Research On Railway Track Anomaly Detection Based On Image Processing

With the continuous construction of railways,more and more people travel by trains on rails,and the safety of trains on rails becomes more and more important.For trains on the track,the track is the basis for the safe running of the train and affects the stability of the entire railway system.Therefore,it is necessary to detect abnormalities in railway tracks.The traditional railway track anomaly detection mainly depends on manual completion,the detection efficiency is low,and it is easily affected by manual subjectivity.By using image processing technology to design inspection algorithms,you can quickly and accurately detect and locate defects.This article first introduces the relevant background and significance of railway track anomaly detection,and shows that railway track anomaly detection has important guiding significance for the later track repair,thereby ensuring the safety of train operation.In addition,this paper also analyzes the current status of railway track anomaly detection at home and abroad,and the detection methods for various anomalies that occur in different periods in railway track anomaly detection.This paper mainly focuses on block anomaly detection,wave grinding anomaly detection and positioning design algorithm of longitudinal joint of track plate.And put forward the following innovative solutions:1)Track interception for block anomaly detection.Different from the traditional use of the average threshold method or vertical projection,this paper first uses the average threshold method to locate the light band boundary,then uses the slope threshold method to determine the linear type,and uses different track boundary processing methods for different linear types to achieve the positioning of the track boundary.Use the track boundary to intercept the image for block anomaly detection.Next,the gray scale compensation is used to eliminate the influence of the gradation of the gray value of the light band,and the abnormality is detected using morphological calculation.2)Abnormal detection of corrugation.Different from the traditional direct analysis of the Fourier transform waveform,this article filters out high-frequency noise signals,and at the same time adds a reverse Fourier transform to restore the waveform.The restored smooth waveform is used for orbital wave grinding anomaly detection.Finally,the anomaly detection is realized by analyzing the continuity of the wave grinding inflection point.Specifically,the edge of the optical band is detected by the edge,and a partial image from the center of the optical band to the boundary of the actual left track is cut out.This partial image is subjected to Fourier transform to obtain a waveform,and then the high-frequency part of the waveform is filtered out,and is restored using an inverse Fourier transform.And continuous analysis of the inflection point of the recovered waveform to find the final abnormal position of the wave grinding.3)For the anomaly detection at the longitudinal joint of the track slab,this article provides early positioning work for the anomaly detection at the longitudinal joint,reducing the detection time and improving the efficiency for the subsequent anomaly detection.First,by setting the reference straight line,the Hough straight line is used to detect and position the lateral reference.A certain range of images near the reference at the interception point is used for secondary detection.Using the characteristic that there is a clear reserved line at a certain distance between the fasteners on the same track board,the straight line is detected by the secondary Hough line detection,and a straight line processing algorithm is designed to filter out the extra straight line and use the specific distance between the straight lines Position the longitudinal connection area of the actual track.In view of the detection and positioning of the above design algorithms,this paper uses actual image data to check the actual algorithm design of block anomalies,wave grinding anomalies,and positioning of longitudinal joints of track plates,and analyzes various detection results.The recognition accuracy of the final block anomaly was100%,and the recall rate was 96%.The accuracy of wave anomaly recognition was99.66%,and the recall rate was 99.5%.The positioning accuracy of the track joints is99.5%,and the recall rate is 100%.The experimental results in this paper are good and the efficiency is high.Experiments show that the designed algorithm satisfies the detection and location of track anomalies.