Research On Key Technologies Of Intelligent Image Processing For Catenary Defect Detection Of High Speed Railway

Pantograph is a key component installed on the top of high-speed trains to obtain electricity from catenary.The high speed railway uses the pantograph to contact the contact line to obtain the power needed.The running state of pantograph directly affects the safety of train running.A high-speed train in operation is inevitably to encounter foreign body invasion,such as birds,plastic bags and dirty lens,etc.Foreign body invasion will affect the flow quality of catenary and trigger serious catenary accidents.If foreign bodies can be detected in time and corresponding protection measures can be taken,the operation safety of high-speed railway will be greatly improved.Based on the existing high-speed train video surveillance system,the pantograph video of high-speed train is captured by a camera installed on the train.Based on the basic theories and models of machine vision,this paper studies the detection algorithm of moving objects intruding into the pantograph and the detection algorithm of dirty intrusion lens.Real-time foreign body intrusion detection and analysis are carried out on the collected pantograph video,which realizes the accurate detection of moving objects and lens dirty.The main research results are as follows:(1)Aiming at the motion characteristics of pantograph invaded by objects,a moving object detection model based on inter-frame difference is designed.Aiming at the false positives caused by stationary objects,a moving object detection model based on gray scale,motion vector and edge feature was designed.The detection method based on motion vector table is further studied to improve the real-time performance of the algorithm.Through this moving object detection model,false positives can be excluded and moving object can be correctly detected.(2)A detection model based on cumulative difference is designed for the invariable location of dirt intrusion.an information entropy smudge detection algorithm combining gray scale and structure is designed to realize the smudge detection under the condition of poor video consistency.(3)All abnormal samples and normal samples at different times and scenes were taken as test sets to test the moving object detection model and dirty intrusion detection model respectively.The results of moving object detection model show that the correct recognition rate of birds invasion is 100%,and the false detection rate is 0%.The correct recognition rate of plastic bag hanging pantograph was 83.3%,and the missing rate was 16.7%.The accuracy of the dirty intrusion detection model is 100% and the recall rate is 99.6%.The experimental results show that both the moving object detection model and the dirty intrusion detection model have high accuracy and recall rate.It can effectively detect the foreign body invasion in the process of train operation and give an alarm to road man,therefore,the driving safety is improved.