Research On Key Techniques Of Line-Structured Light Vision Measurement And Its Applications In Railway Dynamic Inspection

Recently, railway transportation has been developed rapidly in China. The incremental mileages of operated lines show that the railway fundamental network has been constructed. In the future, the development of railway transportation will be transferred to the long-term safe operation, maintenance and management by the large-scale construction. Rail, tunnel and overhead catenary system, as the most important infrastructures, which are not only to withstand all the extreme weathers, but also receive the dynamic loads of high-speed running train. So there are many factors affect the safety of the railways operation, and the field environment is quite complicated. The traditional manual and portable measurement for the railways infrastructure is obviously unable to meet the needs of the modern railway development. It is difficult to ensure the safety operation of railway transportation without the advanced techniques. How to use the advanced techniques for the railway infrastructure management becomes the most important problem.Line-structured light vision measurement technique, an important branch of modern visual measurement, recently receives extensive attentions. If the line-structured light vision measurement technique is applied into the railway infrastructure dynamic inspection, the advantage is obvious, while compared with the traditional manual static measurement. However, it is a comprehensive and application technique involving multiple disciplines of optics, optoelectronics, signal processing, pattern recognition, computer science, instrument science. The related contents are very broad and the practical problems are quite complicated. In different environmental conditions, the measurement results may be influenced by many different factors. Therefore, the research on the technique of line-structured light vision measurement is very necessary. In order to solve problems in railway infrastructure dynamic inspection, the following issues focusing on the line-structured light vision measurement are studied.(1) In order to achieve the line-structured light vision measurement, the measurement system should be calibrated in advance. To solve the problem of global calibration of multiple vision sensors is the basis of the success of the visual measurement system. The global calibration of multi-cameras and structured-light vision system is the key issue in the high-accuracy profiling measurement of large field of view, due to the wide distribution, lacking common filed of view and the complicated calibration model for multiple cameras. In order to overcome the calibration problem, the methods for the vision system calibration is divided into three types according to the spatial distribution of multiple vision sensors, such as public visual sensor calibration, visual sensor calibration and non public visual sensor calibration. The calibration of the vision system is based on the combination of the distribution of different visual sensors.(2) In addition, how to extract the stable and reliable calibration feature points in the structural light plane is the difficult problem for the light plane structural parameters calibration. To solve this problem, a method to calibrate the structural plane based on the combination of needle shaped target and displacement stage is studied. Through the combination of the structural plane and the displacement stage, a sufficient number of feature points on the structural light plane can be obtained. This method has solved the problem of feature point acquisition on the structural light plane and can significantly improve the calibration accuracy compared with the conventional calibration algorithm.(3) Accurate extraction of structural light image features is the key technology of structural light measurement. The traditional image processing operator can not meet the needs of structural light extraction. The methods based on curve fitting, surface fitting and engenvalue decomposition, the sub-pixel level accuracy can be obtained, but these methods generally existing large computation. In order to improve the performance of image processing algorithms in two aspects:recognition accuracy and processing efficiency, an improved Hessian matrix processing method based on ridge line tracking is studied. Through the combination of the ridge line tracking algorithm and Hessian matrix method, the improved method can avoid the large-scale image processing computation, which can still maintain the accuracy of sub-pixel identification and improve the image processing efficiency.(4) In reality, due to the spatial pose difference of multiple vision sensors, the images captured by the vision sensors will be affected by the illumination condition, the surface shape of the object, the optical path and the sensor's photosensitive characteristic. Interactive closest points (ICP) algorithm as the classic mthod in the different vision sensors data fusion, it is widely used in the off-line image data fusion processing. However, due to the different point sets registration need to exhaustively search, so it is difficult to meet the needs of reality. Therefore, a fast registration method based on k-d tree for the detection data of multiple vision sensor measurement is studied. By using the k-d tree fast search method, the efficiency of data fusion of different vision sensor results is significantly improved.(5) In order to reduce the measurement errors brought by the train vibrations, a compensation approach is investigated. The rail full cross-sectional profile measurement results utilizing the compensation approach are carried out. The influence of the vehicle's multiple degree of freedom vibration on rail profile is analyzed. The on-site experiment is carried out in the real railway line, and the experimental results demonstrate the effectiveness of the proposed approach.In this paper, the successful application of line-structural light vision measurement techniques in railway transportation dynamic inspection is carried out. On the one hand, the measurement accuracy, efficiency, and feasibility of the proposed method in the paper can be verified. On the other hand, it can also provide technical ways to solve the field maintaining problem of the transportation infrastructure.