Rule-based multiple-target tracking in acoustic wireless sensor networks

Keywords: Wireless sensor networks Doppler effect Kalman filter Track association Multiple targets a b s t r a c t This paper introduces an efficient rule-based algorithm for tracking multiple fast-moving ground vehicles in an acoustic Wireless Sensor Network (WSN). To reduce the level of power consumption, the acoustic sensor nodes in the WSN are assumed to be very simple in nature, having no capability to discern the direction and range of targets or to communicate directly with adjacent peer nodes. Instead the WSN employs a specialized fusion center that collects detection information directly from the sensor nodes. Accurate tracking of multiple high-speed targets can be affected by the Doppler shift and by the merging and splitting of target trajectories. To satisfy all of these requirements for target tracking in acoustic WSNs, a rule-based low-complexity tracking algorithm is presented in this work. The algorithm performs clustering of the detecting sensors and computes the weighted centroid of each cluster to determine the number of targets and their positions. The algorithm also takes into account the target path merging and splitting that necessitates cluster-to-cluster associations as well as detection-to-track associations. The algorithm compensates for the Doppler effect, has a relatively low computational complexity, and has been shown to have high overall track accuracy. Wireless Sensor Networks (WSNs) are composed of numerous wireless micro sensors and a centralized fusion center. For tracking ground vehicles or underwater targets, WSNs have been regarded as useful and are becoming increasingly available due to recent advancements in low-cost micro sensor technology [1]. Requirements for WSN applications often mandate that the sensors have low computational and communication loads because the sensors themselves are battery powered. Thus, it is desirable to utilize sensor elements that are extremely simple, having no peer-to-peer communication capability and having only intensity threshold detection capability. This allows for the sensors themselves to be better mass-produced and more energy efficient. Such sensors send detection information directly to the fusion center when the detection is higher than a predetermined threshold value. It is the job of the fusion center to implement an effective tracking algorithm that will utilize this information in an intelligent manner. Even though the fusion center has significantly greater computational capability than the individual sensor nodes, the scope of the overall problem mandates that the tracking algorithm be computationally efficient. It should also have high track accuracy even when objects are moving at …