Abstract: Randomized Hough transform is an effective method for circle detection.But when dealing with multi-circle complex image，the random sampling will bring lots of invalid accumulation and calculation.This paper presents a fast multi-circle detection algorithm with randomized Hough transform.Three kinds of noises are deleted.To determine a candidate circle，the other two points are searched according to certain rules when randomly pick a sample point.The original image is used to affirm the candidate circle for true circle.Theoretical analysis and experimental results indicate that the algorithm can detect multiple circles more quickly than other algorithms.And it also has better application value.