Research Of Fabric Defect Detection Based On Hybrid Self-Adaptive Wavelet Basis

In the process of textiles production, the appearance of fabric defects affect the fabric quality seriously, so the fabric defect detection is one of the necessary link in the factory, and it has important significance and application value. However, there are many kinds of fabric images with complex texture, so it has become an important topic of detecting the different fabric defects successfully in this research field. In the automatic detection of fabric defects, wavelet is used extensively by people because of good characteristics of spatial location and multiresolution. However, the adaptability of traditional wavelets is not suitable for detecting different texture background, instead, the self-adaptive wavelet basis can better adapt to different texture background, so that the defect information is easier to identify in each layer of the fabric image transformed by wavelet. In order to improve the adaptability of the adaptive wavelet on the texture of fabric, improvements have been made on image pre-treatment, construction of hybrid self-adaptive wavelets and fusion of decomposed images. The work and research results are as follows:(1) Establish a wavelet library with certain length and choose the self-wavelet by the limited conditions of target fabric texture. In order to improve the adaptability of wavelet transform on target fabric texture, in the preprocessing stage of the fabric image, presents an optimization selection algorithm of fabric image resolution based on the image entropy. Different resolution images of the same kind of fabric are given within a reasonable range, and select the optimal resolution through the analysis of the decomposed images by image entropy.(2) As a single self-adaptive wavelet basis is not suitable for multi-level wavelet decomposition, this paper proposes a fabric defect detection algorithm based on hybrid self-adaptive wavelets. In the multi-level statical wavelet decomposition, optimize the normal fabric image and the low frequency sub image to get the hybrid self-adaptive wavelets which consist of different self-adaptive wavelets. Experimental results show that, compared with the single hybrid adaptive wavelet, sub image decomposed by hybrid self-adaptive wavelets can be more effective on defect detection, and has good defect segmentation and detection results.(3) In order to improve the algorithm with only one limited condition, this paper proposes a fabric defect detection algorithm based on improved self-adaptive wavelet basis. Firstly, the self-adaptive wavelet basis is optimized by three kind of different limited conditions; then the self-adaptive wavelets are used to implement 2-L wavelet decomposition and the Otsu method is used to implement the image segmentation; finally, detection effect is obtained from fusing the multi-segmented images according to the noting mechanism. Experimental results demonstrate that the proposed algorithm can preserve the text information well and effectively reduce the noise points in the test result at the same time.