Research On Speed Measurement And Location Based On Image Recognition And Its Application In Rail Transit Training System

With the rapid development of high-speed railway and subway in our country, the demand for rail-related talent is expanding. In order to meet this demand, domestic colleges are dedicated in training the signal, transportation and other related professional students, and preparing rail transit training system to rapidly promote the students’ practical operating ability. Train control plays a pivotal role in the functioning of the rail transit training system. In order to control the train effectively, it is necessary to monitor different trains’real-time position and speed. Algorithms of train positioning and velocity utilized in the existing training systems have their own advantages and disadvantages. It could make an important practical value to seek for the less costly and time-effective algorithm of train positioning and velocity for the construction of rail transit training system. This paper combines the project requirements of multiple track traffic training system, the foreign studies in the related field, the author derives a series of practical algorithms for the orbit identification, the train positioning and recognition, which could get the accurate real-time position and speed of the train to improve the train control technique.The main contributions are as follows:(1) This study refers to the field condition of the practical, the author collects data of different focal length and shooting angle by testing different video collection equipment and its different installation locations and angles. A suitable combination for the field application is found, which contributes to an appropriate processing method for collecting images;(2) Based on the analysis of common rail identification method, by comparing the experimental outcomes, the author proposes an orbit identification method which is suitable for the application. This method is a combination of the morphological basis theory, the top-hat transform method and the principle of HOUGH transform. The finding achieves good results of the track recognition;(3) An improved train positioning algorithm is proposed. Firstly, the algorithm derives the approximate location of the train by moving object detection, and then, generates accurate coordinates of the center of the train by regional correction. Finally, it could work out the trains’location and speed from the orbit identification data. The experimental results show that the algorithm can effectively get the exact position and speed of the train;(4) This study raises a train number identifying method which could be applied in practice. This method extracts the train center area, then, it gets the binary image with corresponding tilt correction and recognizes the number on the top of trains in the means of template matching. As a result, it gets the corresponding numbers of trains. Combined with the previous outputs, the method can obtain real-time locations of different trains and provide a basis for the train control;The experimental results show that the solution proposed by this paper can well satisfy the requirement of monitoring train positioning and velocity in the rail transit training system.