A sub-pixel detection method of riveting height difference based on optimized Hough transform was proposed for the detection of aircraft riveting height difference. Firstly, the collected riveting height difference image was enhanced by gray scale transformation method of point operation in spatial domain, segmented by the local threshold  segmentation method. Then, Canny operator edge detection was used to extract rough edges and the optimized Hough transform was used to extract sub-pixel edge, and the ring region of rivet hole and rivet head is extracted. Finally, the circle fitting was carried out with the RANSAC algorithm to generate the circle contour, and the average height values of the inner and outer ring regions were respectively extracted by the operator get_current_region_z( ). Then the height difference of the riveted surface can be obtained by the difference of the average height obtained by the function height_Z(). Experiments show that the detection method has strong sub-pixel accurate positioning ability, high accuracy, good stability and repeated measurement accuracy of ±10 μm.