Accurate dimensional measurement of spatial circle arcs based on stereo vision

This paper presents an effective method to accurately reconstruct and measure the radius and centre of spatial circle arc based on binocular stereo vision. Firstly, the sub-pixel centers of curvilinear structures are detected with Steger algorithm, which improves the extraction accuracy of the curve center position in the image. To effectively reconstruct the 3D point coordinates of the measured parts using a geometrical matching, a strategy is proposed, which calculates the intersection of the epipolar-line and a line segment on the right image, while the line segment is determined by the two points that are the first and second points closest to the epipolar-line, and distributed on the both sides of the epipolar-line. After reconstructing the 3D points of the spatial circle arc, a novel fitting method of spatial circle arc based on conditional extremum is proposed and the dimensions of the spatial circle arc can be calculated easily. Experiment results show that the presented method is more accurate and robust than traditional ones for high accuracy measurement of spatial circle arcs.