The present invention discloses a SAR image road detection method based on ratio features. The problems are mainly solved that road edge positioning is not accurate and a false alarm rate is high in the prior art. The method comprises the steps of: 1) performing speckle reduction of an SAR image and extracting 9 texture features; 2) screening 3 texture features having the greatest contribution onclassification from the 9 texture features according to a Bhattacharyya distance; 3) calculating comparison ratio features R1 and similar ratio features R2 in the image point by point after speckle reduction; 4) constructing a road dictionary D1 and a background dictionary D2 by employing samples with results of the step 2) and the step 3); 5) solving one average difference value E1 of the road dictionary D1 and one average mean value E2 of the background dictionary D2 for each pixel point, performing classification of the pixel points through the difference values, and obtaining a preliminarydetection result; and 6) performing optimization of the preliminary detection result, and obtaining a final detection result. The SAR image road detection method based on ratio features can completely and clearly detect roads in the SAR image, and is suitable for detection of roads with different directions and different widths in the SAR image.