Abstract: For detecting the common circle objects, a circular hole size detecting system based on computer vision was developed. Using mixed noise filtering，fuzzy and genetic algorithm, the edge of circular hole image was extracted from the collected parts image. The radius and area of circular hole were measured accurately by interpolation method for roundness measurement. The testing results show that the average error of the system is 0,010 5 mm. It is far less than 0,0273 mm, which is the average error of manual measurement system. It can meet the precise requirement in industrial applications.