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

Infoscitex proposes an innovative and robust solution to this topic, a software tool—HMM4D (Human Modeling and Monitoring in 4-Dimension) that can automatically identify and track a human subject, depict the true body shape, classify shape abnormalities, recognize human activities, and detect hostile actions. Using IST’s dynamic 3-D human shape modeling technology, we will extract a parameterized, dynamic 3-D shape model from 3-D sensor data. De-noising and de-clothing techniques will be implemented to get the ground-truth of body shape. The abnormalities of body shape will be identified and classified using advanced data mining technology. Combined with model-based human motion analysis and supported by dynamic human signatures databases, human activities will be decomposed and analyzed in an ontology framework. Dynamic Bayesian networks will be developed to recognize human behavior and predict human intentions. We have selected a talented team for this project consisting of experienced engineers from Infoscitex and two academic collaborators: Dr. James Davis and Dr. Mariofanna Milanova, who are experts and leading researchers in the areas of computer vision and computerized human modeling. We will perform a case study and develop a software demo to prove our concept and demonstrate the feasibility of our approach in Phase I. BENEFIT: Our technology will provide a creative yet robust solution to the Topic and will offer the following benefits: • HMM4D is designed to work with 3-D sensor data, but it can work with 2-D video imagery as well. This means that our technology can meet current needs as well as future demands from Air Force missions. • HMM4D constructs a parameterized, dynamic human model from multi-modality sensor data. The model not only depicts a human shape with high anthropometric fidelity but also provides a wealth of information about the kinematics and dynamics of human activities and actions. • HMM4D exploits human body shape and motion in four dimensions (three dimensional space plus time). This maximizes the potential uses of the information provided by sensor data, increases the discernment and detection capability, and enhances the robustness of the system so that it can function effectively and reliably in the presence of invalid inputs or stressful environment conditions. • The HMM4D is designed to have an open structure that is adaptable to the development of sensor technology. The interfacing with supporting database and the retrieval and fusion of information are designed to be scalable so that large datasets can be used when they become available. Anticipated Defense and Commercial Applications This technology will have very broad applications in both military and commercial sectors. • As a key technology for human-borne threat detection; • As a screening tool for the security surveillance of public facilities; • For war fighters to efficiently identify enemies from a crowd; • For human search and rescue; • For virtual reality creation; • For the entertainment industry to create vivid human figures and animation.