Detection Of Color Fluff Based On Statistical Characteristic Values

As an important part of clothing products, with the rapid development of China and with the improvement of consumption level, down-filled clothing products showing the distinctive Chinese characteristics. Consumers are increasingly high demand for the quality of down products. In market, the use of white feather as filler to improve the quality of products become many down manufacturing enterprises to pursue. At present, only rely on artificial feather down color detection, sorting quality and production efficiency are not fundamentally guaranteed, so the market urgently needs a kind of automatic sorting equipment, sorting out the different color down rapidly to meet the high demand for pure white wool market, in order to improve the production efficiency and product quality. Therefore carry out the work of feather down color on-line grading automatically technology is of great significance.The main work of this paper was as following:The scope of the area and the gray scale range information was obtained. Firstly, a large number of down and feather color images was obtained by using CMC scanner, and then translated into gray image. By comparing the Otsu method and the maximum entropy method of image segmentation results, determine the maximum entropy method for image segmentation, Through morphological processing, two value image of color fluff is segmented, and color fluff images was marked and region intercepted, then color images of color fluff was obtained by multiplication between two value image of color fluff and color images of color fluff regions. Secondly, area and gray level information was extracted by statistic color image scope information of a large number of color fluffs. Analysis of different color area, R, G, B three channels to be four statistical results as a reference index to test.Detected and analyzed of the down. Scope selection of R, G, B three channel grayscale average three indexes, and the R,G, B three directions of the color image are to be analysis and the corresponding gray range is used as the threshold, Then the image segmentation, image fusion, morphological processing and color fluff marked and intercepted, eventually the color fluff was preliminary detected; then comparison of area scope of color fluff, less than the area of statistical reference index of color fluff was classified as normal down, for such a small area of the color fluff as normal feather in the subsequent detection processing, on the contrary, all other than statistical reference area of color fluff as the final detection color fluff. Then marked centre of gravity of color fluff, the color fluff was sorted.Finally, results analysis of statistical characteristic value of multi-objective combination testing different color fluff, through analyzing the testing results of all the different color fluff, macro analysis of the effect of color fluff, the detection analysis of different color fluff content and different thickness, to identify different distribution state of the color fluff are discussed, the discussion of ratio of missed and false detection, calculation and analysis of a large number of color fluff missed detection number and false number, from the results of the statistical data more accurately detection results, the calculation of perimeter range and draw ratio range, supplement the comparative index to detect down, in order to improve the precision and accuracy of detection of color fluff.The test results shows that this method for multiple different color fluff statistical characteristic values are extracted and obtained, through the combination of multiple objective statistical characteristic value to separate out color fluff and normal feather down, can effectively detect the color fluff, and the experiment proved that this method could be used in the online detection of color fluff.