Fabric defect detection based on fusion technology of multiple algorithm

As the variety of fabric defects, it is difficult that there has an image processing algorithm suitable for the detection of all defects. A new method was presented for defect detection. Normal texture is filtered by Fourier transform in the frequency domain and it is increased to the serious defect information. Wavelet single decomposition and approximate sub-image filtering are combined to inhibit the normal texture high frequency and low frequency information, and to enhance the contrast of common defect information. The normal texture information and defect information are separated by orthogonal wavelet multi-decomposition for discrete small defect detection. On this basis, the image characteristics are extracted in sub-windows of image; then the defects are identified by neural network. Experimental results show that the method is effectiveness.