Research On Detection Technology Of Rail Surface Defects Based On Machine Vision

The railway system occupies an important position in the modern transportation system,because of the advantages of safety and speed,resource conservation,environmental protection,less influence by climate and natural conditions,and large transportation volume.It plays a very important role for the rapid economic and social development.Due to the increasing development of our country’s railway system,trains are running faster and faster,running more and more times,and carrying more and more loads.Under this background,the pressure and wear on the rails will continue to increase.In addition,the environment of the railway track is relatively complicated,and it is corroded by the external natural environment for a long time,so the surface of the railway track is extremely prone to produce defects.If it is not discovered and maintained as early as possible,the defects will deteriorate,which will bring hidden dangers to the safety of the train.In order to achieve the purpose of safe operation of the railway system and safe operation of trains,it is necessary to carry out effective and rapid detection of railroad tracks.The existing rail inspection technologies mainly include artificial visual inspection,eddy current inspection and other non-destructive inspection methods.These methods have problems such as low inspection efficiency,low inspection accuracy,and potential safety hazards.They cannot meet the fast and efficient defect inspection requirements of today’s railways.Therefore,based on machine vision and image processing technology,this paper designs a set of visual inspection and recognition system for rail surface defects.The main research work is as follows:First of all,it introduces the research background and significance of rail defect detection,briefly describes the current main rail defect detection technologies and their shortcomings,and analyzed the research status of domestic and foreign testing equipment and visual inspection technology at home and abroad.Secondly,a machine vision rail surface defect detection system is designed.Analyzed the inspection object and the common types of defects on the rail surface,as well as their causes and hazards,explained the overall function of the rail surface defect visual inspection system,designed the mechanical structure of the system and the imaging system,focused on the selection of the camera,light source,lens,lighting scheme and other hardware equipment in the imaging system and the realization of the motion control system.Next,the characteristics of the rail image are analyzed.In order to remove the ballast,sleepers,sand and other non-rail debris on both sides of the rail in the image,a rail region extraction algorithm based on the minimum value of the vertical projection of the gray value is proposed,which greatly improves Improve the efficiency of image processing.Research and implement the rail image preprocessing algorithm and defect detection segmentation algorithm.Aiming at the problem that the traditional threshold segmentation algorithm has slow segmentation speed and is easily affected by the noise of the rail image,which leads to the poor robustness of the algorithm,this paper proposes an improved threshold segmentation algorithm based on image difference and Pauta criterion.The algorithm performs image difference between the background modeled image and the filtered and denoised image to enhance the defect area of the image and overcome uneven reflections on the rail surface,rust,stains,shadows on the rail surface,and the external natural environment,etc.Then use an improved method based on the Pauta criterion to select the threshold of the differential image,and perform threshold segmentation on the differential image to segment the defect from the background.The experimental results show that the algorithm can quickly and efficiently segment the surface defects of the rail,the segmentation average running time is 6.19 ms,and the segmentation accuracy rate reaches 89.9%.Then,the classification and recognition algorithm of rail surface defects is studied.The characteristics of rail surface defects are analyzed,reasonable and effective feature vectors are selected as the input of the classifier,and the ELM-based rail surface defect recognition and classification method is designed and implemented.The ELM is trained to obtain a classifier,which can identify the types of rail surface defects.Experimental results show that the classification algorithm can effectively identify the types of defects on the rail surface,with a comprehensive accuracy rate of 95.76%.Finally,under the Microsoft Visual Studio 2017 integrated development environment,using the OpenCV software library,using C++ to write the algorithm code,using Qt to design the artificial interactive interface,design and develop a visual inspection software system for rail surface defects.The software system can effectively realize the detection and classification of rail surface defects.