Automated Analysis and Classification of Anomalous 3-D Human Shapes and Hostile Actions

Detecting terrorist activities in urban environments and remote terrains is an important aspect of persistent surveillance. Normally, intelligence analysts examine video to detect suspicious activities. This approach to video analysis can be time- and manpower-intensive. To reduce the cognitive burden on the decision making operators, Vecna Technologies proposes a software solution that automatically classifies anomalous 3-D human shape and hostile actions from 3-D laser scans, 3-D camera images, volumetric data, and motion capture data. Traditional 3-D model-based shape classification frameworks suffer from the limitations of high-dimensional search when trying to deal with the wide variety of human poses and appearance, cluttered scenes, and variable lighting conditions. Our approach resolves these issues by efficiently combining global (nearest neighbor) and local (model fitting) techniques while maintaining multiple shape hypotheses. Given a temporal sequence of 3-D data, our software can identify anomalous behavior with a high level of accuracy using a novel discriminative probabilistic model. In addition, the software generates an anomaly map highlighting potential harmful objects like weapons or concealed explosives that a person might be carrying, allowing quick verification of hostile actions. Surveillance networks incorporating the proposed technology will enable persistent, unmanned monitoring of humans and automate reports of suspicious activities. BENEFIT: The proposed technology is applicable to a wide range of DoD and intelligence community areas including force protection, counter-terrorism, target recognition, human activity monitoring, and surveillance and tracking. We see significant potential for application of this tool to support a range of tactical and strategic systems, including shipboard Navy CIC centers, Army field C3I centers, or USAF theater airborne command posts. A number of programs sponsored by the Department of Defense (e.g. FCS, HumanID, CTS, VIVID) employ video-based monitoring systems capable of directly detecting and tracking humans, identifying threats, and notifying security personnel for further investigation and would benefit from the proposed system. Commercialization opportunities exist in a variety of sectors, including visual surveillance and mobile robotics. Initial analysis of these market segments reveal both unaddressed needs as well as vast potential for rapid adoption and growth.