Warp Knitted Fabric Defect Detection Based On Digital Image Processing

The detection of fabric defects is of great significance in the textile process,especially in warp-knitted fabrics,the fabric often stops due to warp breaking,which affects the production efficiency of the fabric.There are many defect detection systems on the market,but most of the equipment is expensive or not convenient enough.Therefore,this paper applies digital image processing and embedded technology to the defect detection of warp-knitted fabrics to realize the real-time detection of fabric defects.From the perspective of frequency domain and texture features,this thesis designs two defect detection algorithms based on Gabor transform and wavelet transform,performs Matlab simulations on them and verifies their real-time performance and recognition rate,finally chooses an algorithm to be transplanted to STM32 on.Among them,the two algorithms use the same image preprocessing methods,namely histogram equalization and median filtering.In Gabor transform algorithm,the construction of Gabor filter needs to choose the appropriate direction and scale.In terms of direction selection,vertical and horizontal filter banks are constructed.In terms of scale selection,the texture period T is calculated by the method of seeking the extreme value of the distance superposition function,and a filter bank of the size of T×T is constructed.Finally,the gray-level co-occurrence matrix method is used to calculate the eigenvalues of the image after Gabor filtering,and then it is compared with the flawless area to identify the blemish features.The key of wavelet transform algorithm is to select the appropriate wavelet base and decomposition level.In order to highlight the fabric defects,DB wavelet base and four level wavelet transform are selected to binarize the approximate component and horizontal high frequency component of the fourth level.Adaptive threshold is adopted to reduce the interference of different detection environment.Finally,morphological operation is used to dilate and corrode the binary image to get the defect features of the cloth.Two typical cloths are tested on Matlab software.By comparing the real-time performance and defect recognition rate of the two algorithms,it is concluded that the defect algorithm based on wavelet transform is better than the Gabor transform algorithm.Finally,a flaw detection algorithm based on wavelet transform is implemented on STM32.Considering the limited hardware resources of STM32,the lifting wavelet algorithm is used to realize the two-dimensional discrete wavelet transform,which reduces the memory usage.Different types of warp knitted fabrics with a resolution of 1280×240 are tested.This method can effectively identify defects and achieve the purpose of real-time detection.