Automatic Identification Algorithm In Monitoring Foreign Body Intrusion Limit Of High-speed Rail

Intrusion of foreign objects in railway tracks is an important cause of train accidents.Once foreign bodies invade the railway track,it will inevitably cause great damage to the operation safety of trains and even threaten the safety of passengers.Therefore,the detection and tracking of foreign objects in orbit is crucial to driving safety.With the rapid development of high-speed trains,the safety of trains has paid more and more attention.High-speed trains travel very fast.Once they collide with orbital foreign objects,they will inevitably cause incalculable losses to trains and passengers.Therefore,it is crucial to identify orbital invading foreign objects and foreign body tracking and even foreign body movement trajectory trend prediction to improve train driving safety and early warning capability.Based on the extensive collection,reading and analysis of domestic and foreign railway intrusion limits of the latest literature and results,the paper studies the basic principles and methods of image de-noising,background modeling,foreign object detection,target tracking and trajectory prediction.The calculation model and implementation algorithm involved in railway foreign body intrusion are given.Aiming at the problem of orbital foreign body invasion in this paper,an improved image de-noising algorithm is presented.The hybrid Gaussian model algorithm is used to identify and detect the target.The BP neural network is used to modify the IMM algorithm to reduce the nonlinear error in tracking and location.The Elman neural network model is built to predict the foreign object trajectory.Through experimental analysis,the model and method can effectively solve the problems of detection,tracking and trajectory prediction of railway foreign body intrusion limit,and achieve real-time monitoring.Firstly,various commonly wavelet de-noising algorithms are introduced.A threshold adaptive optimization de-noising algorithm based on wavelet transform is proposed and simulated by summarized the theoretical basis of predecessors.These provide a theoretical basis for subsequent experiments.Then,in view of the identification and detection of orbital foreign objects in the complex environment,a pixel filtering idea based on wavelet transform is presented to improve the GMM and construct a background model.In particular,an improved wavelet transform algorithm is proposed,which combines adaptive Bayes Shrink threshold estimation method and template pixel replacement method to perform pixel filtering to achieve background adaptive update.In order to solve the problem of the low accuracy,the reason of Kalman tracking filter linearization error is theoretically analyzed.The BP neural network algorithm is presented to correct the tracking process of IMM algorithm and achieves orbital foreign object tracking.In order to accurately predict the moving target trajectory in the maneuvering state,the previous tracking data is the training sample,and the target trajectory prediction model is built by Elman neural network to realize the prediction of the orbit foreign object.Finally,it is verified by experiments that the de-noising method proposed in this paper has better de-noising effect for image information with Gaussian noise.The simulation of GMM based on wavelet transform is improved.The experiment shows that the average prospect false detection rate of GMM is reduced by 24.94% under normal weather,and the average prospect false detection rate for complex bad weather is reduced by 33.76%.The BP neural network is used to modify the nonlinear error of the IMM model,which shows that the tracking algorithm has better tracking effect than the existing algorithm and reduces the tracking error.The established Elman nonlinear neural network prediction model has better prediction accuracy.