Study Of Corner Detection In Camera Calibration

Camera calibration as the first step of the computer vision research has become one of the hotpots in the computer vision filed. One of the challenges of camera calibration is the precision of corner detection. In this paper, the main purpose is to research the corner detection algorithm in camera calibration.According to the study of the traditional corner detection algorithms, the common disadvantage of them is the sensitivity to noise. In this paper, due to the peculiarity of morphological affine invariability as partial differential equation, AMSS algorithm can be used to remove the noise, properly preserve the image features details with better effect of processing. So the AMSS scale space is used as the filter to solve this problem. After simplifying the AMSS equation in partial differential equations, the original location of the corners can be recovered with sub-pixel precision according to the evolution of the maxima of the curvature across the scales. We apply this technique of corner detection to calibrate a camera system using a calibration object. Appling the obtained location of the corner to the camera calibration experiment, and the performance of the proposed corner detection algorithm is analyzed with the experiment results.The experimental result shows that applying the corner detection algorithm based on AMSS scale space into the camera calibration can produce a good calibration paremeters. The error between the obtained corner location and projection of 3D position is less than 0.5 pixels, which means that the proposed algorithm has the property of high accuracy.