To detect circles fast and correctly has wide application prospect in the field of computer vision and automatic inspection. The conventional Hough transform has the limitation of using too much computation and memory space, and restricted extracting parameter. Aiming at this problem, the authors bring forward a new detection method in allusion to the real-time requirement of detecting circle in simple, image. When the image does not intersect each other, they transform the image into binary image and then perform edge detection to it. After that the chain code of the image is acquired. Basing on that, the circularity of the image and radius of the circle more quickly are calculated and a satisfactory result is gotten.