the optimal parameters of Gabor filter which that match the texture features of training samples, then the optimal parameters of Gabor filter are used in constituting optimal Gabor filter. The detection phase is to execute Golden Image Subtraction between golden template image and the filtered image, and to determine thresholding by Direct Threshold. Then the thresholding is utilized in segmenting fabric defects. The experimental results show that the methodes possess precise detection rate on patterned fabric.(2) In the light of the problem of grey fabric defects, a fabric defect detection algorithm based on Optimal Tree Structure of Wavelet Decomposition and Golden Image Subtraction is proposed in this chapter. In stage of training, PCA is used to reduce the number of sub-images of optimal tree structure of wavelet decomposition. Then the de-mixing matrix is obtained by Fast-ICA. In the detection phase, fabric defect also is extracted from the fabric background by executing Golden Image Subtraction and Direct Threshold. Meanwhile, the paper also presents a comparison between the proposed method and the Wavelet preprocessed Golden Image Subtraction. The experimental results show that the proposed method possesses much high detection rate on un-patterned fabric.(3) For the problem of defect detection of regular patterned fabric with background of high contrast, a fabric defect detection algorithm based on Halcon is presented in this chapter. In this algorithm, gray scale and mean filter are adopted to preprocess fabric image in the first place. Then, defect areas are inspected by setting parameters such as thresholding, shape and area. The experimental results show that the method can detect fabric of high contrast background with high detection accuracy such as star-patterned fabric and box- patterned fabric with black and white color.(4) A set of automatic defect detection system are designed and set up in a laboratory environment, which combines with hardware and software to achieve on-line and real-time defect detection. Experimental results show that the proposed system has better reliability, accuracy and real-time performance for common fabric defect detection. Thus, the system can consider further debugging in the industrial field.