Выделение калибровочного шаблона и определение координат его реперных точек

Abstract

An algorithm of calibration pattern’s feature points detection was proposed for stereo vision systems calibration tasks. The principle of this work is the analysis of the form of feature points and their spatial arrangement. Test results shows a significant detection in conditions of noise and rotation relative to the camera in a wide range (up to 45 degrees) without loss of accuracy. That is achieved by the following detailed calculations and statistical analysis (geometric or spatial) of feature points. When using the pixel accuracy of the coordinates of points of the calibration pattern in the plane of projection proposed algorithm provides an average error of 1.2 pixel.