Research On Attack Angle Detection Method Of Wheel-rail Based On Digital Image Processing

With the development of railway to the trend of high speed and heavy load,the interaction between the wheel and rail becomes more intense,which can cause wheel-rail abrasion,fatigue and derailment.Therefore,the measurement of wheel-rail contact parameters plays an important role in monitor locomotive safety and stability,It will cause the status of wheel-rail connection to be complex and an increase in angle of attack when the train passes through large ramps and small radius turn areas which can bring serious abnormal wear.As a important parameter in wheel-rail contact status assessment,wheel-rail attack angle play a decisive role not only in the safety and comfort aspects,but in wheel-rail wear measurement.This paper takes two scenarios of train goes straight and turns under as a research object,and uses different digital image detection methods to achieve wheel-rail attack angle detection on different basis of scene characteristicsWhen the train goes straight,there is snake movement which can cause displacement of wheel-rail,roll angle and attack angle.The paper uses the CCD sensor and linear laser which are installed on the train bogie to form image detection device.In the same time,it must make sure the laser light shines on the rail surface and with the wheel movement on the same line,so it can indicate the direction of operation of the wheelset.Take the image of the rail surface with a CCD camera to acquire laser line and track edge images respectively.In addition,passing through image filtering,feature detection,line fitting and other steps to determine the track edge line and laser line position,Thus,the attack angle can be calculated by means of the angle between the two characteristic lines.When the train is turning,the different curvature of the track causes more complex wheel-rail contact condition,apart from this,large attack angles and large displacements often exist.This paper proposes a wheel-rail angle of attack detection method based on graphical geometric mapping model,This method is based on the imaging principle of camera and fixes the camera to a specific position on train bogie.Building a geometric imaging model to form a single correspondence between wheel geometry and spatial position in the image,by which can transform angle detection to measurement of the minor axis of the ellipse.We build mathematical model to calculate the relationship between deflection angle and image ellipse information in numerical value and get the attack angle from the wheel shape.To ensure the accuracy of track edge detection,the paper use Generalized Interpolation Fourier Transform accelerate Radon Transform,and adjust minimum interpolation error to adapt GIFT parameters.In addition,aiming at the complex background and noisy wheel images,we optimize the ellipse detection algorithm based edge curvature and regional searching with the established mapping model to achieve the detection of attack angle in the margin of error.Finally,a simulation test was performed to demonstrate the accuracy and effectiveness of the two methods.