Stereo Line Matching Based on the Combination of Geometric and Intensity Data

This paper presents a line matching method for the reconstruction of 3D line segment, based on geometric and intensity information. Line matching is a difficult task in the area of stereo vision, since it relies on incomplete line data. A set of line attributes is used to match a pair of 2D lines. Line attributes consist of intensity information and geometric information including length, orientation, and endpoint position. For the implementation of the proposed matching method, we employ multi-threshold technique, which provides flexible and efficient matching capability. The experimental results show that the rate of the correct match comes up with 96.12%, which is very hard to be achieved in the complex scene used in the experiment.