Research On The Cashmere&wool Image Recognition System Based On The OMAP3530

Cashmere and wool are precious textile material. Distinguishing cashmere and woolcorrectly, quickly and objectively is very necessary, whether to safeguard the interests ofconsumers or to maintain the performance of complementary in blending processing of textile.The method of computer digital image processing has strong advantage undoubtedly.In this paper, the objects of digital image processing are cashmere&wool fiber imageswhich are gotten through optical microscope. In order to obtain effective feature vector, imagepreprocessing is necessary including image binarization, image filling, image spatial filtering,frequency domain filtering, morphological image processing and so on.According to the difference of cashmere&wool,, eight features of Cashmere and woolare extracted including diameter, scales density, scale thickness, radial axle parameters,degree of roughness, while roughness are described by histogram mean, standard deviation,entropy and moment of order three. Before distinguishing cashmere and wool principalcomponent analysis is used in order to reduce dimension. Then Bayesian decision theory andsupport vector machine method based on the cross validation are used to recognition cashmere&wool, the results show that the method of support vector machine is superior to theBayesian decision. The process of Cross validation is: random select90%of200sampleimages (each100) as training samples,10%of the remaining samples as to be identifiedsamples. The average recognition rate is92.45%, the unbiased estimation of variance is0.287%.According to the study from theory to practical application process, the algorithmprogramming language is from MATLAB to C/C++language; the recognition result is shownin the form of MATLAB interface to the WINDOWS interface to the WINCE interface; experiment platform is from the computer system to OMAP3530embedded system.