A particle filter approach to estimating target location using Brownian bridges

Abstract

We study the problem of modelling the trajectory of a moving object of interest, or target, given limited locational and temporal information. Because of uncertainty in information, the location of the target can be represented using a spatial distribution, or heatmap. This paper proposes a comprehensive method for constructing and updating probability heatmaps for the location of a moving object based on uncertain information. This method uses Brownian bridges to model and construct temporal probability heatmaps of target movement, and employs a particle filter to update the heatmap as new intelligence arrives. This approach allows for more complexity than simple deterministic motion models, and is computationally easier to implement than detailed models for local target movement.