Research On The Technologies Of Distributed Intelligent Monitoring For Invasion Limit Of Foreign Body Along The Railway Based On Machine Vision

With the great-leap-forward development of railway in China, the passengertransportation with high speed and the heavy haul freight have become increasingly enhanced,which is facing new challenges to the security and safety systems of railway operation.The intelligent video monitoring is a new application direction and also a hot topic in the fieldof computer vision, which aims to observe and analyze the content within the scene to realizethe automatic and intelligent monitoring, and has shown its great development potential inthe practical engineering applications of railway. In terms of the issues on frequent accidentsof invasion limit of foreign body along the railway, which has severely affected the safety oftrain running, the thesis has focused on the detection and identification of invasion limit offoreign body along the railway using machine vision to realize the monitoring, early warningand safety management in the whole process of safe railway operation.Based on the analysis of the latest technologies at home and abroad related to theinvasion limit of foreign body along the railway, machine vision, artificial intelligence and soon, the basic principle and methods of binocular machine vision, image analysis, patternrecognition and state warning etc. have been deeply researched in the thesis. Thecomputation model and realization algorithm of the camera calibration, image preprocessing,characteristic extraction, determination of invasion limit, identification of foreign body andstate warning of invasion limit, based on the three-dimensional binocular machine vision ofthe invasion limit of foreign body along the railway, have been provided. The system modelof the distributed intelligent monitoring has been established and the prototype system of theintelligent monitoring for the invasion limit of foreign body along the railway has beendeveloped. After the analysis of the experiments and the practical application, the model andmethods have effectively resolved the problems of the detection, identification and earlywarning for the invasion limit of foreign body along the railway, which can meet the rapidand steady performance of real-time monitoring.Firstly, the urgent situations of railway safety and security systems have been deeplyanalyzed, in which the role and significance of the distributed intelligent monitoring systemshas been explained. After the frequent consulting and careful analysis of theoretical andapplied literature, related to the technologies of railway safety monitoring and machine vision,the design ideas of the distributed intelligent monitoring systems for the invasion limit offoreign body along the railway has been proposed.Secondly, based on the analysis of the imaging and detection principle of binocularmachine vision, the camera calibration system of three-dimensional binocular vision has beenestablished in the light of the appropriate algorithm of camera calibration and the inside and outside parameters of the vision monitoring system for the vasion limit of foreign body alongthe railway have been achieved. After the analysis of the related technologies involvingimage denoising, image enhancement, image sharpening and three-dimensional reconstructionin machine vision for the invasion limit of foreign body along the railway, a rapid algorithmof image denoising with recurrent integer GCV threshold has been proposed in the thesis.Thirdly, compared with the several common detection algorithms of image edge, theimproved arithmetic operators with noise immunity, expansion and erosion have beenprovided and a new detection algorithm of image edge based on the multi-structure elementsof grayscale morphology has also been proposed. While the pixel characteristics in rail imagehave been detected and the coordinates in real space have been calibrated, the detectionwindow of two-dimensional railway invasion limit has been determined. Therefore, themethods of background estimation and space shadow elimination of the detected image havebeen put forward after the analysis of the image characteristics of the invasion limit of foreignbody along the railway, and the foreign body within two-dimensional invasion limit can bedetected and judged by background difference with dynamic threshold. Furthermore,according to the detected results of the foreign body within two-dimensional invasion limit,the stereo matching and three-dimensional reconstruction have been used to determine thethree-dimensional invasion limit to realize the detection of the invasion limit of foreign bodyalong the railway within three-dimensional space.Fourthly, fter the scenario analysis of the invasion limit of foreign body along therailway, the level evaluation of and novel method of early warning based on the dynamicBayesian network and fuzzy classification have been provided, which has described thevarious effect on the level of invasion limit by the probability method combined with expertknowledge. The observation data has been processed by fuzzy classification to achieve theforecast results by reasoning algorithm.Then the models of the evaluation and early warninghave been established and the relevant cases have shown the methods are effective.Finally, the model of the distributed intelligent monitoring for the invasion limit offoreign body along the railway based on MAS has been built on the base of the deep researchof the architecture of the railway intelligent distributed monitoring system, which hascomprehensively depicted the composition of the agent in the monitoring system and therelationship between the functional modules in terms of the logical structure and function.The formal description of the monitoring activities has been done and the communicationsbetween agents have also been researched. The prototype system of the distributedintelligent monitoring of the invasion limit of foreign body along the railway has beendeveloped, involving the design ideas and logical structure of the system and the specificfunction realization. The conclusion of the thesis has summed up the related and finished research andbrought forward some needed research in the near future, involving the further researchproblems in the field of theories and application of railway intelligent distributed monitoringsystem.