Study On High Speed Rail Profile Measurement System Based On Structured Light Vision

Periodic assessment of the state of rail is a necessary mean to maintain railway line and ensure the safety of the railway line. Traditional mechanical contact measurement approach, whose disadvantages lie in low speed, heavy workload, high risk and poor reliability, cannot meet the high speed and high density traffic needs. With its large measuring range, high precision, high speed and good flexibility, non-contact measurement based on structured-light vision has become the mainstream of track inspection technology. However, the speed and accuracy of the existing domestic track inspection devices of this kind is not high enough. This paper presented a high speed rail profile measuring system, made theoretical and experimental studies related.Achieving high speed and high density rail profile sampling is the prerequisite of obtaining reliable rail state information. This paper designed a high speed rail profile image acquisition system based on a multi-sensor parallel sampling, demonstrated feasibility of the system, and gave a method based on the measurement speed, sampling interval and the performance of the camera to determine the number of sensors needed. When the system measuring rail profile at high speed, huge amount of data will be generated, which will cause data transmission problem. To solve this problem, rail profile image region of interest (ROI) extraction algorithm based on field programmable gate array (FPGA) was designed, and related experiments were conducted. Experimental results show that the algorithm can achieve real-time ROI image extraction, and ROI extraction can reduce the data amount of communication effectively.Based on theoretical analysis, the mathematical model of measurement system was established. Using a planar target, the measurement system was calibrated, so the unknown parameters in the system model were determined. Based on the characteristics of rail profile image, a sub-pixel center of structured light stripe extraction algorithm was designed. To alleviate adverse effects of track inspection car vibration on measurement system, the mathematical model of the impact of vibration on the measurement system was established, and a data calibration method based on profile curve feature point and iterative closest point (ICP) algorithm is proposed. Dynamic data is calibrated preliminarily by identifying the profile curve feature point, and then the transformation matrix between the preliminarily calibrated data and the standard data is determined by ICP algorithm, finally the data calibration is completed. The impact of vibration on measurement was simulated in the experiment, then the deviation value between the calibrated data and the standard data was measured at both vertical and horizontal wear measuring points, standard deviation respectively 0.088mm,0.085mm, take 0.99 confidence level, precision respectively ±0.227mm, ±0.219mm. Experimental results show that the method has good repeatability precision, can be applied to engineering measurement.