Research On Key Technologies For The Analysis And Forecast Of Urban Traffic Operation State

Urban traffic is a carrier for urban operation, and it is closely related to the work and life of city population. With the development of social economy, the level of urbanization is increasing, and the contradiction between the limited urban traffic resources and people’s growing traffic demands is becoming more and more prominent. Urban transportation network carrying capacity overload, traffic congestion, traffic pollution, and traffic safety have become the basic problem of urban traffic. Practice has proved that it is an practicability and effective way to improve the traffic capacity and service quality with the advanced technology of urban intelligent transportation.In urban intelligent transportation, through the urban traffic basic data acquisition, urban traffic operation state extraction, urban traffic operation state analysis and forecast of three stages, people can realize the effective monitoring and management of urban traffic.The basic data of urban traffic is the basis and foundation of scientific decision-making for the management department to guide the travel and road planning. Compared with the traditional digital signal, image information can directly reflect the real traffic, and that the effective treatment for image information of urban traffic has become the key to obtain basic data.Because the traffic image contains a lot of noise and invalid information, to obtain accurate and reliable data, need to filter the data and processing. In this thesis, the key technology of traffic image processing is studied, and the related algorithms are proposed, which provides the basis for the accurate extraction of traffic operation state.In the extraction of traffic operation state, the basic data of traffic is often isolated digital or image information. It is difficult to be understood and read directly, and it is needed to match the acquired digital and image signal with map. In this thesis, the key technology of digital map matching in extraction of traffic operation state is studied, and an effective algorithm is proposed, which provides the conditions for the analysis and prediction of urban traffic conditions.In the analysis and prediction of urban traffic operation, the state data is needed to be analyzed and studied to get the operating law and capture the key information. Through training and testing the traffic forecasting model, forecasting the traffic operation status, and providing the basis for the decision-making department, and providing the effective information service for the people to travel. Based on the analysis of the historical data of Chengdu City, this thesis puts forward a new forecasting model based on the research of the traditional Markov model, which provides an effective method for the prediction of urban traffic operation state.In this thesis, the key technologies of image processing, state extraction, prediction analysis and other key technologies are studied, which provide some effective solutions to improve the level of urban traffic management.The main works are as follows:(1) Due to the limitation of objective conditions in urban traffic based data acquisition, we often can only get a low resolution traffic image, a single image hybrid Super-Resolution(SIHSR) is proposed. By combining L0 image smooth processing method, bicubic interpolation, and sparse regression to process single low resolution image, we obtain a high resolution image. This technology is used in the low resolution images include traffic network, vehicle, license plate. With the contrast experiments of traditional super resolution technique, the discovery is tha the SIHSR is more effective and better. For details please see Chapter 2.(2) In real urban road network, there are numerous complicated road structures such as parallel road and complex interchange bridge. Traditional map-matching algorithms could not match trajectories on these structures accurately. In this thesis, we propose a map-matching algorithm based on the topological structure of the road networks and transform the problem that matching GPS trajectories in road map into the problem that finding the shortest path in weighted road network. We test the algorithm with the real data of GPS trajectories of 13933 taxis in Chengdu. The results show that the presented algorithm can acquire a high success ratio and accuracy ratio in complicated urban road networks. For details please see Chapter 3.(3) In real urban traffic, the existing traditional model is difficult to reflect the traffic state relationship between multiple roads because the complex interaction between each path. Basing on the research of the traditional traffic prediction model and Markov model, AAT-HO3M(Adjustment Term Higher-order Multivariable Markov Model) is proposed with convergence and estimation of the parameters. This model is applied in traffic congestion, the results of the predictions illustrate that the AAT-HO3 M is better than the traditional higher-order multivariable Markov model and improved multivariable Markov model. For details please see Chapter 4.(4) Using the the research results of this thesis, we analy the Chengdu traffic operation status. We use the map matching technology fom this thesis to get the floating car real running track, the average speed of each road, and draw traffic network congestion status distribution in Chengdu. The operation law of traffic network in Chengdu is analyzed by using the average speed of each road. Using the forecast model, the traffic congestion status is predicted. The static accessibility and dynamic reachability of Chengdu transportation network are analyzed. For details please see Chapter 5.The new ideas and methods proposed in this thesis have some guidance significance and application value to the current urban traffic management.