High-precision Dynamic Rail Section Detection System Design

Non-contact detection technology for rail profile is the key in estimating rail abrasiveness and line irregularity and evaluating the rail line quality, thus is of great importance for railway maintenance and safety. As rail transportation is developing toward high speed, heavy load, and high density, traditional manual and low-speed detection methods cannot fulfill the need. Therefore, developing high-speed detection technology for rail profile has great significance.Based on the analysis of existing rail profiling techniques, this thesis proposes the overall structure and implementation method for the high-speed kinematical detection system of rail profile. The system utilizes the triangulation photogrammetry, and obtains rail profile with the help of line-type laser and high-speed digital camera. The real-time acquisition and compression of rail profile images is then realized by using FPGA-based SOPC technology, and high-accuracy positioning and calibration is acquired by utilizing Gaussian regression based sub-pixel image processing, which greatly improves the system accuracy. The resulted system accuracy is better than 0.06mm, with a measurement speed of 200 frames per second.The focuses of this thesis are on the detection scheme of rail profile, system design for high-speed image acquisition and compression, and sub-pixel image processing and calibration. The thesis also presents theoretical analysis and experimental verifications for overall measurement speed and accuracy of the proposed system. Experiments on the prototype show that the system meets the requirements on both measurement speed and accuracy, and is ready for further tests on line.