Construction And Application Of Woven Fabric Texture Representation Model For Colored Spun Fabric

Texture is an important feature that characterizes the content of an image,which describes the position and arrangement of each pixel in the gray space.Traditional fabric analysis is mainly done manually,which has the problems of low efficiency,low accuracy and poor stability.With the popularization and application of digital image processing technology,it is the main research direction to realize the automation and intelligence of fabric texture analysis through the digital representation of fabric texture.In recent years,two or more kinds of different dyed fibers are blended and spun into color spinning products,which are deeply loved by the majority of consumers because of their rich color levels.Therefore,how to use computer vision technology to characterize and analyze the texture of colored spun fabrics is worthy of further study.Aiming at the randomness and complexity of texture distribution of colored spun fabric,this paper constructs texture representation algorithms in space domain,frequency domain and space-frequency domain respectively,and is used for the recognition of interlacing points and classification of fabric structure of color spun woven fabric.The specific contents are as follows:(1)Aiming at the difficulty of texture feature extraction of fabric interlacing points on colored spun fabric,an automatic recognition algorithm based on texture statistical features and multi-core learning was established.Firstly,the brightness channels of YUV,HSV and Lab color space are fused to construct a mixed brightness channel.On this basis,the local texture statistical features of interlacing point image were extracted to characterize the texture of fabric interlacing point.Finally,support vector machine was constructed by multi-core learning algorithm for recognition.(2)Aiming at the difficulty of fabric texture feature extraction in color spun woven fabrics,a fabric texture classification algorithm based on multi-resolution features and hierarchical hybrid classifier was proposed.Based on the HSV color space,the algorithm used wavelet multi-resolution analysis to decompose the colored spun fabric into three layers in the brightness channel,and extracted the local binary mode features of each layer image for fusion.At the same time,BP network and Naive Bayes theory were used to construct a hierarchical mixed classifier to realize the classification of fabric structure.(3)Aiming at the problem that the local binary model is difficult to accurately characterize the texture anisotropy caused by the uneven distribution of dyed fibers on the surface of colored spun woven fabrics,a classification algorithm for texture structure of colored spun woven fabrics combining spatial and frequency domain features was proposed.In this model,local binary mode and local phase quantization algorithm were used to extract the fabric texture,and then principal component analysis was used to reduce the dimension of fusion features.Finally,a hybrid classifier was used to recognize the fabric texture.The research of this paper has important guiding significance for the construction of stable and effective color spun fabric design and production system,and can provide technical support for the application of colored spun fabric digital management and defect detection,which has important theoretical research significance and application value.