Hi,
The visionDetectionGenerator object holds information about the vision sensor's configuration and provides a method that models the detection of targets according to positions and velocities of the targets and ego vehicle.
As mentioned in the documentation, the arguement 'ProcessNoiseIntensity' refers to the noise intensity used for filtering position and velocity measurements, specified as a positive real scalar. Noise intensity defines the standard deviation of the process noise of the internal constant-velocity Kalman filter used in a vision sensor. The filter models the process noise using a piecewise-constant white noise acceleration model. The noise intensity is defined in units of m/s^2 and should be of the order of the maximum acceleration magnitude expected for a target to be detected by the sensor. Since it is a scalar value, the maximum value required depends on target's acceleration.
As per my understanding, if a detection is generated, the measurements are generated from the bounding box which encloses the target’s visible image. Noise is then added to the position estimate according to the uncertainty of the target’s position introduced by the width and height of the camera’s pixels. It is likely a random Gaussian distributed measurement noise.
For more information on exact mathematical relationships governing the algorithm, I would recommend you to contact MathWorks Technical Support using the following link:
Regards,
Zuber
