Abstract:This paper presents a road threshold segmentation method that taking into account spatial gradient information and color information in order to solve the problem of uneven distribution of road gray values in high resolution images and the difficulty of segmentation of the different things with the same spectrum with simple thresholds, and it has different effects on the extraction of high-resolution road images from different sensors and different resolutions. The gradient value and gray value are combined to get the correlation connection factor, and the original road image is transformed into HSI color space to segment the gray consistency region of the road, and the segmentation results are verified by gradient value. Finally, the validation results and the segmentation results of correlation connectivity factors are fused to remove the non-road information, and filter the non-road information to extract the accurate road. Using images of different sensors and different resolution to experiment. The results show that the algorithm can well solve the above problems and strong adaptability. By comparison with the existing method, the superiority of the proposed algorithm in precision is verified.