Dynamic Node Cooperation Based on Moving Multi-target Tracking in Double Wireless Video Sensor Network

Aiming at the issue of moving multi-target tracking (MMTT) in double wireless video sensor network, this paper proposes a dynamic node cooperative program and a tracking forecasting association program. Particle multi-Bernoulli filtering algorithm is applied to perform the sensor cooperative scheme proposed, including cluster head node selecting scheme and cluster members selecting scheme. At each time step, each cluster head node selects a node that has the best observation perspective on the tracking target among cluster members, and activates this node, making it the cluster head node of next time step. Each new cluster head node collects nodes within its communication range, activates other nodes to obtain more information about the target and makes them their own cluster members. In addition, this paper also applies Gaussian compound filter algorithm to implement tracking – forecasting association program, and combines target identity with the multiple target state acquired from random finite set (RFS). The simulation result evaluates the dynamic node cooperation program proposed in this paper, especially when the dynamic state and measurement process of target are seriously non-linear. The simulation shows that the location estimation of target is more precise by using the target recognition program proposed in this paper.