Fabric Defect Detection Technique Based on Two-double Neural Network

This paper introduces the identification of the defects on the fabric by using two-double neural network and wavelet analysis. The purpose is to fit for the automatic cloth inspection system and to avoid the disadvantages of traditional human inspection. Firstly, training the normal fabric to acquire its characteristics and then using the BP neural network to tell the normal fabric apart from the one with defects. Secondly, doing the two-dimeusional discrete wavelet transformation based on the image of the defects, then wiping off the proper characteristics of the fabric, and identifying the defects utilizing the trained BP neural network. It is proved that this method is of high speed and accuracy. It comes up to the requirement of automatic cloth inspection.

This paper introduces the identification of the defects on the fabric by using two-double neural network and wavelet analysis.The purpose is to fit for the automatic cloth inspection system and to avoid the disadvantages of traditional human inspection.Firstly,training the normal fabric to acquire its characteristics and then using the BP neural network to tell the normal fabric apart from the one with defects.Secondly,doing the two-dimensional discrete wavelet transformation based on the image of the defects,then wiping off the proper characteristics of the fabric,and identifying the defects utilizing the trained BP neural network.It is proved that this method is of high speed and accuracy.It comes up to the requirement of automatic cloth inspection.