SBIR-STTR Award

The primary objective of this effort is to develop tractable techniques to automatically extract terrain and geologic features from interferometric SAR (IFSAR) imagery and Digital Terrain Elevation (DTE) data produced from IFSAR phase data. A secondary objective is to develop performance prediction techniques for the extraction process. Today terrain information extraction from aerial imagery remains mainly a labor intensive, manual practice. Leighly and Associates, Inc., believe the reasons for this lack of progress has been (1) the lack of terrain organization domain knowledge and the use of context in automated image analysis approaches and (2) the lack of data suitable for tractable and robust solutions. This effort addresses both areas of weakness. An approach will be developed which builds on methods for predicting the occurrence of a limited set of geologic landforms in any given physiographic area, and developing strategies which selectively employ terrain partitioning, identification, and attribution algorithms to three-dimensional SAR image data (IFSAR image and associated DTE data). The initial area of interest will concern southwestern U.S. regions where IFSAR data exists. Methods for predicting feature extraction performance for IFSAR data will be developed that are also based on the terrain organization approach to automated feature extraction. Anticipated Benefits/Potential Applications - This effort addresses terrain analysis and in turn terrain analysis supports many dual use applications. Examples of military operations supported by terrain analysis include combat operations planning, automated weapons guidance, IFB, simulation terrain data bases, etc. Examples of civilian value include: site evaluation/engineering, remote surveying, geological surveys, flooding potential/prediction, etc. Methods and techniques developed here may evolve to processing IFSAR satellite data. Documented methods developed will be useful for training terrain analysts.

Terrain information extraction from aerial imagery remains a labor intensive, manual practice. For automated image analysis approaches this is believed to be due to (1) lack of terrain organization knowledge and context and (2) lack of data suitable for a robust solution in three-dimensions. This effort addresses both limitations. Phase 1 developed key ideas for extraction of terrain information from IFSAR data which is composed of digital terrain elevation data (enabling representation of terrain configuration) and co-registered SAR image data. The approach uses concept from the geologic discipline of physiography to organize geographic areas and geomorphic history to predict expected landforms, their properties, and their spatial arrangements. This context allows definition of recognition strategies and use of model-based IFSAR data analysis. Area coverage of IFSAR data was not available in Phase 1 to support implementation. That has changed; IFSAR data is now available that includes Forts Irwin, Hood, Polk, and Benning. In Phase 2, knowledge-based delineation, identification, and attribution of landforms will be demonstrated with IFSAR data and knowledge-based techniques. A secondary objective will determine how this approach generalizes to other physiographic areas. The Phase 2 results will provide the basis for commercialization of IFSAR processing software.

Keywords:
interferometric SAR, terrain organization, image analysis