Research On Color Classification And Color Difference Detection System In Fabric

Color is an indispensable indicator in textile industry, which plays an important role in textile quality, grade and price, even more economic efficiency of an enterprise. So it is of great significant in the actual production. At present, artificial methods are still continued to use in many enterprises. However, artificial methods may influent the accuracy and efficiency of detection. In order to solve the subsistent problem,the color classification and color difference detection of fabric based on image processing technology has been adopted in the paper. It helps improve the quality of textile and efficiency of color detection.Two different methods based on the HSV and the CIELAB color spaces have been proposed in color detection of fabric. The algorithm of histogram intersection based on HSV space needs conversion formula to convert the RGB color space to HSV color space.Secondly, quantize color characteristics and obtain a one-dimensional histogram, and then cumulate the smaller values of the two images in each histogram obtained on the histogram feature images in the public portion of sense. So the color difference of fabric can be detected qualitatively by the method.Levenberg-Marquardt(LM) optimized BP algorithm has been adopted to obtain the color feature values of fabric images instead of color space conversion formula, which is based on the CIELAB color space. The color difference of fabric can be detected quantitatively by this method. Through a large number of samples training for many times to get the best network parameters, so as to establish the relationship between RGB color space and L*a*b* color space. In order to verify the validity of the algorithm, four kinds of formulas are used to obtain the color difference of fabric. Compared with results measured by SP60 spectrophotometer, the color difference can be detected quantitatively with high accuracy and efficiency.In the paper, BP neural network and RBF neural network have been proposed to classify the color of fabric. Color characteristic value obtained through OHTA color space is input of neural network. Color category is output of neural network. The parameters such as the number of hidden layer nodes and learning rate, as well as the distribution density of the RBF neural network can be determined in experiment. The experimental results show that LM optimized BP algorithm with a higher result of classification thanRBF neural network.The implementation of color difference detection and color classification has great significance and practical value.