Research On Several Important Algorithms In Color Recognition Of Vehicle-Body

Color recognition of vehicle-body is one of important auxiliary means in the vehicle recognition system. It plays important role in traffic survey, traffic control and vehicle management system. In practice, color of vehicle-body is often influenced by noise,illumination and so on, so accurate color recognition is somewhat difficult. A research on color recognition of vehicle-body is performed under this background.Images are often influenced by impulse noise in generating and transferring process. For color image, impulse noise not only affects image articulation, but also affects color of the image. So it is necessary to remove impulse noise before color recognition. According to image characteristic of impulse noise, a color image filtering algorithm based on impulse noise detection of rank- ordered pixels difference is presented in this thesis. In the noise detecting process for this algorithm, every channel of pixels in filtering window is ranked, and noise is detected according to the location condition of center pixel and the difference of adjoining pixels after being ranked. Based on the result of impulse noise detection, an algorithm of improved adaptive vector median filtering is then presented to remove impulse noise. Experimental results have shown that the performance of filtering impulse noise is improved remarkably by the algorithm.Recognition area location is the precondition of extraction for color characteristic information. By full utilizing of the correlation of vehicle-head and recognition area as well as axis symmetry of vehicle-head, the recognition area can be located through OTSU algorithm and twice projection.Color characteristic can be descripted by color space, and the difference between every color can be measured by chromatic aberration formula. Therefore, the selection of color space and chromatic aberration formula is important for the result of color recognition of vehicle-body. Based on the comparative research on different color spaces and the chromatic aberration formulas of different color spaces, a practical color space and chromatic aberration formula is found to recognize color of vehicle-body. Further on, a method of restraining the influence of illumination is presented exploringly. The precision of color recognition of vehicle-body and its robustness is improved by the method.