The Algorithm Research On Feature Points Extracted From Digital Railway Curve

The reasonable dimensions and smooth alignment of ballastless is a key factor to ensure a high-speed, securely and comfortable travel for a train. Currently, the long wave and short wave irregularity detection and maintenance measures for the track have been more perfect. However, due to the impact of construction bias and measurement error, etc., the feature points of a flat curve, which were gotten by the horizotal coordinates of the center line and the digital polyline of superelevation, will have errors in distance. The existence of the discrepancies may affect the irregularity of track, badly affect the safety of travel for train. Accurate segmentation is the foundation of rail linear optimization and run speed increase for the existing railways. The automatic methods of segmentation for horizontal alignment had more deep and mature research results, but in terms of vertical alignment, the most commonly used methods were based on artificial segmentation, or artificial segmentation firstly and then determined by an iterative approach precisely. They were not only fail to segment automatically, but also had a low accuracy. To solve these problems, this paper explores the algorithm of automatic recognition of feature points for horizontal alignment of high-speed railway, and explores the algorithm of automatic segmentation for vertical alignment of existing railways. These algorithms will be used for the appraisal of smoothness in high-speed railway and linear optimization, run speed increase for existing railways.In terms of the recognition of feature points for high-speed railway, according to the data of track detection, the algorithms which are based on the coordinates of the center line and the digital polyline of superelevation are studied. According to the results of experiments, following conclusions are gotten:since the static detection of ballastless track has a large number of data, and the sampling interval is small, the traditional algorithms are difficult to express the change of outline on horizontal alignment effectively due to the measure errors. Therefore, they can not be used to distinguish the feature points automatically. The eleven-point method has a strong anti-noise ability. The approximate curvature which is calculated by eleven-point method can express the change of outline on digital curve effectively. This algorithm is used in the automatic recognition of feature points with a good effect. It is not prone to cause the case of wrong choice, leaking selection. The Douglas-Peucker algorithm can recognize the approximate locations of turning points for the digital polyline of superelevation effectively. Based on the result of approximate recognition, the locations of the gradient change points are reliable by iteration fitting algorithm. The difference between horizontal feature points and gradient change points of superelevation picked up by the two algorithms respectively can be used to analysis the irregulation of track. They can be used for the appraisal of track irreguration in new railways and the track detection in the runing railways.The algorithm of the automatic segmentation for vertical alignment based on the approximate curvature is studied. The experiment results show that approximate curvature which is calculated by eleven-point method can be also used in the automatic recognition of feature points for vertical alignment of existing railways. It has a more effective result than traditional methods. Based on the result of segmentation with iteration, the coordinates of the center of the rail circle can be effectively and precisely calculated by Radius Immobilization Least-square Circle Fitting Method (RILCFM). Thereby RILCFM may increasing the accuracy of the locations of feature points. These combined algorithms can effectively achieve the automatic segmentation of vertical alignment with a high accuracy. They will lay the foundation of rail linear optimization and speed increase of existing railways.