Three-dimensional Contour Measurement Method For Surface Defects Of High-speed Heavy Rail Based On Point Cloud Feature

With the rapid development of China’s high-speed rail technology,high-speed rail has become an efficient,comfortable and safe way of travel.Therefore,the safe production of highspeed rail heavy rails has become an important issue to ensure the safety of travel.The detection of heavy rail surface defects is an important task for the safe production of heavy rails.At this stage,China’s domestic machine vision-based high-speed rail heavy rail defect detection method started late,most steel mills still stay in the stage of manual visual inspection,low detection efficiency,high miss detection rate,subject to the subjective factors of the test personnel,and there is a big hidden danger in the safety of on-site workers.At present,the machine vision detection method mainly uses two-dimensional image information for detection,but in practical applications,it is very limited by the image acquisition quality,and is affected by the detection area range,and the occurrence of missed detection or false detection is more common.In the research of this paper,in view of the problems existing in the two-dimensional image detection method,the method of 3D point cloud and line array image fusion is used to detect the surface defects of heavy rails,extract the 3D defects,reduce the missed detection,and improve the pseudo defects and 3D.The ability to distinguish defects and reduce the rate of false positives.The main research contents of the thesis include:(1)Designed the calibration method of linear array camera,and built a heavy-duty threedimensional defect detection platform consisting of high-precision linear array camera and mobile platform to collect the surface of heavy rail.(2)Studying the application of two-dimensional image registration method on the surface of heavy rail,providing initial registration for 3D point cloud registration,selecting the initial registration method based on phase correlation through experimental comparison,and the traditional point cloud registration method Compared with the better convergence,the complexity of 3D point cloud registration is reduced,and higher registration accuracy and computational efficiency can be obtained.The method is applied to the heavy-track point cloud registration,and good results are obtained.(3)To study the defect location extraction method of point cloud on heavy rail surface.In this paper,the point cloud method is used to segment the heavy-track defect point cloud based on the region growth algorithm,and the minimum bounding box is calculated to measure the size of the three-dimensional defect.In the heavy rail defect detection experiment,compared with the traditional two-dimensional image detection method,the method has better detection rate and lower false detection rate.The method of heavy-rail surface defect based on point cloud feature can cooperate with two-dimensional image detection method,which can reduce the influence of image quality and improve the computational efficiency of point cloud,which lays a foundation for the overall system construction.