Tool/software:
Hi,
I’ve been reviewing the parameter tuning guidelines in the "Group Tracker Parameter Tuning Guide for the 3D People Tracking Demo – Rev 1.2", specifically the configuration of the maxAcceleration parameters:
- Ceiling Mount (CM): [1, 0.1, 1]
- Wall Mount (WM): [0.1, 0.1, 0.1]
I understand that the idea is to model the deviation from the constant acceleration assumption, and that higher values imply less trust in the motion model. However, I'm trying to make sense of the different behavior depending on the mounting position.
When tracking people, we typically assume they are standing upright and mostly move along the floor, with little variation in vertical height. So, regardless of whether the radar is ceiling- or wall-mounted, there should be one dimension (height) in which acceleration is minimal and the motion is more predictable.
Given this, I’m puzzled by the difference in maxAcceleration values between CM and WM. In both cases, I would expect one axis to have a significantly lower value than the other two axes while the parameters for CM match this behavior, the parameters for WM are uniformly small.
Is there something I’m missing — perhaps related to measurement noise or motion modeling differences between mounting positions — or could this be a documentation inconsistency? I’d really appreciate any clarification on whether this behavior is intentional, and whether my assumptions are correct regarding expected acceleration in each axis.
Thanks in advance!
