Automatic Extraction And Inspection Of Railway Track Edge Based On Close Range Photogrammetry

Railway track is an important infrastructure for maintaining the high speed of the train.Maintaining a good geometric state of the railway track is one of the key factors for the long-term and smooth operation of the railway.In order to realize high efficiency detection of track geometric smoothness of high speed railway,close range photogrammetry is introduced to detect track state parameters.Digital image processing is used to match uniform homologous points in close range railway images.Based on the structure characteristics of the track in the image,the inner edge of rail is automatically extracted and matched.The three-dimensional coordinates of the rail edge point are obtained by a stable and efficient self-calibration bundle adjustment through hybrid program design.With these coordinates the geometric state of the track is calculated and the track measurement is carried out.Considering the high similarity and resolution of the railway track in close range images,the ORB algorithm combined with the partition matching method is used to obtain the uniform distributed homologous points according to the relative position relationship between the images,and the mismatching points are eliminated by the image kernel line.On this basis,pyramid images are used to accelerate the matching of three degree homonyms.The experiment shows that the partition matching has higher distribution uniformity than the whole matching.The ORB algorithm is highly efficient and accurate for partition matching,and basically meets the requirements of real-time matching.With the railway track image extends along the track in strips,the number of images is huge.A hybrid programming method is used to realize the self-calibration bundle adjustment.The data is stored by the sparse matrix,and the text files are used to exchange data for a stable and fast calculation process.Experiments show that a series of methods can improve the efficiency of adjustment and occupy less system resources.In order to solve the problem that the inner edge of railway track is not prominent in digital image,a strategy is designed including the coarse edge detection and candidate edge selection layer by layer from the track bottom to top for accurate recognition of track feature.The experimental results show that the accuracy of automatic extraction of the inner edge line of the orbit in the left rail image is about 1.5 pixels,and the accuracy is about 1 pixels in the right track image,which verifies the reliability of the automatic track edge algorithm.Based on the above theories and methods,a close range photogrammetry system for railway track edge recognition and detection is developed.The continuous 3D coordinates ofrail edge points is used to fit the spatial linetype of track,and calculate the track smoothness parameters.The results of the calculated data on the experimental section show that the close range photogrammetry technology can be applied to track static detection effectively and is expected to become a new technology for rapid detection of static state of railway tracks.