Based Particle Filter For Maneuvering Target Tracking Technology

The traditional Kalman filter is well used in the aspect of the maneuvering targettracking, since it is the best result under the condition that the systemic dynamic model islinear and the noise Gaussian. The state posterior probability distribution resolved by theKalman filter is not approaching resolution. But because of the condition of practicalapplication such as projects, we have to aid non-linear and non-Gaussian elements to thesystemic dynamic model. I.e. the system dynamic model becomes non-linear and the noisebecomes non-Gaussian. So the traditional Kalman filter can't resolve the problem well.The particle filter completed by Gordon can resolve this problem well though approachingstate posterior probability distribution. In the non-linear and non-Gaussian conditions, itsresult of maneuvering target tracking is much better than Kalman filter.At first, we introduce the traditional Kalman filter and the Extended Kalman filterbriefly. Secondly, we expatiate the theory of particle filter though the probability statisticaltheory in order to resolve the problem of maneuvering target tracking under the non-linearand non-Gaussian conditions. Under several applied conditions, we make somemodifications based on the particle filter theory, and thus we get four related particle filters.We make computer simulations to the maneuvering target tracking though MATLAB andcompare with their performance by Root Mean Square Error(RMSE). At last, we applyparticle filter in the Wireless Sensor Net(WSN) which attracts people these years.Corporating data fusion, entropy of expected posterior distribution, kernel representationand so on, we realize the maneuvering target tracking. And we simulate the applicationthough MATLAB.