Circle recognition and parameter measurement methods with the similarity constraint on the slope curve of the tangent lines to a contour

Abstract This paper presents an effective method to extract the circle information employing the geometrical feature constraint with a wheel rim image as an example. This process utilizes the segmentation method of maximizing inter-class variance to binaryze the original wheel image by a global threshold. In order to discriminate the real circle contour from the other irrelevant parts in the initial recognized contour, we construct the slope curve of the tangent lines moving on the initial contour and compare it with the slope curve of a perfect circle. The deduced similarity constraint comprises the increasing feature of the tangent function curve and the membership degree of a partial contour belonging to the real circle. The membership degree is defined by the ratio between multi-partition integrals of the discontinuous curve and the sum of the integrals. The homologous contour of the partial curve with a membership degree larger than 30% is considered as a component of the real circle. The least square method is performed on the improved circle equation to obtain the optimal solutions. The experimental effects prove that the outlined method has the potential to recognize circle as well as validity and appropriateness for the optical inspection and vision-based measurement.