An Inspection for Warp Knitted Fabric Using Image Processing

The factory which manufactures cloth textile product needs an automatic detection defect detection system to improve efficiency. As the conventional method, the method of irradiating a laser beam, the method of using a statistical model in an image processing, the method of using the difference of an image, and the method of depending on wavelet transform are proposed. However, the method of irradiating a laser beam cannot be used for cloth partially with uneven thickness. Furthermore, when an image was blurred by oscillation of a machine and cloth, the method of using a photographed image had the problem to which detection precision falls. This method had detected the device behind the cloth as a defect if thickness of the cloth is thin. These conventional methods were difficult to use for inspection of the moving cloth being weaved. This paper proposes a new method of detecting the defect of cloth automatically. First, the image of cloth photographed by a camera is divided into a block at equal intervals. Then a defect is detected by filtering two kinds of filters to the luminance average value in a block. The proposal method uses the brightness sum of the block set by the configuration of the defect. Since the difference of the brightness sum of the defective part and the normal part is clear, the proposal method can respond to both the defect to which the hole of cloth spreads, and the defect in which cloth is thin. Moreover, since the proposal method uses the filter which has resistance in a small noise and large noise, this method can detect only the defective part, without detecting the device behind the thin place of cloth. It was confirmed at the production plant that the defect of the cloth textile product during manufacture can be certainly extracted by the proposal method.