Optical flow matching with automatically correcting the scale difference of tunnel parallel photogrammetry

Abstract

Using parallel photography to model tunnels is an efficient method for real scene modelling. Aiming at the problem that the accuracy of optical flow matching in tunnel parallel photography sequence photos is severely affected by the scale deformation of stereo images, a novel optical flow matching method with automatically correcting the scale difference of tunnel parallel photography stereo images is proposed from the perspective of imaging relationships. By analysing the distribution pattern of scale difference in stereo images, a model is obtained in which the scale difference of image points is symmetrically distributed radially on the image and follows a power function growth. Introduce it into traditional optical flow matching to correct image scale differences based on the model to improve matching accuracy. The mean square error of the optical flow matching after correcting scale difference in the experiment is less than 0.3 pixels, which is at least 34.3% higher than before correction and a maximum improvement of 45.5% in the experimental results. The research result indicates that the proposed optical flow matching method with automatically correcting the scale difference has a significant effect on improving the accuracy of tunnel parallel photography image matching and modelling.