A fast image segmentation algorithm for detection of pseudo-foreign fibers in lint cotton

Detection of pseudo-foreign fibers based on Automatic Vision Inspection system is crucial to improve the classification accuracy of foreign fibers in cotton, and justify the quality and grade of lint. Pseudo-foreign fibers are dried cotton leaves, seed-coats, barks, grass and dust which mixed into cotton during production, they usually have relative small sizes and large quantities compared to the foreign fibers, so it is hard to detect and segment them from images. In this paper a novel algorithm is proposed to solve the problem. 110 images with pseudo-foreign fibers are tested, and the result showed that both spots and slender shaped pseudo-foreign fibers can be segmented clearly with average speed of 0.67 s, which can meet the requirement of online cotton trash detection.