A Real-time Semantic Video Parsing and Tagging System

Video sensors have undoubtedly become important sources of intelligence in the modern warfare to provide battlespace situational awareness. However, burgeoning numbers of battlespace sensors and the sheer volume of collected video data make it very difficult for battlefield decision makers to grasp a global picture of developing events as well as to detect specific events in real time. Video data need to be made available to users and systems across DoD through the Discovery Metadata Specification (DDMS) set as the common format for a wide-range of DoD data assets. UtopiaCompression (UC) proposes a novel real-time semantic video parsing and tagging (SAVANT) software system to help decision makers effectively process information from live streaming videos so as to accurately and timely respond to the development of battlespace events as well as conduct forensic analysis on archived video to satisfy warfighter priority and specific intelligent requirements. Innovative intelligent video analytics algorithms are employed to extract in real-time video metadata through automated object detection, recognition, context-based scene parsing and event recognition, and to generate DDMS compliant video content description representation. The proposed software will help to improve the successes rate of intelligence gathering missions as well as to reduce manpower requirements.