Detecting Foreign Strands From Cotton Using RGB Saliency Map And Morphological Segmentation

Abstract — This paper presents an image segmentation method aiming at detection of foreign strands from cotton. Detecting foreign strands from the background is difficult due to their various shapes, size and colors of foreign strands. In this method, saliency features are identified. Color objects are identified by their color feature while gray objects are by brightness feature. Final foreign strands is obtained by fusion of the color and gray objects. But this method does not work well when it encounters situations of uneven thickness of the layers of cotton and the different colors and shapes of foreign strands. So, to get better result, in the proposed method, morphological edge detection is performed to detect the edge of the foreign strands. Also, an iterative thresholding method is used to segment the gray gradient images. The experimental results show that the foreign strands are detected with high accuracy than the conventional image segmentation methods