Railway Gauge Detection By Using Image Processing Technology

With the continuously constructions of high-speed railway and passenger dedicated lines, the speed of trains has been becoming higher and higher recently. Accordingly, railway transportation safety has been becoming issue. So researching and improving of railway gauge detection technique has great social value and economic value because it affects transportation safety directly.The detection technique of fixed-point contact is always used to detect the static railway gauge in China. But it is not acceptable for busy main lines to spend several minutes or more on detecting single section. It is urgent to develop a new detection equipment that can obtain the data of railway gauge quickly and correctly and complete the detection task. This demand can be satisfied by non-contact measuring method because the technique obtains the data of railway gauge by the means of processing the images which were acquired by image sensor. Although being faster, more exactly and more laborsaving than fixed-point contact method, non-contact measuring method was not given enough attention.Two CCD image sensors and image acquisition cards were used to acquire the frontal image of train in this paper. Before image acquisition, the cameras were calibrated respectively. In the process, Marr vision theory was applied to solve the calibration matrixes, lens radial distortion parameter and polar line equations. Images were acquired after calibrating in the condition that the internal and external parameters of cameras were keeping invariant. Then images were processed by many methods including smoothing, sharpening, edge detection successively. The next matching of images combined of calibration parameters and polar line equations. Similarity method, relaxed iterative algorithm and Sectional Method were used in matching. Calculate spatial position of object with the data of matching. Spatial position and gauge standard were compared to verdict overrun or not after spatial position had been calculated by matching result and the object which is overrun were signed.The tests show that the system in this paper can calibrate the cameras correctly and acquire clear and right images. The system also provides many methods of preprocessing and can choose reasonable methods according to the qualities and features of original images. The system can match left and right images correctly and can calculate and recognize the overrun objects correctly. Detecting the basic structure gauge for standard gauge railway correctly is realized.In view of the test conditions, the system only achieved a non-contact static detection. This paper is mainly about the digital image processing and machine vision technology in the railway gauge detection. With the development of sensors, as well as high-speed processors, the application of image processing to detect gauge railway has a more broad prospects.