All-Weather Feature-Based Combat Identification Navy SBIR FY2005.1 Sol No.: Navy SBIR FY2005.1 Topic No.: N05-030 Topic Title: All-Weather Feature-Based Combat Identification Proposal No.: N051-030-1116 Firm: Referentia Systems Incorporated 550 Paiea Street Suite #236 Honolulu, Hawaii 96819 Contact: David Parker Phone: (808) 423-1900 Web Site: www.referentia.com Abstract: Referentia Systems Incorporated proposes a new and unique technology for real-time all-weather combat identification. The SBIR solicitation suggested a feature-based solution, but our technology delivers superior performance because it is not feature-based. Instead of using templates, and instead of extracting features from sensor data, Referentia's technology fits 3D models of objects directly to the sensor data, bypassing feature-extraction and all of its pitfalls, such as feature explosion in visually complicated scenes. After fitting known objects to the data, our technology can then perform the reverse operation: convert the remaining unfitted data to 3D models, for immediate classification by a human operator (e.g. friend/foe/neutral). It can then save new objects in a local or global knowledge base for immediate use by any linked systems. Our Phase I technical objectives are to research, implement, test, and demonstrate the core technologies needed to successfully complete the entire project. Benefits: This technology has many important applications in other military and commercial fields besides real-time all-weather combat identification. Some representative examples are: o It could be used to automatically measure sea waves for Naval weather operations. The system could fit 3D models of various types of waves (swells, seas, and breakers) to real-time stereo optical images of the ocean, and from the fitted 3D models determine crucial wave parameters such as height, celerity (speed), and direction. Currently, the wave height is measured by human estimation and the other parameters aren't gathered. o It could be used to create a high-resolution 3D model of an area by combining the information from many satellite photos into one 3D model. The advantage over traditional methods is that no human input is needed to find control points that match from photo to photo, and that a higher resolution 3D image can be obtained by combining information from several 2D images instead of relying on a single stereo pair. o It could be used to create 3D templates for self-targeting munitions. A 3D model of the target would allow the munition to approach along any path. Currently, some munitions are guided by 2D templates and can only approach along the path the photo was taken. o It could be used to minimize collateral damage by making munitions smart enough to recognize mission abort objects in the scene. For example, the munition might abort if it recognized a school bus or a group of children near the target. Commercial applications could include: o Economical creation of immersive 3D environments for remotely exploring real estate property o Dynamic real-time reconstruction of sporting events to allow fans to "get in the middle of the play" as it happens o Automated on-line searching of image and video databases (much richer capability than recently publicized efforts) o Effective 3D face matching for security Return