Track Geometric State Inspection Of High-speed Railway By Vehicle-borne Close-range Photogrammetry

The railway tracks are the infrastructure facilities that support the vehicles running, leading the vehicles to move smoothly and safely. During the long-time period of operation, the tracks may generate a variety of geometric distortion, since the tracks are affected by the long-term interaction effect between wheel and rail, roadbed subsidence and other factors. To guarantee high-speed train running safetly, it is necessary to strengthen the examination and maintenance of railway track for the good geometric state of track structure.This thesis studies the theory and method of track geometric state inspection of high-speed railway by vehicle-borne close-range photogrammetry. The image data can be obtained continuously by the close-range photogrammetric image acquisition system. The image matching algorithm is utilized to extract the homologue points of track images. The control points set on the rail surface are employed as the exterior constraints in the self-calibration bundle block adjustment processing for calculating the coordinates of track surface. The track geometric state parameters can be determinated through three-dimensional coordinate of the homologue points that are obtained by edge extraction and matching algorithm in the track inside edge.There are some significant characteristics for track close-range photogrammetry, such as high similarity of image gray and large amount of data. The ORB algorithm and nearest neighbor algorithm are employed to extract homologue points with the advantage of eliminating matching error and less redundancy. The experimental results show that the image points which are obtained by ORB algorithm have high accuracy and good spatial distribution with the almost real-time processing requirement.The two types of systemic errors often affect the image point coordinates in track close-range photogrammetry, which are interior orientation elements and lens distortion. This thesis calculates the interior orientation parameters, exterior orientation elements and distortion parameters by the self-calibration bundle adjustment. The experimental results show that the geometric accuracy of the self-calibration bundle adjustment can reach0.7mm and1.6mm respectively in the along-rail and transverse-rail directions, and1.3mm in the vertical direction. The results can meet the high precision requirement of track inspection.Considering the feature of less gray variation in track images, the Canny operator with the adaptive threshold and polynomial matching model are employed to extract the homologue points in the inside edge of track. The three-dimensional coordinate can be caculated in the track control network. The testing results show that the geometric accuracy of polynomial matching model can reach1.6pixels in the transverse-rail direction,1.3pixels in the along-rail direction. The results can meet the high matching precision requirement of homologue points.The three-dimensional coordinate in the inside edge of track can be used to calculate the track static geometry state parameters. This thesis calculated the static geometry state parameters for the straight line area including the rail gauge, height difference, and central line shapes. The experimental results show that vehicle-borne close-range photogrammetry method is promising as a new technology for measuring the track geometric regularity.