Accurate Modeling Of Railway Tracks Based On MLS Point Cloud

The track is a key object in the railway system,and the precise modeling of the railway track can be applied to many fields,such as railway asset surveys,boundary analysis and track measurements.In the railway MLS,the laser scanning system is installed on the tail of the train to scan the track.Rapid acquisition of three-dimensional coordinate data of railway corridor with large area and high resolution is very suitable for the track modeling and track parameters measurement based on track model.At present,the main method of railway track modeling based on MLS point cloud is to use simplified parameter model of rail to fit rail point cloud.The track model established by this method has low precision and does not contain track parameters such as gauge and superelevation.In this paper,two aspects of research were carried out to establish an accurate model including track parameters based on MLS rail point cloud.1)This paper introduces the cylindrical neighborhood elevation difference with positive and negative signs,and proposes a MLS point cloud extraction algorithm for rail based on column neighborhood elevation difference.The algorithm introduces a cylindrical neighborhood with the normal direction of the top surface of the rail top instead of the traditional spherical neighborhood to calculate the neighborhood elevation difference of the MLS point cloud.The algorithm solves the problem that the spherical neighborhood used in rail point cloud extraction algorithm based on spherical neighborhood elevation difference can not accurately reflect the difference between rail points and sleepers,ballast points neighborhood elevation difference,which leads to low extraction accuracy.The optimal value of the algorithm parameters is determined by experiments,and the algorithm is compared with rail point cloud extraction algorithm based on spherical neighborhood elevation difference.The experimental results show that the proposed algorithm can correctly extract the rail point cloud of the whole railway line,and its F-measure is 88.73%.The optimal values of the three parameters of segment length,neighborhood radius and histogram group distance are 13 m,50mm and 10 mm respectively.Compared with rail point cloud extraction algorithm based on spherical neighborhood elevation difference,the recall is basically the same,the accuracy is increased by 59.57%,and the F-measure is improved by 42.68%.2)A matching algorithm between the track model considering gauge deviation and MLS rail point cloud is proposed.In the matching iterative process,the track gauge of the track model is continuously adjusted by calculating the distance between the left and right rail point clouds to the working edge of the track model,so that the gauge of the track model gradually approaches the true gauge of MLS rail point cloud.The model is precisely matched to the MLS rail point cloud.The problem that the left and right rails of the orbit model established by single rail model matching algorithm are not parallel and the matching precision and gauge accuracy of fixed gauge track model matching algorithm are low are solved.Through experiments,the matching results of the algorithm were tested by rail point cloud simulation data,and the algorithm performance was evaluated by matching accuracy,gauge error and rail parallelism.Comparing with single rail model matching algorithm and fixed gauge track model matching algorithm,the effects of gauge deviation,track superelevation,slope and guard point cloud on the matching results of the three algorithms are analyzed.The experimental results show that the matching accuracy of a matching algorithm between the track model considering gauge deviation and MLS rail point cloud is 0.16 mm,and it is not affected by gauge deviation,track superelevation,slope and the guard rail point cloud.Compared with single rail model matching algorithm,the matching algorithm proposed in this paper is basically equivalent in matching accuracy and gauge accuracy,and rails is relatively improved.Compared with fixed gauge track model matching algorithm,the matching algorithm proposed in this paper is basically parallel,and the matching accuracy and gauge accuracy are improved.