Research Of Low Power Vehicle Detection Technology Based On Mangneto-resistive Sensors

Traffic information collection is an important part of intelligent transportation system (ITS),the monitoring, decision-making and scheduling of ITS is based on real-time and accurate traffic information. With the widespread use of ITS, traditional vehicle detection technologies are difficult to meet the growing demand for traffic information collection. Vehicle detection using magneto-resistive sensors can acquire traffic information by sensing changes in the ambient magnetic field. Compared with traditional detection technologies, magnetic vehicle detection has potential advantages in cost, accuracy, longevity, ease of installation, and also has broad application prospects. Magneto-resistive vehicle detection technology is in the stage of development with low maturity, and is difficult to meet the demand of application, especially in terms of detection accuracy and working life. So the research on magneto-resistive vehicle detection can help to get a better performance in accuracy and longevity, which is important to improve the practicality of the magneto-resistive vehicle detection and accelerate the development of ITS.First of all, this paper describes the research significance and status of magneto-resistive vehicle detection, and the basic principles of detection are also described. Secondly, a vehicle detection node prototype is designed and implemented, hardware selection and circuit design is introduced, a limited competition protocol is designed to meet communication needs, and the software is implemented under slot-driven structure to improve code efficient. Thirdly, after detailed analysis of vehicle magnetic signal, magneto-resistive vehicle detection algorithms are proposed and validated. Through analysis and selection of detection features, single-lane volume detection algorithm is proposed based on multiple features, and the overall detection accuracy rate reaches97.5%in different traffic situations, with better performance against exiting algorithms for low velocity and dense traffic situations. Multi-lane volume detection algorithm is proposed based on vehicle position recognition and results fusion. With proposed nodes layout scheme, the vehicle position is classified by signal feature extracting and linear discriminant analysis, and the actual number of vehicles on multiple lanes can be obtained by matching and fusing the information of time and position. Multi-lane detection accuracy rate is99.6%, improving the accuracy by13%to18%compared with non-fusion detection methods, while the proposed algorithm has strong scalability and solves the problem of ride-lane and adjacent lane interference. Vehicle direction detection algorithm is proposed based on magnetic signal phase characteristics, the algorithm has detection accuracy rate of98.7%, while also has low computational complexity and good adaptability for vehicle velocity. Fourthly, power consumption is optimized and evaluated, and a detection method of dual complementary sensor nodes is proposed with the use of dynamic sampling intervals and alternate complementary working. The complementary detection releases the redundancy of existing dual-sensor architecture and can reduce power consumption by90%in idle state. The experimental results show that the complementary method is capable of traffic information collection for general accuracy, and can reduce power consumption by50%in dense traffic situations.