Description
Detection of circles in basic gray scale images is explored in this thesis. Impetus for the work derives from an interest in high speed detection of tires on cars. The circle detection process used is based on the pixel direction concept of J. F. Canny and the nonmaximum suppression process for contour tracing described by Neubeck and Gool. In the detection process, edge segments are evaluated to determine if they are part of a candidate circle and the center and radius of the circle is computed. The edge threshold property of Canny edge detection is not used. The algorithm has been tested on a group of 30 images of cars using Matlab. The image set includes identical mages having different image intensities. Edge images of tires are compared with images obtained using Canny edge detection. Useful results for detecting circles in images are found. VHDL implementation of the extraction algorithm is discussed. Synthesis results indicate that the design code can perform evaluations at a 30 frame per second rate. Memory usage is small compared to that required for the circular Hough transform. The time taken to determine the center and radius of a circle, and memory usage by the proposed algorithm, are discussed.
