Using The Scale Pattern To Identify The Cashmere And Determine Its Content

At present there are many identification methods of cashmere and wool, but the industrial standard method is still microscopy, which mainly use the differences of scales to identify cashmere and wool, and identification inspectors’experiences have great affects on the test results. So an objective fact is that cashmere and wool may be confused and uncertain, and now the diameter is coarsening seriously, So it’s imperative to identify cashmere fibre accurately. In order to reduce the influence of subjective factors and improve the accuracy, the method which combines computer technology and microscopy detection technology is used to identify cashmere and wool.Against this background, this paper researches and improves a optical microscopic imaging analysis system, it can capture images form fabric. Then comparing characteristics of images from three kinds of lighting systems, it can be concluded that images of the reflecting illumination are clear and this lighting system is suitable for fabric samples. The image quality of coaxial fall shot illumination is poorer than reflecting illumination, but better than transmissive illumination.Then, this instrument is used to capture images from samples. Through image processing and feature extraction,8characteristic parameters and4combination parameters representing fiber scale are extracted form these images. And the relation of each parameter of fiber scale in samples is analyzed, the conclusion is that the relation of8characteristic parameters have three cases, including linear correlation, independence and nonlinear correlation, and the relation of4combination parameters isn’t independent. So it is necessary to screen identification indexs, and there are two bases, one is the relation of parameters, i.e., its correlation coefficient is smaller, then it should be elected as a identification index. The other is the correct rate of parameters, i.e., its correct rate is bigger, then it should be elected as a identification index.Moreover, according to the overlapping features of parameters’theoretic distribution curve between two kinds of fiber, this paper has calculated the critical value, two types of errors, overlapping area and correct rate. And the correct rate is a basis of parameter significance and identification accuracy, but also constitutes the weight coefficient of identification formula.Finally, based on the relation and differences of distribution among8characteristic parameters and4combination parameters,4kinds of multi-parameters Bayesian classification mode are used to identify samples. Their results are basically consistent, which prove that the parameter screening is important and the way of screening is effective. In order to improve identification accuracy, this paper puts forward a optimal combination, i.e.,9parameters identification model, its largest absolute error of sample1,2,3and4is only1.3%and largest absolute error of sample5,6and7is4.9%, which is superior to the standard test A and B. This method is more accurate and objective. It can make up for the experiences of identification inspectors, week the influence of artificial factor and improve the identification accuracy.