Research On Distributed Multiple-Target Tracking Method And System Implementatation In Wireless Sensor Networks

Wireless Sensor Networks is a new kind of ad hoc network platform which is composed by large number of small size, low-cost and wireless communication sensors. It can real-time acquire, analyse and integrate the information of surrounding environment, then feed back to the remote users. Consequently, it has a wide range of use in environment monitoring, equipment management, public security, medical treatment and military affairs. Target tracking is an important research branch in applied research area. It plays a pivotal role in wild-animal tracking, intelligent transportation, and battlefield surveillance and so on. In this article, we improved the single-target tracking algorithm in sensor networks, then based on it, we proposed a multi-target tracking framework.Firstly we compared many localization algorithms and proposed a new "distance-measure" localization algorithm. In the new algorithm we stored the advanced information of targets and divided the perceptive region of nodes. It can improve the localization accuracy and reliability and reduce the computational complexity. The efficiency of localization has been improved.To reduce energy consumption of nodes is a very important issue of target tracking in sensor networks. Based on dynamic prediction cluster algorithm, we proposed "Prince Cluster Header - Abnormal-Competition" algorithm. Firstly we classified the nodes based on the historical prediction results, and then according to the classification results to choose the most suitable node as the cluster header for next locate moment. This mechanism avoided the competition and election among the nodes, reduced the signal transmission and saved energy.Under the circumstances of multiple targets, as the distance between the targets become nearer, the conjunct impact to nodes become larger, thus bringing signal association problems. According to the scene analysis that the distance changes continuously we proposed the multi-target tracking framework: "Long distance single-target precise tracking—Sparse multi-target close-precise tracking and Dense multi-target group tracking."Multi-target close-precise tracking modified the "Prince Cluster Header - Abnormal-Competition" algorithm in single-target tracking. Introduced timestamp mechanism, based on predictions of information we decomposed the nodes'measure, thus associating the signal measures for each target and tracking each target respectively.Multi-target group tracking regards all the targets as a whole unit. Based on the group clusters'formation, localization, sustainment, mergence and decomposition process, we described the whole framework of the group tracking algorithm and the conversion process between the close-precision tracking and the group tracking.Finally, we designed and implemented the prototype system based on Mica2 Platform and the simulation system. The prototype system was used for valida- ting the feasibility of the target location algorithm and the "Prince Cluster Header - Abnormal-Competition" tracking algorithm for single target. The simulation sys -tem validated the multi-target close-precise tracking algorithm and the multi-tar- get group tracking algorithm.