Color Target Detection Based On Color Constancy

Color constancy is one of the characteristics of the human visual system, in machine vision, color constancy algorithm can correct color images ignoring scene illumination, restore the original color of the objects, the use of this method can apply to many occasions with a good effect.In this paper we analyze the different classes of color constancy algorithms, and handle different types of color cast images:(1) We analyzed the color constancy of different color spaces, and found that the H and S components in the HSI color space is not sensitive to the shadows and specular highlights, we improved Canny edge detection algorithm based on this, by calculating the shadows and highlights quasi-invariant part eigenvalues and direction to achieve the color images of the road detection;(2) We analyzed several major algorithms in a single light color constancy algorithms: max-RGB, Grey-World, Shades-of-Grey, Grey-Edge algorithm comparison experiments confirmed that such algorithms due to illumination color misregistration image good color correction, and to achieve the detection of the skin area;(3) We analyzed several major algorithms of multi illumination color constancy algorithm such as SSR, MSR and MSRCR, comparative experiments prove that such algorithms underexposed due to color cast image image enhancement, which by MSRCR algorithm adaptive color histogram linear stretch improvement, can achieve a better color fidelity of the image enhancement and improve the brightness, contrast, information entropy of the images. We used this algorithm to realize the detection of the image of the colored traffic signs.