Abstract：Research purposes：In order to improve the speed and efficiency of the detection for high-speed railway track, the close-range photogrammetry is introduced to examine the geometric regularity of railway track. The accurate image coregistration is the key process for track photographs orientation. The track photographs have few image characteristics, single color information, and high similarity of image grey value. So track image edge extracting and matching is relatively difficult. The method on extracting and matching track image edge for vehicle-borne close-range photographs was proposed in this paper. Based on the grey feature and its change information between track photographs, the coordinate geometric mapping model for homologous points of track edge was derived, and accurate image matching was carried out.
Research conclusions：(1) The cross-track coordinates for homologous points of track edge was cubic function model, and the along-track coordinates was linear model. The transformation relationship at homologous points was established, automatic and accurate image matching was carried out. (2)The transformation relationship which acquired from one image pair can apply to the other images with same acquisition conditions. (3)The accuracy and reliability of track image matching were validated from the experiment, which indicates that it can solve the matching difficulty, and it is a promising technique for examining the static geometric regularity of high-speed railways track with close-rang photogrammetry.
