Automatic corner detection of checkerboard based on LSD

To improve the robustness of current automatic corner detection algorithms, a novel algorithm based on Line Segment Detection (LSD) was proposed to extract the corners automatically. First, the LSD algorithm was used to process a checkerboard image to obtain all lines including checker edges. Then, the pseudo permutation of lengths and angles for obtained lines were done respectively to filter fake edges. Furthermore, the neighboring endpoints of the remaining lines were combined, and the coordinates of the corners were optimized with the sub-pixel algorithm. Finally, an energy method was utilized to recover the chessboard's structure and the corner points were ranked at the same time. Experimental results indicate that method proposed here automatically detects corners in images with noises and shadows. The maximum locating error and average error for the corner coordinate extraction are less than 0.2 pixels and 0.15 pixels respectively as compared with those of modified Harris method. This method has a higher robustness and its locating accuracy is almost as the modified Harris method, which shows it is suitable for a real factory environment. © 2015, Science Press. All right reserved.