Data-Driven Tracking Based on Kalman Filter
A good model of the target will extract useful information about the target’s state from observations effectively. There are many models used to maneuvering target tracking, such as constant-velocity (CV) models, Singer acceleration model (zero-mean first-order Markov model) and “current” model (Mean-Adaptive Acceleration Model), etc. While due to the complexity of maneuvering target, to seek the target model which can get better performance is still a subject worthy of study. For the AR process, autocorrelation function is estimated by the random sampling points in this paper. We have the statistics relation between the autocorrelation function and variance based on a first-order stationary Markov process. Then the system parameters are obtained and a model is developed based on statistics relation, which needn’t set unknown parameter. Simulation shows the model developed can adaptively get the model parameter and obtain good performance.