Image points matching between multiple uncalibrated images has always been an important task in Close-rangephotogrammetry and difficulties are often experienced. Epipolar geometry constraint method is considered to be aneffective way to solve this problem and it has been studied in detail. However, large amount of applications have shownthat epipolar geometry is hard to be recovered accurately due to image distortion and positioning error, which leads tonumerous mismatches when image points are distributed intensively. In order to solve this problem, an alternative imagepoints matching technique to the epipolar geometry constraint method is investigated, which matches the image pointscombining with a bundle adjustment process and using their corresponding spatial collinear lines intersected on 3D spaceinstead of finding on 2D epipolar line. In this paper, the matching principle and process of this method are described indetail. Experiment results and comparison with epipolar geometry constraint method are displayed, which verifies theflexibility and effectiveness of this method.
展开▼