Research Of Fabric Defect Detection Based On Low-rank And Sparse Matrix Decomposition

Fabric defect detection plays a key role in the textile quality control system,directly developing the performance of the system.It is of great application value to detect various kinds of defects from complex texture images.The solution of this problem is also helpful to provide new ideas to the surface defect detection of other industrial products,which has important value of application.The existing methods of fabric defect detection mostly adopt the traditional methods of pattern recognition.In recent years,derived from the theory of compressed sensing and sparse representation,the model of low rank and sparse matrix decomposition has been widely used in image processing and pattern recognition,and achieved good detection effect in object detection and saliency detection.The model of low rank and sparse matrix decomposition is consistent with the low rankness and sparsity of the human visual system.The image matrix is decomposed into a low rank matrix and a sparse matrix to achieve the effective separation of the object and the background.Especially,fabric image has a high visual redundancy,and compared with the object detection in natural scene,the task of fabric defect detection can better conform to the model of low rank and sparse matrix decomposition.In addition,the feature extraction of fabric image is also a key step of defect detection.For a given image,extract a proper feature,and construct suitable model of low rank and sparse matrix decomposition,and utilize the optimization method to solve the model.Finally,the distribution map of defects can be obtained via to using effective threshold segmentation algorithm,thus it can be effective and accurate to locate the location and regions of the defects.Therefore,this thesis presents some algorithms of fabric defect detection based on the application of gradient histogram and low rank decomposition,Gabor filter and low rank decomposition,GHOG and low rank matrix recovery and feature extraction of biological vision and low rank representation.The work and research results are as follows:1).In order to accurately detect the fabric defects in production process,an effective fabric detection algorithm based on Gabor filter and low-rank decomposition is proposed.Firstly,the Gabor filter features with multi-scale and multiple directions are extracted from the fabric image,then the extracted Gabor feature maps are divided into the blocks with size 16×16 by uniform sampling;secondly,we calculate the average feature vector for each block,and stack the feature vectors of all blocks into a feature matrix;thirdly,an efficient low rank decomposition model is built for feature matrix,and is divided into a low-rank matrix and a sparse matrix by the accelerated proximal gradient approach(APG).Finally,the distribution map of defects generated by sparse matrix is segmented by the improved optimal threshold algorithm,to locate the defect regions.2).An effective fabric detection algorithm by using histogram of oriented gradients(HOG)and low-rank decomposition was proposed.Firstly,the test fabric image was divided into the image blocks with the same size;a feature matrix was generated by extracting the HOG features of each block.Secondly,an efficient low-rank decomposition model was constructed,and augmented Lagrange method was adopted to decompose the feature matrix into a low-rank matrix and a sparse matrix.Finally,the distribution map of defects generated by sparse matrix was segmented by the improved optimal threshold algorithm,to locate the defect regions.3)A novel detection algorithm based on Gabor-HOG(GHOG)and low-rank decomposition is proposed.Defect-free patterned fabric images have the specified direction,while defects damage their regularity of direction.Therefore,a direction-aware descriptor is designed,denoted as GHOG,a combination of Gabor and HOG,which is extremely valuable for localizing the defect region.Upon devising a powerful directional descriptor,an efficient low-rank decomposition model is constructed to divide the matrix generated by the directional feature extracted from image blocks into a low-rank matrix(background information)and a sparse matrix(defect information).A non-convex log det()as a smooth surrogate function for the rank instead of the nuclear norm is also exploited to improve the efficiency of the low-rank model.Moreover,the computational efficiency is further improved by utilizing the alternative direction method(ADM).Thereafter,the distribution map of defects generated by the sparse matrix is segmented via the optimal threshold algorithm to locate the defect regions.4).A novel fabric defect detection algorithm based on biological vision feature extraction and low-rank representation is proposed.The representation of biological vision to the objective world is complete,and it can support all kinds of complex high-level vision tasks.We introduce a feature representation method combined human visual perception with retinal representation mechanism.On the basis of this feature representation,we use KSVD to train a dictionary of the normal fabric image block on the test image.Based on the dictionary,a low rank representation model of feature matrix is established,and the ADMM method is used to solve the problem,so as to improve the detection performance and adaptability of the algorithm.