SBIR-STTR Award

Converted from advanced automatic target recognition (ATR) processors, this new feature extraction/GIS system (a) uses rasterdata as input, and thus eliminates manual digitization; (b) uses expert systems to extract features automatically, and thus eliminates time-consuming and error-prone manual object tracings and (c) vectorizes the extracted objects and converts them into GIS layers automatically, and thus eliminates the intermediate, man-in-the-loop steps of conventional GIS procedures. We achieve such data processing efficiency by first combining Image processing, Multi sources analysis and GIS into one single system, second by using pseudo English as a programming language to perform pattern recognition, third by using a target identification system to perform numeric character recognition, and fourth by combining a rule-based object recognizer and an image-library-based object matcher into one integrated system. In this environment, each processor is an English key word, and a set of key words becomes an expert system that controls the entire feature extraction and GIS processes. We will demonstrate these claimed system capabilities and GIS benefits from testing with scanned NOAA nautical charts and government-supplied high-resolution image data, if appropriate.Commercial Applications:Currently, recreational boats are equipped to display nautical charts if a proper media exists. Vector-based data are much easier for electronic display, as a minimum scanned nautical chart can bevectorized for use as an electronic chart. Potential users are 14Umillion recreational boat owners. The same technology is applicable to extracting map features from scanned topographic maps. Potential users are owners of ground vehicles. Additional applications are in medical imaging for early detection of breast cancer, and in pattern recognition for character identification.

To generate raster/vector hybrid databases, a three-step process is proposed: 1) map features are extracted; 2) each extracted feature is identified, and 3) it is vectorized as a GIS coverage. Our Phase 1 effort proves that Steps 1 and 2 are feasible. In addition, test results with the SRE's Vectorizer indicate that variation in size, shape, and location after vectorization is absent. Potential features to be vectorized include: 1) Rocks; 2) Obstructions; 3) Ship Wrecks; 4) Soundings; 5) Cables; 6) Pipelines; 7) Bridges; 8) Channels; and 9) Shorelines. Proposed sources include NOS nautical charts, U.S. Corps of Engineers blueprints, and text documents. Phase 2 effort will prove that: a) the SRE segmenter is effective for feature extraction and identification; b) error in raster to vector conversion is either absent or extremely insignificant; c) SRE generated vector data can be transitioned into the ARC/INFO system; d) a division-of-labor approach - between automated and manual method - to vector database generation is optimal; and e) commercialization of the SRE's ImaG system can be a by-product of data analysis. The Marine Board estimated that vectorizing 1,000 NOAA nautical charts with manual methods will cost the Government $20 million; thus, enabling NOAA to achieve the same task with a significant saving in costs and time is one of the Phase 2 benefits. Commercial applications:Potential users are 14 million recreational boat owners. Additional applications are in map feature vectorization, military target recognition, urban, regional and environmental planning, medical imaging, pattern recognition, and industrial robotic vision.