Checkerboard Subpixel Corner Detection Based on Local Area Intensity Response and LoG Response
Fuping Wang, Yang Luo, Guanzhuang Duan · 2024
To address the issue of imprecise subpixel localization of checkerboard corners, we propose a subpixel corner detection algorithm based on local area intensity response and LoG (Laplacian of Gaussian) response. Initially, we utilize the Canny edge detection algorithm to extract the edges from the original checkerboard image. Then, radon transform is used to detect the lines and extract the edge region between the intersections of the lines. Then, the edge pixel localization and edge direction based on the local area intensity response are estimated to produce the subpixel edge, the intersection points of which are determined as the initial coarse corners. Around the corners, we calculate local LoG responses in the feature space to interpolate the four local response peaks. Finally, an ellipse is fitting using the peaks and the final subpixel corner is located as the center of the ellipse. Experimental results validate the robust performance of our proposed algorithm, even in the presence of noise interference and affine transformations.