Data Association For Multi -target Tracking Using Bearing-only Sensor Networks

Target localization and tracking has always been a research focus in wireless sensor network fields. Passive bearing-only acoustic sensor network excels at non-emitting, low-power feature, making it useful in many applications in the military fields, environment monitoring, etc. In this scenario, the unknown relationship be-tween targets and untrusted measurements leads to failure in the existing multi-target localization algorithm. And also, measurement error from the bearing-only sensors and the interference of the environment make multi-target localization and tracking problem even more difficult to solve.In this paper, we design a data association method based on sensing probability to solve the problem of unreliable association. Data association includes three parts: measurement to measurement association, measurement to track association, and track to track association. Simulation and experiment has been performed to show the performance benefit of the designed algorithms, which are also applied to the moni-toring system software. The main work and contributions of the paper include the following aspects:Firstly, for the unknown or unreliable matching of measurements and targets, we analyze and design a measurement to measurement association algorithm without pri-or information of targets. We formulate a sensing probability which indicates that if certain intersection of bearing lines is a real target. Also we give a method of setting the gate threshold.Secondly, we design a measurement to track association algorithm based on sensing probability, aiming at solving the track initialization and maintenance prob-lems. The simulation and experiments show that the data association algorithm based on sensing probability has a good performance to capture most of the targets and gen-erates track faster with lower computation loads.Finally, focus on the practical use of the algorithm; we analyze the impact of sensor measurement such as environmental noise, false alarm, missed detection and other factors. Based on real measured data, we design a smoothing algorithms using in the actual scenario, and improve the formulation of the sensing probability. In the end, the paper describes the development and design of the system software, as well as issues related to migration algorithms.