Close-Range Photogrammetric Imagery Matching And Modeling For High Speed Railway Track

The track of high-speed railway is the important ground infrastructure that supports the normal operation of railway system. Because the surface of tracks is often affected by the long-term interaction effect between wheel and rail, the regular geometry of track is easily changed. In addition, the subgrade and ground subsidence are also the influencing factors. The geometric deformation of track may lead to safety risk for the train running. So how to monitor the geometric state and stability in the railway operational phase becomes an important problem to be investigated.The common method of checking the geometric state of railway track is based on the technological system that uses track inspection vehicle and the total station. In order to improve the efficiency of measuring track regularity for high-speed railway, a new method of close-range photogrammetry is proposed in this thesis. The close-range photogrammetry method can be used to collect the railway track images and carry out track images matching to obtain homologue points. The image orientation process can be used to obtain the coordinates of homologue points which provide the basic data for the absolute orientation and aerial triangulation. The ground coordinates of track points can be solved to calculate the geometric regularity parameters of tracks.This thesis presents a second derivative operator of Harris feature points. It takes second derivative of image grey values to pick up feature points. Meanwhile, the track image is divided into blocks with varying window size of interest value comparison. The adaptive threshold determined by the number requirement of feature points is employed to extract the interest points.On the basis of track image features extraction, a matching strategy based on coarse matching of pyramid image layers and fine matching from least square method is proposed in this thesis. The matching strategy first makes pyramid layer images for coarse matching with the correlation coefficient of track images. Then the least squares algorithm for fine matching with the coarse matching results as initial values is used to seek homologue points.The process of track image matching can provide the necessary homologue points for relative orientation of consecutive track images. The relative orientation process is performed with track model coordinates extracted. The connection of stereo model is carried out through relative orientation parameters of consecutive images. The photogrammetric experiments is conducted with track images collected from the ballast track in the testing field and ballastless track in Shaoxing section of Hangzhou-Ningbo high-speed railway. The testing results show that the proposed second derivative Harris operator can not only greatly increase the efficiency of extraction of track feature points, but also improve the number and accuracy of feature point extraction in comparison with the traditional Harris operator. The derived feature points present uniform spatial distribution, which avoid effectively the feature points clustered and data redundancy. The proposed image matching strategy based on pyramid layer coarse matching and fine matching from least square can realize track image matching with grey similarity. The number and accuracy of homologue points from the proposed matching method can meet the requirement of relative orientation of close-range photogrammetry. The experiments validated the potential of close-range photogrammetry applied to track regularity measurement for high-speed railway.