Microwave video vehicle detection and classification technology fusion research
|School||Wuhan University of Technology|
|Course||Communication and Information System|
|Keywords||Vehicle Detection Vehicle classification Microwave radar sensor Data Fusion Bayesian networks|
With the rapid economic development, transport and infrastructure needs of the contradiction between the increasingly prominent, so intelligent transportation systems came into being, which is a kind of advanced information technology, data communications transmission technology, electronic control technology, computer processing technology focus on the application of new transportation system, it makes transportation infrastructure to maximize performance. In our intelligent transportation system has been a great deal of attention and development. Vehicle detection and classification of intelligent transportation system is a very important aspect, it can provide a wealth of traffic flow information, such as traffic density, vehicle speed, vehicle characteristics, vehicle type and road events such as traffic flow parameters for the traffic signal control, road monitoring, road planning information. However, the existing vehicle detection and classification techniques confined to a single sensor operating mode, regardless of the classification results from the detection or classification from the detection efficiency point of view, are subject to a lot of constraints, such as the traditional loop detectors, laser detection, infrared detection, video detection and radar detector do have their inherent disadvantages, and some by the installation conditions, and some by the external environment and the weather, some of the information provided is too simple, are unable to meet the development needs of intelligent transportation systems. This paper presents the outstanding problems of microwave sensor fusion is accomplished video vehicle detection and classification. In the construction of the hardware platform to complete microwave vehicle detection and video vehicle detection and classification of vehicle integration work through the microwave detection module to get the vehicle height profile, video detection module to get the vehicle flat profile, and then use the extracted mixture Gaussian distribution vehicle characteristics modeling, the last in the framework of the Bayesian network to classify vehicles. Mixed Gaussian distribution can better fit vehicle characteristics, sensor fusion based on Bayesian network vehicle classification method that can be of different types of sensor characteristics make scientific judgments, the possibility to quantify the results of the evaluation, and the entire system is highly configurable and robustness. Through experiments showed that microwave video vehicle detection and classification sensor fusion system has a very stable vehicle detection performance of traditional classification methods have greatly improved vehicle and breakthrough, the accuracy of classification from a single microwave sensor fusion for 79% to 87 sensors %, in particular small and medium sized vehicles and large vehicles between the major classification error rate decreased from 9% to 2%. Finally, a microwave video vehicle detection and classification sensor fusion system applications realized, after the actual test illustrates the system is stable and reliable system design to achieve the intended purpose. Microwave video sensor fusion vehicle detection and classification systems in intelligent transportation systems and vehicle monitoring multi-sensor fusion has a broad application prospects.