Visual attention mechanism-based fabric pilling grade evaluation method

Abstract

The invention discloses a visual attention mechanism-based fabric pilling grade evaluation method. The method comprises the following steps of: 1, acquiring a fabric pilling image; 2, carrying out wavelet multilayer decomposition on the fabric pilling image; 3, determining a central layer subgraph and a peripheral layer subgraph of the fabric pilling image; 4, constructing a fabric pilling saliency map; 5, obtaining a fabric pilling target; 6, extracting fabric pilling features; and 7, evaluating a fabric pilling grade. According to the visual attention mechanism-based fabric pilling grade evaluation method, the central layer subgraph and the peripheral layer subgraph are correctly selected and central-peripheral operation is carried out between the two subgraphs, so that the significance of the pilling target can be improved and the calculated amount can be decreased; threshold value segmentation and filtration are carried out on the fabric pilling saliency map, so that the fabric pilling target can be effectively obtained; and on such basis, the pilling features are extracted, so that objective evaluation can be rapidly and correctly carried out on the fabric pilling grade.

Description

technical field [0001] The invention belongs to the technical field of fabric surface evaluation methods, and relates to a method for evaluating fabric pilling grades based on a visual attention mechanism. Background technique [0002] Fabric pilling not only affects the appearance of the fabric, but also reduces the wearing experience of the fabric. Therefore, the evaluation of the fabric pilling level is an important indicator for testing and controlling the quality of the fabric. The traditional evaluation of the pilling level of fabrics is mainly done by the naked eye of inspectors. The effect of this grade evaluation is affected by subjective factors such as personal experience and psychological factors of the inspectors, so it is impossible to make an objective evaluation of the pilling level of fabrics. . [0003] At present, with the development of computer and image processing technology, new algorithms for automatic detection of fabric pilling and objective evalua...

AI Technical Summary

Problems solved by technology

[0004] In the space domain, many researchers have done a lot of research and achieved certain research results, but there are still some deficiencies: the gray histogram statistical method mainly performs gray histogram statistics on the gray image of the fabric, and then Carry out threshold segmentation, select the area number of pilling as the feature parameter, but when the fabric texture is complex, the floating points on the fabric are easily mistaken for hair balls (Konda A, X in L C.Evaluation of Pilling by computer imageanalysis .Journal of the Textile Machinery Society of Japan, 1990, 36 (3): 9 6-10); Support vector machine objective evaluation method, using the performance of data mining fabric pilling of support vector machine (SVM) and making predictions, but The parameters of SVM are mainly selected based on experience, which is highly subjective (Poh Hean Yap, Xungai Wang, Lijing Wang, and Kok-Leong Ong. Prediction of Wool Knitwear Pilling Propensity using Support Vector Machines[J]. Textile Research Journal, 2010, 80( 1):77-83); Gaussian template matching method, has used two-dimensional Gaussian fitting theory, trains the pilling template with the actual pilling image, and determines reasonable threshold with histogram fitting technique to segment the image, then extracts the pilling Number, average area of ​​pilling defects, total area of ​​pilling, pilling contrast and density to train pilling features, and evaluate pilling level by establishing corresponding formulas, however, whether the trained pilling template is the best match The pilling template will directly affect the evaluation effect (Binjie Xin, Jinlian Hu, and Haojin Yan. Objective Evaluation of Fabric Pilling Using Image Analysis Techniques [J]. Textile Research Journal, 2002, 72 (12): 1057-1064); The visual objective evaluation method, using stereo vision for stereoscopic evaluation, is mainly suitable for the evaluation of the pilling level of soft thick yarn fabrics, but not suitable for the evaluation of the pilling level of hard fine yarn fabrics (Bugao Xu, Wurong Yu, and RongWu Wang.Stereovisionfor three-dimensional measurements of fabric pilling [J]. Textile Research Journal, 2011, 81( 20):2168-2179; Kim S C, K an g T J.Evaluation of fabricpilling using hybrid imaging methods[J].Fibers and Polymers, 2006, 7(1):57-61)

The fabric defect detection method based on salient texture features, using the best window based on local texture, extracts and fuses roughness, contrast and direction to generate a visual saliency feature map to highlight fabric defect areas, but only improves the saliency from fabric defect features. It is difficult to improve the significance of small and numerous pilling defects, and it is also difficult to filter out noise and other information by using the Ostu segmentation method; plain fabric defect detection method based on visual saliency (Guan Shengqi, Gao Zhaoyuan, Wu Ning, Xu Shuaihua. Plain fabric defect detection based on visual saliency[J]. Textile Journal, 2014, 35(4):56-61), using wavelet to decompose the fabric pilling image, and performing central-peripheral operations on sub-images of different layers , the differential subgraphs are fused to form a saliency map, and then, the maximum inter-class variance segmentation method is used to segment the fabric defects. This method selects the subgraphs of all layers for central-peripheral operations, without considering the information of the layer subgraph itself , the central-peripheral operation on some sub-image information of the central layer and the peripheral layer has little difference, or the difference is not the pilling defect information, which will not only increase the calculation amount, but also reduce the significance of small defect information such as pilling. It has not been significantly improved. In addition, the maximum inter-class variance segmentation method also has the defect that it cannot eliminate the noise in small defect information such as pilling.

[0008] It can be seen that the current method based on the visual attention mechanism can be used for the detection of a wide variety of fabric defects, but it is not suitable for the detection of fabric pilling images with small areas, low contrast, and large numbers. How to effectively improve the conspicuousness of fabric pilling And the binary segmentation method directly affects the final objective evaluation effect of the pilling level

Images