Bundle Block Adjustment Of Close-range Photogrammetry For High Speed Railway Track

The smoothness of railway track may become gradually weak under the influence of long-term train load, roadbed settlement and geological environment after the high-speed railway is put into use. In order to ensure the safety and smooth of railway track, it is necessary to strengthen the detection of track geometry and other infrastructure. In addition, with the increased speed of train operation, it is necessary to improve the efficiency and accuracy of track smoothness examination.At present, the track smoothness detection is usually carried out with the combination of total station point. This method has the disadvantages of limited efficiency and heavy field operation workload. In this thesis, the close-range photogrammetry is used to do single track strip adjustment and double tracks adjustment to obtain the coordinates of homologue points. The initial value of exterior parameters can be calculated by using the homography matrix. The parameters and coordinates of homologue points, which can be regarded as the initial value for track orientation and modeling, are employed in bundle block adjustment of track images to obtain the accurate value of image parameters,The process of close-range track image strip adjustment for high speed railway shows that the single track strip adjustment has high efficiency and less memory usage, so this adjustment method can be used to provide initial value for bundle block adjustment. Then in the processing of bundle adjustment, Levenberg-Martquardt algorithm and damping strategy can ensure the convergence speed and stability of matrix operation, and improve adjustment efficiency. For the large sparse matrices in adjustment, the method of block compression row storage can save computer memory consumption and accelerate the speed of matrix operation. The adjustment method can be applied to close-range image processing for railway track examination.There are two experiments in this thesis including the Pengzhou branch line in Cheng-Guan railway and Jianyang section in Cheng-Yu high speed railway. The close-range track images of the two experiment areas are used to calculate digital camera parameters, the interior and exterior orientation elements of images, and geometric coordinates of homologue points. The results show that iteration times is reduced, memory usage is greatly reduced, and time of adjustment is obviously shortened after introducing the Levenberg-Martquardt algorithm and damping strategy. And the geometric accuracy of the self-calibration bundle adjustment can reach 0.8 mm in the along-rail direction. It can meet the requirement of track smoothness examination. The experiments show that the bundle adjustment with Levenberg-Martquardt algorithm and damping strategy in close-range photogrammetry technology can be used to image orientation and modeling for track geometry examination.