Object detection is video analytics that identifies what is moving, not merely that something moved. A trained model classifies targets such as people, vehicles or animals and can filter alerts by class. Accuracy depends on resolution, angle and available light.
Plain motion detection reacts to pixels changing, which includes trees, shadows and headlights. Object detection asks a harder question and answers it far more usefully.
On the camera or the recorder
Some cameras run detection at the edge and send results; others stream to a recorder or server that analyzes the video. Edge processing uses less network capacity, while centralized processing is easier to update across many cameras.
What it needs to work
The target must be large enough in the frame and reasonably unobstructed. A camera aimed too wide sees a person as a few pixels, and no model recovers detail the lens never captured.