UniformPoissonModel

Uniform Poisson birth model

Since R2025a

Description

A UniformPoissonModel object contains a birth model that assumes new targets follow a Poisson distribution with a uniform birth density. A Poisson distribution of birth is a common assumption in many multi-object tracking algorithms. In this model, the appearance of new targets is uniformly distributed across the sensor's entire measurement space. You can use this model to define the birth model for a CustomSensor object.

Creation

To create a UniformPoissonModel object, use the sensorBirthModel function with the "uniform-poisson" input argument. For example:

birthModel = sensorBirthModel("uniform-poisson")

Properties

BirthDensity — Spatial density of new targets

1e-6 (default) | positive scalar

Spatial density of new targets, specified as a positive scalar.

BirthDensity is defined as the number of measurements originating from new targets per unit volume of the measurement space. The volume of measurement space is connected to the measurement model of the sensor. For example, if a radar reports azimuth and range measurements, its birth density can be calculated as:

$$\text{BirthDensity}=\frac{\text{BirthRate}}{\text{AzimuthResolution}\cdot \text{RangeResolution}}$$

The denominator represents the volume of one resolution cell of the radar. BirthRate specifies the average number of new targets appearing in one resolution cell per update. Similarly, if you want to calculate the birth density of a camera reporting bounding box as [u;v;w;h], use this formula:

$$\text{BirthDensity}=\frac{\text{NumNewTargetsPerImage}}{\max U\cdot \max V\cdot \max W\cdot \max H}$$

The denominator represents the volume of the measurement space assuming the minimum value for each measurement is 0. The numerator specifies the number of new targets detected on average in one image.

Example: 5e-6

Version History

Introduced in R2025a