Study On The T-CPS Oriented System Of Detection And Recognition For Vehicle

Transportation system is typically nonlinear and strongly coupled in a large-scale spacetime.Consequently,its running law is extremely complex.Limited by vehicle detection technology,traditional way for traffic information perception is not enough to meet the needs of modern traffic control.With the integration of computation,communication and control(3C)technology,the Cyber Physical Systems(CPS)provides a feasible development direction for the next-generation Intelligent Transportation System(ITS).It is necessary to perceive the road vehicles accurately,efficiently and deeply oriented for the T-CPS.And the vehicle perception is the prerequisite and foundation for grasping the full traffic status.As a result,the vehicle perception and recognition become important direction in the field of ITS study.The study on the theories and technologies of information extraction,information fusion and information transmission is of great theoretical significance and application value.On the basis of the study of CPS,a system of vehicle perception and recognition is built.The main contents researched in this dissertation are as follows:1)For the sophisticatedly nonlinear and strongly coupled characteristics of modern transportation system,the CPS of vehicle perception and recognition is established,after the theory and technology related to CPS is studied.Then the system architecture and function model are designed.The system is divided into perception layer,transmission layer,operation layer and decision layer.As a result,the system has wide scale layout,wireless network and the characteristics of multi-information perception for vehicles.2)By analyzing the principle of the effect of vehicle motion on geomagnetic field,the design of sensor node is proposed.In order to effectively identify magnetic disturbance signals,an autonomous recognition algorithm is presented based on the Gauss model of the background data.Experimental results show that the algorithm can effectively separate the vehicle disturbance information from background data.It has solved the problems that geomagnetic detection is easily influenced by the external environment disturbance.3)According to the wide layout of system sensors,the clustering algorithm is designed in accordance with daily laws of the traffic flow,after the rules of typical intersections traffic flow are studied.Based on the energy dissipation model,the adaptive protocol about single-hop to multi-hop is provided.The simulation results show that the protocol is able to improve the energy-efficiency,balance the energy consumption and prolong the life cycle of the system.4)Based on the vehicle detection model with multiple sensing nodes,the normalization preprocess is accomplished in dimensions of time and amplitude.By means of Maximum Likelihood Estimation,the algorithm for information fusion from multi-nodes are proposed according to the disturbance from adjacent lanes.Based on the energy norm,three axis vector information is fused together.For the vehicle information recognition from the data sequences,the double-threshold vehicle dividing algorithm is proposed.The experimental results show that those algorithms are able to solve the problems of error identification and leak identification.5)The properties of geomagnetic signals both on time domain and frequency domain are studied.After that,on the basis of Gauss Process Classification(GPC)model,the method for estimating the length-scale parameter is given with combining the spectrum of geomagnetic signals for computing the radial basis function.The algorithm of GPC for vehicle binaryclassification is given based on spectrum analysis and the auxiliary one for multi-classification based on directed acyclic graph(DAG)are proposed.The experimental results show that the proposed classification method has a good performance for vehicle type identification.In conclusion,four key problems are studied based on the theories and technologies of CPS.The problems include the perception and recognition for vehicle geomagnetic information,the clustering algorithm for balancing the energy consumption of the wireless network,the information fusion from multi-nodes and the vehicle type identification by the spectrum of geomagnetic signals.By the examples analysis on real data,the simulating experiments and the field tests of the verification system,the proposed methods and algorithms are proved to be effective and practical.