The Study On Through-type System For Wheel-set Detection Based On Structured Light

With the rapid development of social economy and rail transit in recent years,the passenger volume of high-speed railway and urban rail transit is increasing.As an important part of running gear,the wheel-set not only bears the weight of the train,but also directly contacts with the rail and bogie braking unit,providing power and braking force for the train.In addition,the bad working environment constantly affects the service state of wheel-set and the safety of train operation.Because of the unique shape of the flange tread of wheel-set,the geometric dimension of wheel-set has been detected by a way of off-line manual measurement for a long time.The detection efficiency is low,and it is difficult to evaluate the service status of wheel-set timely and accurately.Therefore,it is of great significance to study the on-line dynamic measurement system of wheel-set.Based on the measurement method of linear structured light,this paper studies the hardware and software of the through-type geometry measurement system for wheel-set,and studies the algorithms of image processing and feature point extraction involved in the detection process.Based on the introduction of the principle of structured light generation,the measurement process of linear structured light and the coordinate system commonly used in the mathematical model of linear structured light measurement,the direct linear transformation model of camera based on perspective projection model is deduced.Combined with the distortion correction of camera lens,a nonlinear model of linear structured light measurement is established.Aiming at the determination of unknown parameters in nonlinear model,the calibration principle of camera internal parameters and light plane equation is studied.According to the measurement model of linear structured light,the coordinates of the light strip center line in the wheel set light strip image captured by CCD camera are converted into two-dimensional point cloud data of the wheel-set flange tread.The accuracy of the extraction of the center line coordinates affects the accuracy of the subsequent algorithm.In this paper,based on the selection of Hessian matrix algorithm to achieve the sub-pixel extraction of light strip center line,combined with region growing algorithm,the shortcomings of Hessian matrix algorithm are improved from image segmentation,and operation speed.In the meantime,the image denoising and image enhancement in stripe image preprocessing are studiedThe line fitting algorithm based on random sampling consensus algorithm and the arc fitting algorithm based on radius constraint determine the rotation parameters and translation parameters to achieve profile matching and complete wheel-set point cloud data acquisition.Through the method of local fitting,the coordinates of feature points used in the calculation of geometric parameters of wheel-set are accurately extracted.The algorithm studied in this paper is applied to the through-type wheel-set online detection system,and the system is tested.The experimental results show that the detection accuracy of the detection system meets the inspection requirements of the maintenance department for the geometric parameters of wheel-set,which offers it great application prospects.