Study On The Computer Vision Evaluation Of Fabric Luster

Fabric luster has always been an important basis for evaluating the quality of its appearance,and it is one of the main indicators of its visual style.The test and evaluation technology of fabric luster is of great importance not only for improving the visual effect of fabrics,but also for improving the style of finished products,as well as for the design of gloss,process improvement and quality control of fabrics.Therefore,it is very important to study the test and evaluation technology of fabric luster.At present,there are two kinds of evaluation of fabric luster: sensory evaluation and instrument test.However,the human evaluation and determination of the reflected light intensity of instrument are lacking for a complete cognition and characterization of fabric luster,and only come out the analysis result,it is difficult to achieve fabric design and development from the fabric surface properties.The purpose of this paper is to propose an objective and subjective evaluation system for fabric luster based on computer vision and image analysis technology.Based on the existing fabric luster evaluation and measurement methods,firstly,a scale for characterization of fabric luster visual physical quantity is constructed,the "semantic scale" that is combined of the fabric luster visual physical and visual psychological is proposed,and a Fabric Luster Testing Experiment System is also constructed by using optical principle and computer vision technology,and then through the experimental study for a certain amount of fabric samples from the luster image,five characteristics-gray contrast,luster filling range,GLCM contrast and saturation contrast are presented that include four aspects-gray feature,direction features,texture feature and color features,and the analysis and integration of the relationship between each characteristic parameter and the fabric luster grade is made by using the optimization extraction algorithm of multi dimensional feature parameters and the analytic hierarchy process(AHP)method.Finally,the complex nonlinear relationship of fabric luster grade and each index is simulated by the nonlinear approximation ability of discrete Hopfield neural network,the fabric luster classification model is established,from which the grade evaluation and visual evaluation can be obtained according to the fabric luster.The experimental results show that the accuracy of the simulation results is 94%.The evaluation of fabric luster system based on computer vision has replaced the uncertainty of human eye evaluation and the lack of optical physical measurement of existing instruments.It provides a new solution for the objective evaluation of fabric surface luster from theory and method.