Research On Geometric Structure Feature And Track Fastener Location Algorithm Based On Spike Center Point Location

As an important part of fasten the rail on the sleeper to maintain the stability of the track,the loss,broken and displacement of the track fastener will lead to the risk of derailment when the train is moving,which will seriously affect the safe operation of the train.Therefore,the defect detection and maintenance of the track fastener is particularly important.This paper researches on the key technology of fastener positioning in track image.The main work is as follows:Track image pre-processing.The collected source image has problems like complex image information,poor image quality,much noise interference and uneven illumination.Operating the source image directly will make the overall time-consuming of the algorithm too high and the result is not ideal.In view of the above problems,through the experimental researches on the related technologies of image pre-processing,this paper proposes an image pre-processing method based on image scaling,graying,filtering,binarization and edge detection,which has effectively reduced the amount of algorithm calculation,lower the influence of external factors such as noise and environment,meanwhile,it has laid a foundation for improving the positioning performance of components and parts of track.Research on spike positioning.As an important track component for fixing fasteners and backing plate,the accurate positioning of spike is the premise of fastener defect detection.At present,there are few research results on spike positioning technology,and the existing algorithms have some problems,such as low positioning accuracy,high time-consuming and so on.This paper presents a spike center location algorithm based on EDLines.Firstly,based on the edge image obtained by image pre-processing,the edges of track spike in the image would be characteristic of roundness after being corroded and dilated.Then,by means of the improved Hough transform circle detection algorithm,the rough area of the spike was located and expanded,and the image of spike area is roughly extracted;Then,by analyzing the performance of the existing eight traditional line detection algorithms,The EDLines algorithm is used to detect the lines in the spike area image.Finally,used the straight-line screening algorithm and the regular hexagon fitting algorithm which proposed in this paper to accurately fit the spike hexagon and calculate the spike center point.Fastener positioning research.As an important part of fastener defect detection,the accuracy and real-time performance of fasteners positioning need to meet high requirements.Mostly,traditional algorithms use the global edge features of the image to locate the edge of the backing plate,combined with a priori knowledge to locate fasteners,which is limited by the image quality,poor anti-interference and poor performance,while the way of using deep learning to locate fasteners is limited by the number of image samples.By analyzing the geometric feature relationship between track components,this paper proposes a geometric feature fastener location algorithm based on the location of the center point of the spike.the positioning accuracy of the new algorithm is 99.33%,the precision is 0.997,and the speed is29.8 fps,superior to the algorithms compared.