Research Of Vehicle Detection Based On Wireless Magnetic Sensor Networks

With the acceleration of urbanization in China, the ownership of vehicles in city sustainedgrowth in recent years. Traffic problems have become prominent increasingly. In the parking system,for example, some survey data shows that on-street parking problem is particularly prominent. Inthis context, intelligent transportation systems (ITS) formed gradually.Traffic information collected, such as vehicle detection, occupies a very important position inITS. Only after accurate acquisition of the vehicle occupied information, ITS may have made greatprogress. Since the vehicle is a ferromagnetic substance, the vehicle will be present disturbances inthe magnetic field. So vehicle detection can be achieved by the node with anisotropicmagnetoresistance (AMR) sensor in the ground. Wireless sensor network (WSN) with a number ofsuch nodes can be widely used in ITS. However, vehicle detection technique using wirelessmagnetic sensor networks is still immature currently.For accurately detecting a parking vehicle in a parking lot, we propose a single-node detectionalgorithm named Relative Extremum Algorithm (REA) and a multi-node detection algorithm namedCollaborative Decision-making Algorithm (CDA) based on the large number of experimental dataanalysis and the existing vehicle detection algorithms. REA using process-based approach, itconsidered the case of the parking signal change process and through the state machine to extractreal-time characteristics of the fluctuations. CDA concerned about the relevance of the adjacentparking spaces respectively to do a comprehensive judgment in routers by the integration ofinformation from different nodes.REA have been applied to the sensor nodes and CDA to the router nodes respectively. On theparking lot at Shenzhen Institutes of Advanced Technology (SIAT), more than100sensor devicesare deployed to evaluate the performance of the both algorithms. By running the system for morethan half a year, it was observed that the vehicle detection accuracy of the proposed algorithms isabove98%.