The Identification Of Fabric Defects Based On Curvelet Transform And BP Neural Network

Defect inspection is one of the indispensable steps in the fabric detections. The detecting depend on manual labor has many drawbacks, by which there is high rate of missed detection, low efficiency and low reliability. With the development of image processing technology and artificial intelligence, automatic inspection of fabric defect as a replacement of manual labor is becoming possible.Fabric images can be contaminated by noise on the processes of formation, transmission, reception and processing. To receive high quality fabric images, de-noising is important. De-noising based on curvelet transform can receive better fabric defect image than de-noising based on wavelet transform.Fabric defect inspection using image processing technology is a problem of extracting of textural features and pattern recognition. Extracting of textural features of fabric images is a key process. Energy and entropy which are statistic features of sub-image of curvelet transforming are employed to describe the fabric images. For the part of identification of fabric defects , a three layer BP neural network is designed. After testing, six classes fabric image can be correctly identified by the rate of 95.83%.