An extractive streaming system to reduce monitoring traffic

Abstract

Generally, a monitoring system has tradeoff between network fee and monitoring precision. If the system user wants to monitor a something target from a remote office precisely, the network fee becomes higher because of taking lots of monitoring packets (i.e., video frame). Here, I propose an extractive streaming system to address this problem. The proposed system is installed for both the video sender and receiver. The sender system extracts the pre-determined target area (e.g., license plate) with high quality from a video frame, and sends both roughly compressed entire video frame (e.g., encoded H.264) and the extract area to the receiver in parallel. Subsequently, the receiver system overlays the extracted area on the entire video frame. Through these process, the user can monitor a target area (i.e., focused area) precisely with low network fee. I evaluated the proposed system and investigated the extent to which additional latency and communication traffic was incurred by the proposed system and conducted that the system results in reduced network traffic with decreased latency.