Research On Feature Extraction And Optimization Based On Digital Image Of Cashmere And Wool Fiber

Cashmere production is thin and soft, beautiful and elegant, and it is more and more popularto people. The proportion of cashmere in cashmere production is closely related to the cost ofproduction and the production pricing, therefore, it is necessary to find quick, accurate andobjective identification for cashmere. It can achieve the desired effect for identifying cashmere byusing computer image recognition method.This topic chooses cashmere and wool fiber digital images as the research object, thosecashmere and wool digital images are obtained by differential interference contrast microscopeprovided by China Fiber Inspection Bureau. And during the recognition, pretreatment and featureextraction are done, and finally pattern recognition algorithm is used to recognize the cashmereand wool fiber.The preprocess methods mainly include: image rotation and cropping, image binarization,morphological image processing, image processing in time domain, frequency domain processing,etc. The features extracted include: the diameter of scales, the height of scales, the diameter toheight ratio of scales, the density of scales, the diameter shaft parameters of scales and the texturefeature of scales, and the texture feature of scales containing mean value, standard deviation,smoothness, third order moment, consistency and entropy. At last, support vector machine (SVM)is used to recognize the cashmere and wool fiber digital image.During the feature optimization, single and double features are compared, in order to achievethe purpose of better identify cashmere and wool, at last, recognition rate of cashmere wool fiberdigital image is affected synthetically by the diameter of scales, the height of scales, the diameterto height ratio of scales, the density of scales, the diameter shaft parameters of scales and thesmooth in the texture feature of scales, the rate is89.25%.