Research Of Gauge Detection Algorithm Based On Image Processing

Track irregularity mainly includes track longitudinal level, along-track, track level, tracktwist irregularity, and gauge etc a series of track irregularity parameters. Among them gauge isone of the basic track irregularity parameter and it has always been one of the necessaryparameters to measure ever since the track inspection car developed. If the gauge can notbe measured precisely, it will lead to weighty rail and urban rail hidden trouble, forexample derailment and aggravated wheel-rail wear. In order to implementation measuredprecisely, high efficiently, and non-contact. Gauge detection is still an important researchof the train department in every country to be in face. This paper after introduces thedeveloping background and the research status quo of the automatic detection system ofrailway gauge, the detection of gauge combined computer binocular vision with imageprocessing is put forward.The detection based on binocular vision is one kind of novelnon-contact detection technique, which developed from the computer vision realm,primarily integrated the image processing technique and regarded images as the carrier andmeans to detect or transfer the information.In this paper mainly researches the following aspects:（1）Camera imaging model and the basic theory of calibration based on pinholeperspective model of camera, linear and nonlinear two methods of calibration. In the visionmeasurement system, point reconstruction by different methods is put emphasis on.Compared the methods one is world coordinate built on the calibration board, the other ison the camera optical center, and using the light plane to point reconstruction. The rangingaccuracy on two different world coordinates and the measurement methods using lightplane with binocular vision system are compared. Found the method which is fit for thesubject, and the measurement error reasons are analyzed.（2）The image preprocessing methods and two different spatial filtering methodsduring the image processing are researched.The methods of image segmentation are analyze and compared.For some disadvantages in gray image segmentation, the imagesegmentation algorithm is put forward based on RGB.The algorithm segments image by Rcomponent is combined with difference image method.Using the algorithm has extractedthe binary image of rail cross-section and gauge measurement matching point.And has gotthinning image by Hilditch. The extraction of gauge measurement matching point hasreached sub-pixel.Finally, the author summarizes the accomplished main tasks in this paper andprospects this algorithm for gauge detection.This paper key works include:（1）Comparision point reconstruction on two kinds of world coordinate and usinglight plane.（2）Comparision and analysis many kinds measurement schemes and rangingaccuracy.（3）The image segmentation algorithm of rail cross-section and gauge measurementmatching point.