Research And Application Of Dynamic Measuring Method Of Rail Profile Based On Structured Light Projection

The measurement technology based on structured light is currently the research focus in the field of visual measurement and has been widely used in various fields of industrial measurement.The technology mainly uses a laser emitter to generate a light field with a specific pattern,which intersects with the measured object to form a three-dimensional profile section,and then the profile acquired by the image sensor is analyzed to obtain the depth information of the object surface.Chinese railway has been developed rapidly in the past decade.With the high-speed,heavy-duty and high-density operation of the train,the dynamic detection and analysis of the rail profile and its matching relationship with the wheels have gradually become a concern of the railway sector.However,under the dynamic environment,the undesirable illumination changes,noise interferences,train shaking,and shape changes of the target caused by the high-speed train which running on long-distance tracks have raised many new challenges.Therefore,how to realize the dynamic real-time and high-precision measurement of rail profile under the dynamic,high-speed and high-noise running environment is the research focus in this dissertation.The research content and innovation results of this dissertation are as follows:1)Based on the holographic principle of visual imaging,a geometric calibration method of laser line structured light systems using planar targets is proposed.The calibration method is used to capture the planar calibration target images at different positions,and the laser is turned on and off when getting the same target image;based on the intersection between the laser line and the target virtual line in the two images,the position of the laser plane is determined,and the projection/photography relationship between the line laser and the camera,also called the homography matrix,is further obtained;finally,the calibration of the structured light system is realized.Compared with the existing methods,the planar calibration target of this method has the advantages of simple fabrication,good flexibility in calibration process,high measurement accuracy,and good practicality.2)An algorithm based on deep learning is proposed to realize the recognition of different types of rail profiles,and a template-match driven temporal and spatial contextual tracking algorithm is used to achieve rapid tracking of the rail head laser stripe.This method effectively solves the problem of profile measurement when the trains pass the crossings,and realizes accurate positioning and fast tracking of many types of laser stripes;3)A fast,accurate and reliable method is proposed to extract the center line of the laser fringe image in driving environment.Firstly,the boundary-based connectivity analysis and deep learning segmentation algorithm are used to realize the semantic-level segmentation of the projected laser fringe,and then based on the method of segmentation smoothing and seamless splicing,high-precision extraction of the centerline of the laser fringe is achieved.This method can effectively remove the noise interference caused by track reflection,fouling,and foreign object obstruction on the laser fringe.4)A polynomial fitting method based on the truncated residual histogram is proposed to find the optimal fitting curve of the profile,which reduces the influence of noise on the profile fitting and proposes a dynamic window based on the curvature distribution feature of the fitted curve.The maximum curvature entropy interval search algorithm is presented to achieve the segmentation of the profile feature region,and then the alignment registration between the measurement profile and the standard design profile is achieved with the feature region as a reference.This method effectively overcomes the impact of train shaking and ambient noise on the target profile measurement.