Automated Scene Understanding
Agency / Branch:
DOD / OSD
Automatic visual content extraction and scene understanding is an enabling technology for video surveillance, situational awareness, and force protection applications. High-level scene comprehension requires a deep understanding of objects, scene elements and their inter-relations. Current systems lack a general visual knowledge framework and efficient computational algorithms for detecting large number of object categories. We propose to develop video inference algorithms and a modular architecture for scene understanding based on the computational framework of And-Or graph. Various image inference modules can be easily integrated to the framework through an API for scene understanding. The architecture should also support the input of user-provided context to improve inference. Leveraging earlier work on semantic annotation, we will develop algorithms to infer complex relationships between scene entities. Plain text reports of the scene will be automatically generated to describe these relationships, contextual information, as well as events of interest. To achieve high compression rate for bandwidth-constraint applications, the text description are used to synthesize image and video to provide a representative rendering of the scene and events. In addition, we propose a method for indexing hierarchical data as well as a scalable framework for searching large imagery dataset.
Small Business Information at Submission:
11600 Sunrise Valley Drive Suite # 290 Reston, VA 20191
Number of Employees: