Introduction to multi-target tracking in Wireless Sensor Networks

The principles, energy saving, construction design, and related software and hardware design of wireless sensor networks are introduced. This article focuses on an application of wireless sensor networks-Introduction to multi-target tracking in wireless sensor networks.

Target Tracking of wireless sensor networks has always been a hot topic in research. Previously, many of the researches were single-object tracking, and multiple or all nodes in the sensor network were used to track the same target.

Mechitov K and others use binary-detection collaborative tracking to determine the target location based on the collaboration of multiple sensors by checking whether the target is within or outside the sensor detection distance, this method requires clock synchronization between nodes and requires the node to know its own location information. Zhao F and other information-driven collaborative tracking ideas, use the information detected by sensor nodes and the detection information of other received nodes to determine the possible Motion Track of the target, and wake up appropriate sensor nodes to participate in tracking activities at the next moment, A proper prediction mechanism can effectively reduce inter-node communication, thus saving limited energy resources and communication resources of nodes; zhang w s and others proposed a convey tree tracking algorithm for solving single-target tracking in wireless sensor networks. This algorithm is a distributed algorithm, most of the previous tracking algorithms were centralized, and the transfer tree was a dynamic tree structure consisting of nodes near the Mobile target, some nodes will be dynamically added or deleted as the target moves, ensuring efficient tracking of the target Reduce the communication overhead between nodes.

The current target tracking algorithm is mainly for single-target tracking in different environments. How to efficiently integrate effective information at a low cost of energy, increase the measurement accuracy and prolong the network lifetime, multi-target tracking has become a hot topic in the current research of wireless sensor network target tracking. When researching multi-target tracking in wireless sensor networks, limited energy needs to be taken into account. The distributed tracing algorithm can prolong the network life. sensor measurements may be the synthetic measurements of multiple targets, these challenges traditional multi-target tracking algorithms.

Jaewon Shin adopts a distributed multi-scale framework and uses the transfer matrix to optimize the computing capacity of multi-object recognition. The algorithm updates local node information to provide global target information, this algorithm framework is feasible for multi-target tracking in wireless sensor networks. Lei Chen and others also propose a Distributed Data Association Algorithm for Multi-target tracking in wireless sensor networks; maurice Chu uses Bayesian Estimation to solve the data association problem of multi-target tracking. It also uses a distributed algorithm to achieve multi-target tracking in wireless sensor networks.

Multi-target tracking in Wireless Sensor Networks

Wire Sensor Network tracking is one of the main purposes of sensor networks. It is also a difficult and key issue and has many basic elements for commercial and military applications, such as traffic monitoring, organization security, and access to battlefield conditions. Using node collaborative tracking in wireless sensor networks is an important aspect of wireless sensor network technology.

The earliest wireless sensor network system tracking experiment was the implementation of some tracking methods in the SensIT project of Defense Advanced Research Projects Agency in the United States. Many of the current tracing application solutions are still in the research phase. Due to the limitation of many hardware resources on sensor nodes, sensor nodes are often affected by the external environment. Louis wireless links are subject to interference, and the network topology structure changes dynamically, the sensor network's activity Target Tracking Applications have strong real-time requirements. Therefore, many traditional tracking algorithms are not suitable for sensor networks. Activity target tracking has been studied in the radar field for many years. As a result, many typical activity target tracking systems are single-sensor tracking systems, such as the Nearest Neighbor Method (NN), set theory description method, generalized correlation method, classic allocation method, multi-hypothesis method, probability data association (PDA) method, Joint Data Interconnection (JPDA) method, interactive multi-model (IMM) and other Data Interconnection algorithms.

In the 7O era of 2O century, the multi-sensor information fusion technology emerged to process multiple sensor data in multiple levels, in multiple ways and in multiple layers, resulting in new meaningful information. The centralized multi-sensor comprehensive tracking algorithm is developed directly on the basis of a single sensor system, such as the multi-sensor combined probability Data Interconnection (MSJPDA) and the generalized S one-dimensional allocation algorithm; distributed Multi-sensor track association algorithms mainly include statistical methods (such as weighted method, independent sequential method, classic allocation method, Nearest Neighbor Method (NN) and K-NN method) and the Method Based on Fuzzy Mathematics (Fuzzy double-door restricted track association algorithm and fuzzy comprehensive function-based track association algorithm ). For WSN, because of their limited capabilities of a single node, multiple nodes must be combined for target tracking, and there is no powerful central processor, obviously single-sensor and centralized multi-sensor tracking algorithms are not suitable; the concept of the distributed tracing algorithm is that the sensor has its own information fusion center, which is different from the distributed tracing algorithm of our WSN. It does not consider the ability of the integrated nodes and the computing complexity. Although the above method has high precision, it cannot be implemented in the WSN or the efficiency is not high.