How motion detection works

OdinCam keeps a running model of what the scene looks like when nothing is happening, and compares each new frame against it. Pixels that differ by more than a threshold count as changed.

Rather than simply counting changed pixels, the frame is divided into a grid and a cell only counts as active when enough of it has changed. This is what separates a person walking through the shot from camera noise sprinkled evenly across the whole image — noise changes just as many pixels, but never concentrates them in one place.

Motion has to persist across several frames before it triggers, so a single odd frame is ignored.

The background model updates slowly, which has one consequence worth knowing: something that moves into the scene and then stays perfectly still will eventually be absorbed into the background and stop registering.