Research On The Recognition Model Of Cashmere And Wool Based On Digital Image Processing

The research of this paper is based on the Research of Parameter Detection of Cashmere Fiber in China-Arabia Cashmere Cooperation Industry of China-Arabia university cooperation research program. The cashmere and wool fiber of Lingwu market were selected randomly as the research object, each 1500 sample image of two types’fiber is collected by digital image acquisition system, preprocessing fiber image and extracting eight characteristic parameters fibers with digital image processing technology. The focus is on building recognition model based on Bayesian classifier, BP neural network model, FCM and BP network model. Finally, this paper compare the three recognition model, obtain the best identification method, achieve high-precision automated identification of cashmere and wool fibers. The main research achievements are listed as follows:(1) This paper research the acquisition condition of cashmere and wool images, determine the appropriate optical microscope, and use the reflex lighting way, the sample should be extracted in accordance with ISO17751:2007 strictly. It gives the best fiber image preprocessed program:graying, edge enhancement, denoising, edge extraction and the morphology processing. There are 8 characteristic parameters (the fiber diameter, the scale height, the scale perimeter, the scale area, the relative scale perimeter, the relative scale area, the diameter to height ratio, the squareness)are selected for describing fiber properties.(2) This thesis propose a recognition model based on Bayesian classification algorithm, analyze correlation and the weights of cashmere and wool eight characteristic parameters. Through research, it selects relative flake size, fiber diameter and height scale to build three-parameter identification model, determine bayes formula, obtain simulation map.(3) Research on the BP neural network model to establish the BP network identification model, then obtain training results, which based on the number of nodes determine by measured data, and training rate and minimum error rate, and the training and hidden layer function. Also studied BP recognition model based on FCM algorithm, the data in the BP network training before entering the first cluster analysis, and the resulting output as input BP neural network training, obtain recognition results. Finally, comparison of two recognition methods by the error count and identification. The results showed that recognition algorithm based on FCM and BP network combining high recognition rate and lower error count.(4) The recognition model of FCM and BP network with more accurate recognition rates can reach 99.6%by counting error points, recognition rate of cashmere and wool. More followed BP algorithm is to identify the model, but this model can also be used for cashmere wool fiber automatic identification technology, its recognition rate is 99.3%. In which model-based Bayesian classification algorithm to identify the worst, recognition rate is only 93.4%.