Research On Image Detection Method Of Rail Surface Defect And Rail Fastener Absent Pattern In Complex Environment

Keeping and maintaining the security of track is an important task to ensure the safe operation of the train,with the rapid development of railway construction in recent years,the railway transportation of China has presented the new characteristics of high speed,heavy load and large operation density,and the traditional manual inspection has been unable to meet the needs of railway detection at the present stage.With the great development of hardware and software in these years,the superiorities of machine vision based detection methods are becoming more and more obvious.In this dissertation,according to the machine vision detection technology,the defects on the surface of rail and the missing state of rail fastener are detected and the defects are classified by analyzing and processing the track images,on the basis of this,a portable defect detection system for rail parts is constructed to realize the integration of rail surface defect detection and fastener missing state detection.First of all,Gamma transformation method is used to improve the uneven illumination of images,and the wavelet transform is used to decompose these images,the weight of the image edge subband is increased in the process of image reconstruction to enhance the edge information of image.A method that combined columns gray accumulation and gray mutation statistics is used to locate the target area in ballast track images and ballastless track images respectively.The specific position relationship between regions in the track image is used to complete the location of the rail fastener area.Next,the membership function is combined with entropy theory to complete the segmentation of the defect area in the surface of rail,and the edge of the defect region is optimized according to the path ergodic method.In view of the special interference situation that residual water stains on rail surface,the water stains region is detected according the difference of gray value distribution between water stain region and defect area.The edge contour of water stain region is eliminated by using morphological processing and image difference.A simple classification of rail defects is achieved by extracting the geometric features of the defect area and using BP neural network theory.Then,according to the characteristic that the fastener as a standard workpiece has specific size and geometry,the initial detection of the missing fastener is completed by matching the edge curve of the fastener to the template image.Considering that the rail fastener may be covered by occlusions,image samples that containing and without fastener are established and a twi-difference is carried out on images to reduce the effect of occlusion on the detection.The minimum distance classifier is used to complete the detection of missing fasteners.Finally,according to the demands of actual track detection and the working principle of the machine vision based detection system,a portable defect detection system for rail parts has been preliminary designed,realizing the integration of different detection contents.Through the field experiment in the existing railway line,the experimental results show that the track detection system has a certain adaptability and can meet the real-time requirements of hand push detection,the track detection system can reduce the burden of worker to a certain extent and improve the efficiency of detection.