Compared with traditional global threshold algorithms,which can not extract the pavement crack exactly under non-uniform illumination,an improved image segmentation and classification method is proposed to overcome the drawbacks.Firstly,a new image segmentation method combining sobel and maximum entropy is presented and the solitary noise is reduced based on the pixel geometrical shape of the image.Secondly,a pattern classifier based on RBF neural network is designed to recognize different cracks according to the geometrical shape differences of different cracks.The experimental result indicates that compared with other image segmentation algorithms,the method proposed in this paper has a good ability to extract the image edge,and a strong ability to suppress the noise in the image.Moreover,it can achieve classification accuracy.