Research On Computer Image Recognition Method For Down Fibers

As the excellent properties of goose fiber, as well as many similarities of the appearance of morphological structure and chemical composition with duck fiber, so the falsified phenomenon that duck down instead of goose down or mixed with duck down in goose down and feathers are becoming increasingly serious recently, which makes a fast, accurate and effective method to identify goose and duck down become a top priority, meanwhile, detection of goose and duck down fiber have been recognized as the most difficult issues in the field of down fiber testing. Now, down the types of identification are testing through the microscope with the naked eye by personnel. This method requires a lot of training and practice experience, the discriminant subjective, inaccurate, time-consuming and labor-intensive.In this paper, the computer image processing method to identify goose and duck down fiber was studied, which includes:(1) Through a detailed description and characterization of the second order down-filament structure, the microscopic structure of the down were qualitative analyzed. First introduced four characteristic parameters of the quantitative characterization of the differences.we get the judging character that are fit for the automatic recognition of computer and provide for semi-automatic recognition judgment method.(2) Study the methods of image Processing of the image of down. Methods of image enhancement, image gray level change and image smoothness are mainly researched and we get the suitable method for the image of down. After researching the image segmentation, we get the binary image of down with OSTU algorithm.And binary image erosion, dilation, refinement and morphological processing.(3) Classification extraction the down image after the pre-treatment, then the nodes which have been recognized will be matched each other and the distance between the matched nodes and node density were calculated. With discriminant analysis method to establish discriminant function to finally determine the types of down.(4) The whole project of the system is proposed and through the semi-automatic recognition to assist down recognition, the software of down category recognition system is designed. In the end, under Windows XP,we develop a system with friendly interface and simple operation with MATLAB of the GUI development tools.In our experimental conditions,20down samples were image processing by using this algorithm, the results were consistent with the semi-automatic identification, which illustrate that the proposed method of goose and duck down fiber identification is effective. The subject of the image processing technology to introduce the field of down fiber. using image processing techniques to identify the down fiber so that the down identification to the objectification and quickly development direction.