| Along with the acceleration of urbanization in china, the development of subway construction is especially rapid. It is expected that metro lines in china will reach 158 by the end of 2015, with the total mileage of more than 4189 kilometers. Rail is an important infrastructure for guiding wheel forward and supporting the trains. Outline dimension of rail is an important indicator to reflect whether metro vehicle can run safely. When deterioration change of outline dimension is severe, it will pose a threat to the operation of metro vehicle and even cause serious traffic accident. How to ensure the safe operation of subway lines has attracted much attention. In order to ensure operation safety of metro vehicle, the change of outline dimension need be detected quickly and accurately.Detection system installed in vehicles is designed to detect non-contact complete rail profile. Using four sets of the laser-photogrammetric sensor to acquire image, the system detects complete profiles of left and right rail respectively. The characteristics of laser-photogrammetric sensor are large measuring range, high precision and high speed. Based on these characteristics, high precision camera calibration method, image processing technology and numerical optimization method, the system can achieve non-contact complete rail profile detection.The laser-photogrammetric sensor calibration is a key factor of profile detection, because calibration precision affects measurement precision directly. Based on the analysis of the existing calibration methods, single camera calibration method and global calibration method are studied respectively. Four kinds of coordinate systems are established, and the transformation between different coordinate systems is elaborated. Based on laser-plane constraint, linear and nonlinear measurement models and the global calibration model are established. World coordinates and image coordinates of target points are acquired in checkerboard, and the parameters of laser-photogrammetric sensors are solved by the least square method, so the calibration experiment is completed.Complete rail profile detection system is required to process images acquired by four sets of sensors in real time. Real-time requirement is strong and data processing is big. So the center-line of light-strip extraction algorithm must meet the real-time detection requirement under different conditions. This paper has analysed image preprocessing and center-line of light-strip extraction algorithms to process images in real time. The gradient gravity algorithm is improved to extract center-line of light-strip, and the result is verified by the experiment.In process of dynamic measurement, the vehicle vibration has non-negligible influence on detection data of rail profile. Vibration compensation model is designed for floating up and down, left and right yaw and roll vibration.This paper uses static measurement data of rail profile as the reference, the detection system measure rail profile dynamically. Then translational and rotational vibration components caused by vehicle vibration is calculated. Vibration components are substituted into the vibration compensation model to reduce systematic errors.Windows is used as a development environment, and this paper designs software of complete rail profile detection system. Under different working conditions, mileage positioning accuracy and data precision of detection system is tested on-site. The test results show that mileage positioning accuracy and data precision of detection system are able to meet working requirement. |
PDF Full Text Request