Research on the Multi-Target Tracking Scheduling Algorithm in the Wireless Sensor Network

Abstract

The main problem is how to improve the tracking accuracy and reduce the energy consumption in the multi- target tracking research of the wireless sensor network. The detection extent should be selected according to the relation- ship between the prediction target state and the sensor nodes. In addition, the selected detection extent can be waked up and then form the clustering set to track the targets. The successful tracking for the multi-targets can be achieved by ad- justing the detection extent and adopting time to separate the conflict time in the conflict nodes. The simulation results show that the method proposed in the paper can actually improve the probability of successful tracking. The wireless sensor network is composed of a large number of sensor nodes which are deployed in the specific regions. Each sensor node is equipped with the embedded processor, the sensor component, the storage and the wire- less communication. The wireless sensor network can be applied in various fields, such as environmental monitoring, military application, healthy application, household applica- tion, business management, target tracking and other fields. Its framework is mainly divided into Sink node and Sensor node. The Sink node mainly collects the data obtained from the environmental monitoring of the sensor nodes. The Sen- sor node mainly collects the environmental information ob- tained from the Sensing range. The Target Tracking in many kinds of environmental monitoring application in the wireless sensor network is al- ways an important application, such as detecting the enemy movements in the military application, tracking the animals' moving routes, nursing the family and others. Tracking the mobile purpose is more difficult than tracking the fixed re- gions. The target's position information must be updated at any time for obtaining the real-time position information of the target. Therefore, the sensor nodes need to do the local- ization work to the target at any time and constantly monitor the target's moving method for reducing the chance of losing the target and offering users target information. 2. RELATED RESEARCHES The paper mainly adopts the prediction method to predict the target position. The sensor nodes can be waked up in terms of the prediction target position and the clustering set can be formed to track the target which is the Prediction Based Clustering and the paper does researches aiming to the Cluster-based target tracking. Many sensor nodes will be formed into the clustering set during the process of the track- ing task and then the Cluster Head should be selected from the clustering set, while others belong to the Cluster Mem- ber. The cluster member should measure the target informa- tion and transmit the information to the cluster head. Later, the cluster head should collect the data, compute and predict the target information and then decide which sensor nodes should do the tracking task in the next time so that the cluster head is very important and energy-consumed. The above description is the Dynamic Clustering framework. The reference (5-7) introduce an adaptive Sampling in- terval. The reference (6) proposes a distributed adaptive multi-sensor node scheduling which implements the target tracking with the cooperation of the sensor nodes. The uncer- tainty of the prediction target position based on the Extended Kalman filter judges whether the uncertain results of the position meets the requirements of the threshold value and then chooses the optimal detection extent. If the sensor nodes can detect the target's probability and set up the probability threshold value during the process of selecting the sensor nodes, n sensor nodes should be selected to form the cluster- ing set according to the sensor probability and the threshold probability so that the tracking target task can be done. The cluster head can be selected to collect member's information from the clustering set and then the next step's tracking task should be computed and do a signal target tracking. If the sensor node just can track a target at the same time, the DMTT proposed in the reference (7) can adopt the tar- get's historical route to compute the detection extent. When the target moves in a straight line, the detection extent be- comes more larger, while the detection extent becomes more smaller whose target is moving in a irregular line. The EKF should be adopted to predict the target position during the process of choosing the nodes, and the distance between the sensor nodes and the prediction target position should be computed on the basis of Predicted covariance matrix in the Extended Kalman Filter. Then the shortest n sensor nodes in