Infrared Multi-object Tracking Technology

Multi-target tracking technology has been widely investigated for decades in the world. In-depth analysis of IR multi-target tracking is done in this paper.Firstly, the principle of IR multi-target tracking and the observation character of IR targets are summarized, as well as several common target motion models. Then, aiming at target tracking and filtering problems, two types of target filtering and estimating methods: One is linear filtering method(Kalman Filter). The current and future time step motion-state of targets can be well estimated in this method. For the other nonlinear method, performances of many nonlinear filterings are compared by simulation and analysis, focusing on Particle Filter and its better algorithm such as Gaussian Particle Filter, it indicates that the performance of GPF is better than other algorithm in the simulation.To solve the data association problem of multi-target tracking, we analyze some typical data association algorithms and advance a MC-JPDAF method based on Particle filter principle. Furthermore, a method based on cost function is applied to track the targets, it can have a good effect.At last, the develop direction of multi-target tracking is forecasted.