| The continuous increase of train speed and axle load will aggravate the track wear.If the track is not detected and maintained in time,it will lead to accidents such as train derailment,which will seriously affect the safety of railway transportation.With the rapid growth of high-speed railway mileage in China,manual inspection and measurement of rail status is difficult to meet the needs of railway on-site maintenance because of its heavy workload,low efficiency and precision.It has become an urgent problem for railway engineering departments that how to accurately and quickly detect track wear and master its damage rule and wear changes.Thus,the track can be maintained in time to prolong the service life and avoid safety accidents.Under this application background,Structured-light measurement technology has been used in track wear measurement instead of traditional manual measurement methods because of its advantages of non-contact,high precision and rich measurement data.Track wear usually occurs in different parts such as rail,frog and switch rail,which can be divided into different types,such as vertical wear,horizontal wear,wave wear and so on.Rail can measure wear by obtaining rail profile for its wide distribution and simple structure,while complex structural components,such as frog,switch rai,and wave rail can not be measured accurately through two-dimensional cross-section profile,so it is necessary to obtain its three-dimensional shape.In this paper,according to the characteristics of structured-light vision measurement,the linear structured-light vision measurement technology is applied to the dynamic measurement of rail full profile wear,and the surface structured-light vision measurement technology is applied to frog,switch rail and other complex structures and shortwave wear measurement.The accuracy of the linear structured-light binocular vision dynamic measurement system for rail full profile wear is easily affected by ambient light,on-site calibration,vibration in the process of measurement and error analysis compared with the standard model.However,the difficulty of the wear measurement of track 3D shape by surface structured-light stereo vision lies in the data preprocessing of the measured point cloud and the comparative analysis with the standard 3D model.In order to use structured light vision measurement technology to accurately measure track wear,the main research contents of this paper are as follows:(1)The high precision calibration method of the linear structured-light binocular vision measurement system in the railway site is studied,which is the basis for accurately obtaining the full profile of the rail.It is not easy to adjust the equipment on the railway site,so the system calibration is realized based on the free plane target.By obtaining the characteristic points produced by the intersection of the light plane and several checkerboard plane targets at different positions,the spatial position of the light plane in the camera coordinate system on both sides of the rail is fitted,and then the unilateral profiles of the rail measured by the cameras on both sides are unified to the same coordinate system,thus the full profile of the rail is obtained.However,due to the photosensitive characteristics of the camera and the influence of external factors,there will be a slight deviation in the profiles of the both sides of the rail.Therefore,according to the same characteristics of the rail head tread profile on both sides,the full profile rail data fusion is realized.(2)The accurate and fast method of extracting the sub-pixel centers of rail light-strip profile in the linear structured-light vision measurement system is studied,and then they are transformed into rail full profile through the system calibration parameters.The railway field ambient light and the rail head bright band will affect the rail light strip profile image quality,so the camera automatic exposure method is used to ensure the rail image quality,so that the light-strip profile occupies a certain amount of pixels in the image.On this basis,the sub-pixel centers of the light-strip are extracted.In order to ensure the accuracy and efficiency of sub-pixel extraction of the centers of rail light-strip profile,an extraction method based on image dynamic threshold segmentation and Hessian matrix is proposed to improve the adaptability of measurement to the environment and the speed of image processing.(3)The measurement error of rail wear caused by the vibration of track detection vehicle in the dynamic measurement process of linear structured-light vision measurement system is studied,and then the influence is eliminated from the hardware design and measurement algorithm.The vibration of the track detection vehicle leads to the deflection of the light plane,so the measured rail profile will be stretched horizontally and longitudinally relative to the standard profile.If the measured rail profile is compared with the standard two-dimensional rail profile,there is an error in the wear value.In the aspect of hardware design,through the dynamic adjustment of the spring compression device,the measurement unit of the track inspection vehicle is always close to the side of the rail.In the aspect of measurement algorithm,the two-dimensional rail full profile data is introduced into the 3D space and compared with the standard rail 3D model.The measured small arc centers of the rail web on both sides of the rail are used as the initial registration benchmark,and the rail web profile is used as the accurate registration reference to realize the accurate alignment between the measured rail profile and the standard rail 3D model.Finally,the full profile of the measured rail is projected to the plane perpendicular to the longitudinal direction of the rail to measure wear,which can effectively eliminate the problem of light plane deflection caused by vibration in the process of measurement.(4)The pre-processing of the track original measured point cloud obtained by the surface structured-light vision measurement system is studied,and the accurate registration analysis between the measured data and the standard model is analyzed.The track original measured point cloud has a large amount of data and is accompanied by a large number of noise points,so the European clustering segmentation is used to eliminate the noise points,and then the curvature-based point cloud adaptive sampling is used to simplify the point cloud while preserving the detail features of the point cloud,so as to complete the data preprocessing.The track 3D wear measurement is to compare the measured 3D shape with the standard 3D model,so it uses the accurate registration algorithm which combines the principal component analysis method and the nearest point iteration method.Compared with the common registration algorithms,this method improves the accuracy and efficiency.(5)The linear structured-light binocular stereo vision measurement is deeply studied,and the dynamic measurement of rail full profile wear is realized.On the basis of theoretical research,a linear structured-light binocular stereo vision measurement system is designed and manufactured,and the track inspection vehicle is carried out on the rail to obtain the full profile wear condition.For complex structures such as turnouts and rail wave wear,through the design and manufacture of surface structured-light stereo vision measurement system,the track 3D shape point cloud is obtained by static measurement and the track 3D wear measurement is realized.Through theoretical research and many field tests,the fast and dynamic measurement of rail full profile wear by linear structured-light binocular vision measurement system and the3 D measurement of rail wave wear,frog and switch rail wear by surface structured-light stereo vision measurement system are realized.The measurement systems can realize the extraction of works measurement parameters,verify the feasibility and practicability of the structured-light vision measurement method,and provide a new technical means for track wear measurement. |
PDF Full Text Request