Research On Target Classification And Multi-target Localization For Monitoring System Based On WSN

Monitoring System based on wireless sensor network(WSN) is made up of several WSN node clusters, which are composed by a number of sensor-nodes and a sink node. Each sensor-node is configured with a magnetic sensor and a noise detection sensor for detecting the slender varieties of the magnetic field intensity and noise strength around the sensor-node. The sensor-nodes send measured data to sink node by wireless, and sink node would make a multi-sensor data fusion to get more information.The main task of the monitoring system based on WSN is target detecting in monitored area. Target detecting comprises target localization, velocity estimation of target and target classification. Based on the above application background, the thesis makes some researches on target classification and multi-target localization. The thesis is organized as follows.1. Aiming at the target classification problem in monitoring system based on WSN, the study analysis the propagation characteristic of acoustic wave and the property of the geomagnetic intensity variation that caused by ferromagnetic targets. And then, the characteristic of geomagnetic intensity variation and the characteristic of noise strength variation around the sensor-node caused by different kind of target were summarized. According to these characteristic, intuitionistic fuzzy reasoning(IFR) method was adopt for targets classification on this study. Simulation results show that target classification method based on IFR in WSN monitoring system can classify the target accurately with low computation.2. For estimating the positions of multi-targets in WSN, the multivariate equations set based on the multi-target sensing model of acoustic sensor is concluded in the thesis. Then the objective function of multi-target localization, which transformed the multivariate equation set solving problem into an optimal estimation problem, was established. The study adopt Genetic algorithm(GA) for solving the optimal estimation problem. Simulation results show that multi-target localization method based on GA can locate the multi-target with high precision.3. The environment noise has a big influence on the precision of multi-targets localization. Aiming at this problem, the thesis proposed a novel denoising method based on autoregressive moving average(ARMA) model. An ARMA model of noise sequence, which is the measurement of the environment noise before target detected, is established for predicting the aftertime noise sequence mixed in the target acoustic signal after target have been detected. Then the predicted noise power is subtracted from the measured sound signals for revealing the target signal’s power. Simulation results indicate that for any kind of noise which can be whitened, the performance of the proposed method was found to be effective, and can reduce the multi-target localization error.Target classification and multi-target localization for monitoring system based on WSN are subjects of great concern in WSN researching field heretofore, and there are still many problems need to be further studied. For promoting the accuracy of the target classification and simplifying the algorithm of multi-target localization in WSN, the thesis proposed some novel method based on existing target classification algorithm and multi-target localization method for WSN, and achieved the anticipated purposes.