Research On 3D Inspection System Of Railway Track Using Structure Light

With the stricter requirements for the safety of railway transportation,the requirements for railway track inspection are also improved.Railway track is made up of many components,and how to carry out the inspection of each component has become an important factor affecting the railway safety.At present,the inspection of railway track is still manual mainly,with the increase of railway detection items,manual detection is is not able to meet increasingly stringent inspection requirements in terms of inspection efficiency,inspection accuracy,data processing and storage.In recent years,with the development of information technology,machine vision technology has been rapid development,the application of machine vision to railway track inspection can not only meet the accuracy requirements,but also improve the detection efficiency,so the study of 3-D inspection is great significance.In this paper,based on the 3-D imaging of structured light,has done the post processing of 3-D point cloud data.Through the study of shape characteristics of rail,seam rail,switch rail,presents a method of railway track recognition,to facilitate the subsequent processing of point cloud.Through the research on the section between the point cloud data and two point cloud registration methods are proposed based on the normal rail and based on cloud head feature,the advantages and disadvantages of the two methods are analyzed and compared,and a suitable method is selected according to the needs of field inspection;the inspection parameters of railway track are studied,and the measurement of parameters is realized in the detection system.By analyzing the inspection method of the rail corrugation,the wavelet analysis is applied to the inspection of the rail corrugation,and field test is successfully carried out.In the process of research and development of the measurement software,aiming at the problem of low efficiency of 3D point cloud boundary extraction,improving the computational efficiency based on OpenCL Heterogeneous computing;in order to solve the problem of 3-D point cloud segmentation,the kd-tree is used to optimize the segmentation results;the human-computer interaction of the measurement software has been optimized and interactivity is improved.