Research On The Key Technology Of Rail Surface Defect Detection In Rail Image

The rapid development of modern railway industry is faced with the challenge of the increasing speed,mileage and load of trains,which increases the safety hidden danger of railway infrastructure.Rail is an important part of railway infrastructure.Under the action of high-density operation,high load and external natural environment,the surface of rail will produce certain abrasion damage,which will affect the comfort and safety of train running.It is of great practical significance for railway operation and maintenance to timely detect the defect of the derailment surface and provide reliable maintenance data to relevant departments to ensure the reliability,safety and service life of the rail.The actual track image collected includes not only the track surface area,but also the interference areas such as sleepers,fasteners and gravel,etc.In order to reduce the difficulty of track surface defect diagnosis and reduce the search space and diagnosis time of track surface abrasion defects,accurate and rapid extraction of track surface area is required.In different degrees,the traditional method of orbit surface area extraction requires to set the width of the orbit surface in advance,assume that the orbit surface is in the middle of the orbit image and manually select the boundary,and has some problems,such as poor self-adaptability,light sensitivity,and the inability to extract the orbit surface completely when there are noises such as dust and mud at the rounded corners of the orbit head.In order to solve the above problems,a YUV space based greedy selection and slope detection extension of the orbit surface region extraction method is proposed.Firstly,the RGB orbit image is converted to YUV space,and its V component is extracted to reduce the interference of ambient light and noise.Secondly,the gray projection inversion curve of The V component is drawn,and the gray mean and median of the curve are used to divide the candidate orbital intervals.Then the greedy algorithm is used to calculate the maximum suborder and interval of the divided curve,and the track surface is extracted roughly.Finally,the slope detection extension method is used to extract the rail surface accurately,and the slope detection at a certain distance is carried out on both sides of the coarse-extracted boundary.The rail surface boundary is updated with the appropriate position where the deflection Angle is greater than the set threshold value,so as to realize the location extraction of the rail surface area.After the orbital surface area is extracted,the orbital image still has some new features,such as uneven illumination,limited identifiable features,low contrast,and changeable reflection characteristics,which hinder visual detection.To solve the above problems,a background differential rail surface defect detection method based on defect proportion limitation is proposed.The method mainly includes five steps: track image preprocessing,background modeling and difference,defect proportional limit filtering,defect proportional limit maximum entropy threshold segmentation and connected area labeling.Firstly,a fast background modeling was carried out by combining the mean value and median value of the gray column of the track image,and the pre-processed image and the background image were differentiated.Secondly,the upper limit of defect proportion is truncated to enhance the contrast of the difference graph by using the feature of low defect proportion in orbit image.Then,this feature is used to improve the segmentation of maximum entropy threshold,global variable weighting of target entropy is adopted by adaptive weighting factor,and an appropriate threshold is selected to maximize the entropy value,so as to reduce the interference of noise,such as shadow and rust,while retaining the real defects.Finally,the connected area marker method is used to count the defect areas in the binary image after threshold segmentation,and the defect area lower than the rail damage standard is judged as noise removal,so as to realize the rail surface defect detection.