A corner detection algorithm and evaluation method basing on measurement

A sub-pixel detection corner detection method based on fitting is presented, and mathematic algorithm of the method is in accordance with sub-pixel edge detection, improving the precision of the image measurement. In this paper, based on the geometry that the corner must be the point on the edge and intersection of two edge or many edge points, an algorithm based on the sub-pixel edge to get the intersection is presented. Then the edge equation is acquired by use of the sub-pixel edge detection algorithm in the sub-pixel edge area, the sub-pixel corner position is found by center of gravity method. And a new evaluation method is also presented, whose criterion is the precision of CCD calibration. In the experiment, comparing to the Bouguet's corner detection algorithms using the new evaluation method, the method of this paper is more precise. Therefore the algorithm is more effective in the field of measuring the dimensions in industry.