A Fast Epipolar Rectification Method Based on Fundamental Matrix

Construction machines are in a state of vibration during operation, which causes the misaligned horizontally of image pairs taken by the binocular camera installed on machines. Therefore, epipolar rectification must be carried out before stereo-matching. We propose a novel approach for epipolar rectification, which can be divided into two steps: the rectification of the left image and the corresponding rectification of the right image. The rectification of the left image consists of the transformation of mapping the epipole to infinity and the shear transformation. The rectification of the right image can be achieved on the constraint of zero vertical disparity. The images can be resampled by the transformation matrix, which realizes epipolar rectification of the image pairs. The experimental results show that the proposed algorithm has good accuracy under the premise of minimal geometric distortion, and the linear least square method is used only once in the algorithm, which shows the fewer computing cost and better stability.