tracking::DetectorConfig Struct Reference

Motion detector tunables. Defaults are tuned for an indoor scene 1-3 m away. More...

#include <tracking_types.h>

Public Attributes
int	diff_threshold = 20
	|gray - background| above this is foreground
int	bg_shift = 4
	background cells learn 1/2^bg_shift of the difference per frame (1/16)
int	fg_shift = 7
	foreground cells learn 1/2^fg_shift (1/128): no trails, slow absorption
int	fg_age_limit = 34
float	scene_change_fraction = 0.35f
	raw foreground above this fraction of the grid (+ mean shift)
int	scene_change_mean_shift = 25
int	warmup_frames = 17
int	min_blob_cells = 2
	smaller blobs are dropped (rejects isolated noise cells)
int	max_blobs = kMaxBlobs
	keep at most this many, largest first (clamped to kMaxBlobs)

Detailed Description

Motion detector tunables. Defaults are tuned for an indoor scene 1-3 m away.

Member Data Documentation

fg_age_limit

int tracking::DetectorConfig::fg_age_limit = 34

after this many consecutive foreground frames (~2 s) a cell learns at bg_shift again, so ghosts of a wrong background clear quickly

scene_change_mean_shift

int tracking::DetectorConfig::scene_change_mean_shift = 25

...and only if |frame mean - background mean| also exceeds this (gray levels); a big close object barely moves the mean (auto-exposure)

warmup_frames

int tracking::DetectorConfig::warmup_frames = 17

frames (~1 s) after a (re)seed with no blobs; the background copies the frame each of them, so exposure settling and lamp afterglow are absorbed

---

The documentation for this struct was generated from the following file:

• lib/tracking/include/tracking_types.h