Aiming at the problem of low precision of rivet size detection

an improved Zernike moment algorithm is proposed to detect the sub-pixel edge of rivet self-adpptive method is used to replace threshold of fixed step. First

in order to detect the diameter and length of the rivet

six regions of interest（ ROI） are constructed and the coarse positioning edge is extracted by the Canny operator. Then

the step threshold of the Zernike moment is adaptively determined according to the difference between the rivet and the background in the ROI

and the sub-pixel edge point is extracted from the coarse positioning edge. Finally

the rivet structure information is introduced to define straight linear cluster

the extracted sub-pixel edge points are fitted

and the rivet size is calculated according to the straight linear cluster equation. Experiments show that the improved algorithm has higher precision than other subpixel methods

and the detection precision of rivet length and angle reaches 0. 01 mm and 0. 1°

respectively.