Research On Detection And Pre-warning Of Foreign Objects Invasion On Railway Track Based On Object Tracking

With the rapid development of high-speed rail technology in our country,the speed of high-speed rail continues to increase,the operating mileage continues to increase,and high-speed rail traffic safety is also facing new challenge.Studies have shown that foreign object intrusion is one of the main factors that threaten the safety of high-speed trains.Therefore,the detection and tracking technology of invading foreign objects is particularly important.On the basis of the existing detection technology,this thesis is combined the detection technology with the tracking algorithm to realize the real-time monitoring and pre-warning of invasive foreign objects along the railway.The research contents of the thesis are as follows.Firstly,familiarize the theoretical basis of foreign object detection and research the detection method of invading foreign object.Consider the change of illumination in railway monitoring video and the slow ghost elimination of Visual Background Extractor(Vibe).Vibe algorithm is improved to quickly eliminate ghosts and effectively overcome the interference of strong light.Thereby,the accuracy of foreign object detection is improved.After two sets of comparative experiments,the detection precision and recall rate of the optimized Vibe algorithm are higher than the original algorithm.Secondly,study the tracking method of intrusion foreign object.In complex railway scenes,the Kernel Correlation Filter(KCF)algorithm cannot effectively deal with object scale change,image blurring and non-rigid deformation of foreign object.Through weighted fusion of multiple features and building branch scale models,the improved algorithm can better deal with the above complex situations and achieve accurate tracking of invading foreign objects.In view of the defects of the observation model of the Correlation Filter algorithm,that is,the spatial regularization weight has no trustworthy relationship with the intruding foreign object and the related filter model degradation problem.Fully excavate the expressive ability of deep spatial features to ensure the reliability of the spatial regularization weight and the invading foreign object.Thereby effectively suppress the interference of complex backgrounds and realize the accurate tracking of intrusive foreign objects in complex backgrounds.After simulation experiments in different scenarios,both improved methods have achieved high success rate and accuracy.Finally,study the monitoring area division method and pre-warning scheme.The track edge is extracted by the method of perspective transformation and polynomial function fitting.According to the railway boundary standard,the monitoring area is divided into warning area,early warning area and safety area.In order to improve the real-time performance and reduce the false alarm rate of the system,the train identification and dangerous foreign object judgment are carried out.A reasonable and efficient pre-warning scheme is formulated.According to the research results of foreign object detection and tracking in this thesis,real-time early warning is carried out.After several sets of verification experiments,the results show that the method in this thesis can achieve good pre-warning effect.