Techniques are disclosed to identify a previously-not-present stationary object within a sequence of image frames. A system determines a first frame of the plurality of image frames where an object is not within a field of view of the camera of the A/V device. The system then analyzes other frames to determine that the object becomes present and is stationary within the field of view of the camera. The system may identify a person approaching and moving away from the A/V device. The system may identify a stationary vehicle within and subsequently not within the field of view of the camera. The system may use identification of the parcel to generate a control signal that changes the operational mode of the A/V device to parcel protect mode.