Research On Target Tracking Algorithm Based On Wireless Sensor Networks

Wireless sensors are so cheap,lightweight and flexible that wireless sensor networks are widely used in people's lives and production.The target tracking is one of the most important applications of wireless sensor networks.Therefore,target tracking accurately is a very important research direction of wireless sensor networks.This thesis is to study the target tracking of wireless sensor networks.The main work of this thesis is as follows:1.A target tracking algorithm based on the prediction of trajectory for fusion is proposed.The traditional Kalman filtering algorithm relies on the target motion we estimated,which result in some unstable target tracking results.The tracking result obtained at the previous time contains the information of the target trajectory,which can be based to predict the current state information of the target.We derive the equations of the trajectory based on the sequence of the precursor optimal positioning results.The target positioning results based on the prediction of trajectory are merged with extended kalman filter and maximum likelihood and kalman filter respectively by introducing the fusion parameters,and two target tracking methods based on fusion are obtained.Experiments show that the tracking accuracy can be improved by changing the fusion parameters,whether in linear motion or circular motion.The smaller the motion noise of the target is,the more obvious the improvement of the positioning accuracy is.Comparing to the original algorithm,the fusion methods in the experiment do not significantly increase the running time.2.A target tracking algorithm based on the prediction of trajectory for initial positioning is proposed.The traditional Kalman filtering algorithm relies on the target motion we estimated mainly in its time update phase.The fusion result obtained by combining the prediction value after the time update phase of the Kalman filtering with the result of the prediction of trajectory is taken as the initial positioning result for the target,which is input to the estimation phase of the Kalman filter to continue the filtering calculation.The initial positioning method based on the prediction of trajectory is applied to extended kalman filter and maximum likelihood and kalman filter respectively.Experiments show that the new method obtained by adding the prediction of trajectory for initial positioning is more accurate than the corresponding original method when tracking whether it is tracking the target of linear motion or circular motion,and the running time is not significantly increased.3.A dynamic interactive multiple model supported with trajectory prediction and Singer model is proposed for tracking the maneuvering target.The target motion in reality is generally maneuverable,and it is difficult for us to describe its motion with a single mode.The trajectory equations have the advantage of the dynamics which can be used to play the role of a motion model.We can describe the dynamic linear characteristics and maneuverability of maneuvering target motion by combining this trajectory model with the Singer model.The two models interact with each other in a probabilistic form,and the probability can be adjusted in real time through calculation.The new target tracking algorithm is carried out under two maneuvering motions of curve maneuver and linear maneuver respectively.Experiments show that this improved interactive multiple model method can not only achieve more accurate target tracking,but also automatically enter the packet loss mechanism to predict the target when the sensor data packet is lost,which makes it impossible to miss the target.