| As a passive and secret method of investigation, wireless distributed sensornetworks (WDSN) have attracted more and more attention in recent years. It must behighlighted that information fusion is the most important issue.Combined with the subject of “WDSN based vehicle identification and trackingsystem” derived from some institute, this thesis studies the subjects of multi-sensorbased classifier fusion and target tracking. The main work in this paper contains thefollowing aspects:(1) Classifier fusion of vehicle type recognition. This paper studies two differenttypes of physical signal used to recognition fusion--sound and vibration. At sensor node,the paper simulates the KNN based decision fusion method and Bayesian classifierfusion, while at the sink-node, simple voting decision fusion as well as its derivativemethod is analyzed. At the end of chapter two, a perfect classifiers fusion system whichexhibit good performance of classifier fusion is proposed.(2)WDSN used to track target can be divided into two cases, single-target andmulti-target. For a single target, the paper studies the centralized and distributedtracking system, and compares the improvement of multiple measurements to locationaccuracy. To deal with the problem of data association brought up with multi-targettracking, Emphasizes are put on the association of measurement with track based onfuzzy clustering analysis. At the end of this thesis, a comprehensive tracking method isproposed. That is using NN data Association within Centralized fusion system. Throughanalysis and simulation, it is proved simple and effective.(3)The platform of OMAPL138, a Texas Instrument (TI) product, is used toachieve the main function of the vehicle identification and tracking system. In the GPPside, we use ubuntu10.4virtual machine to develop the ARM Linux embedded systemwhile in the DSP side, with the help of CCS5.2IDE I successfully complete anddebugger dsp/bios program. |
PDF Full Text Request