Identification of cotton contaminants using neighborhood gradient based on YCbCr color space

In view of the harm of cotton contaminants, the image processing method based on machine vision provides a good solution to eliminating the foreign fibers. After analyzing cotton contaminants characteristics adequately, this paper presents a practical recognition algorithm for cotton foreign fibers images. By converting the collected images from RGB to YCbCr space, gray image Y, blue color component Cb and red color component Cr are obtained, and green color component Cg is constructed. In doing so, it is helpful to distinguish gray information and color information effectively and helpful to use them respectively. For brightness images Y, the image sharpening algorithm based on neighborhood gradient is proposed after contrasting arctangent histogram and automatic histogram enhancement. The algorithm fully considers effects of gray variation level of pixel itself, neighborhood pixel gradient and the distance between neighboring point and the centre point. Iterative threshold segmentation is applied to binarize the image Cr, Cb, Cg and the gradient-sharpened image respectively. The optimum threshold obtained by iterative algorithm is free of the effect of noise. With the increase of iterations the evaluating gray value is tending to the true value. Performing fusion technology to the four binary images can fuse gray image and color image information together. In order to eliminate the noise points and false feature, this paper proposes the shape feature recognition method based on minimum area, the ratio between perimeter and area and objective center distance. Finally the fused image is conducted morphological processing. It is proved by experimental results that the algorithm is effective and credible.