The invention discloses a road image detection method. The method comprises: image denoising and enhancement are carried out on a road image obtained by a sensor; edge detection is carried out on the processed image to form a binary image with image feature peripherization; and Hough transform is carried out on the binary image and a peak point representing the road edge is selected from Huough parameter space, thereby completing road edge detection. According to the invention, Hough transform is carried out by using the binary image obtained after original image edge extraction, so that the Hough transform operation load can be reduced substantially. Meanwhile, a corresponding road screening constraint condition is put forward according to a practical road image, so that wrong road detection and identification information can be rejected effectively. Road edge identification can be carried out rapidly and accurately on the condition of interference factor influence existence. Complex road scenes are used and the method has certain practicability.