The invention relates to the technical field of pavement detection methods, and discloses a pavement detection method based on a multi-dimensional information probability model, and the method comprises the following steps: 1, obtaining a plurality of images of a road segment needing to be detected through a camera device, storing the collected images in a memory, and building an image library; 2,selecting any image, extracting position information of pedestrians and vehicles in the image, and additionally adding position information needing to be extracted; 3, removing pedestrians and vehicles on the selected image according to the extracted position information of the pedestrians and the vehicles; and 4, calculating through a computer, and obtaining position information of pedestrians and vehicles according to the position information extracted in the step 2. The road surface detection method based on the multi-dimensional information probability model solves the problem that an existing detection method is only suitable for objects which do not have road signs, barriers, pedestrians and the like and affect computer exporting of information characteristic quantities of road surface chromaticity, colors, textures, lane lines and the like on a road surface.