Bayesian sensor fusion: A framework for using multimodal sensors to estimate target locations and identities in a battlefield scene.

摘要

We develop a Bayesian framework for battlefield situations that enables unified inference by incorporating multi-modal data obtained by multiple sensors. We concentrate on a military battlefield scene and on problems that arise in tactical decision-making. We propose a marked homogeneous Poisson process as a prior model for target placements in the scene. The sensor suite includes an infrared camera, an acoustic sensor array, a human scout, and a seismic sensor array. The likelihood functions for sensor data include new models for seismic classification and for a scout's spot report. We implement a Metropolis-Hastings algorithm that samples from the posterior distribution of the scene given the sensor data. We define a pseudometric on the scene space and indicate how its use can lead to optimal scene estimates. We present results of our methods applied to simulated battlefield scenes.