Supervised identification algorithm on detection of foreign fibers in raw cotton

Foreign fibers accounted for a small proportion in cotton, but there is serious impact on the quality of textile. Foreign fibers are removed by hand, which is low efficiency. Generally, the methods of fixed threshold are used to identify foreign fibers in cotton, but high speed flow of cotton is easy to result in fluctuations on light, the color of captured images will be changed accordingly, then misidentification possibility will be increased. But the suitable amount of sample libraries are used in the identification algorithm of supervised classification, which eliminate this defect to meet the requirements of accuracy and real-time. In this paper, according to the character of image gray of foreign fibers in cotton, and mathematical model is established. Further, important image features are enhanced by image processing, foreign fibers´ characters are drawn. At last, Euclidean distance and k-nearest neighbor classification are adopted in identification algorithm, and finally foreign fibers are identified.