Research On Rail Corrugation Detection Based On Three-dimensional Structured Light

With the rapid development of high-speed railway technology,the requirements for the smoothness of railway tracks are getting higher and higher.Rail corrugation is a longitudinal irregularity of railways.Domestically,the detection for rail corrugation is mainly manual,while being time-consuming and labor-intensive.The detection accuracy and reliability of commonlyused contact detection equipment such as corrugation instrument would greatly reduce as time of usage increases,failing to effectively and precisely detect the rail corrugation.Recently,with the development of information technology,more and more innovation of computer vision and three-dimensional structured light technology are being introduced into industrial applications.Applying computer vision and three-dimensional structured light technology to the field of railway inspection can greatly accelerate inspection process,while leading to a more precise result.The problems of manual detection such as low efficiency and precision can thus be solved.Therefore,it is of great practical significance to apply three-dimensional structured light technology into the research and analysis of rail corrugation.Rail corrugation is a wavelike periodic irregularity at the top of the rail in the longitudinal direction.At present,railway engineering departments widely use contact corrugation detector for corrugation testing.The data is essentially obtained through the static two-point string measurement method.The method is conducted by measuring the longitudinal fluctuation of the rail track between the two reference points,and the longitudinal detection section is relatively single.In contrast,the three-dimensional structured light technology can obtain the complete three-dimensional point cloud data of the rail.Constructing a traversal plane perpendicular to the rail cross section and parallel to the vertical plane,traversing the top plane of the rail to obtain several equally-spaced longitudinal sections at the top of the rail,data regarding the longitudinal detection sections would thus be enriched.At the same time,since the three-dimensional structured light technology is non-contact detection with high precision,the longitudinal crosssection data obtained is more precise.Therefore,this paper uses a binocular structured light system to scan the rail with the purpose of obtaining the original point cloud data of the rail.In this paper,based on the rail point cloud data obtained through the binocular structured light system,the preprocessing of point cloud is further carried out,including the comparative analysis and experiment of different filtering and impurity removal,segmentation,and recognition algorithms,in order to determine the preprocessing flow and corresponding algorithm suitable for rail point cloud.Based on the on-site detection needs,the relevant detection specifications and the characteristics of the rail per se,the registration process of PCA rough registration based on the rail top and ICP accurate registration based on the rail waist is determined,and the registration precision obtained is verified through experiments.Also,due to the limited size of the single point cloud and the limited range of corrugation that can be detected,multiple point cloud stitching is required.In this paper,based on the automatic stitching of the scanning system,the ICP registration process is further added,thus the stitching precision is improved and verified.By traversing the longitudinal section,several longitudinal sections of the rail top are obtained.Combining the detection principle of the corrugation detector and the calculation principle of roughness,we raise two corrugation detection analysis algorithms: one based on chord measuring method and density clustering and the other based on roughness.Through the laboratory sample testing and on-site testing,we compare the data with the one obtained through the rail corrugation testing,finding out an efficient and cost-effective method of rail corrugation detection: using chord measuring method and density clustering algorithm to detect corrugation with short-medium wavelength in single point cloud,and combined with the method for roughness to detect corrugation with medium-long wavelength in multi-splicing point cloud.