Regional variance in target number: Analysis and application for multi-Bernoulli point processes

In the context of multi-target tracking application, the concept of variance in the number of targets estimated in specified regions of the surveillance scene has been recently introduced for multi-object filters. This article has two main objectives. First, the regional variance is derived for a multi-object representation commonly used in the tracking literature, known as the multi-Bernoulli point process, in which the multi-target state is described with a set of hypothesised tracks with associated existence probabilities. This model is exploited in multi-target applications where it can be assumed that targets evolve independently of each other and generate sensor observations that are uncorrelated with other targets. An illustration of the concept of regional statistics (mean and variance) in target number, and how to interpret them in the broader context of multi-object filtering, it then provided. Possible applications include performance assessment and sensor control for multi-target tracking.