SBIR-STTR Award

There is a current need for a dynamic target knowledge base for automatic target recognition to adapt to the variations in targets encountered in the real world. The proposed research will study methods to determine differences between 2E image and range data of a target and a library of 3D CAD model data, and update the CAD model to accurately represent the real world target. We will detect features, within the 2D data and attempt to detect corresponding features in the 3D CAD model.Creating a match quality metric to measure the difference between thefeatures will allow us to determine which CAD model best matches the 2D data. The same metric will allow us to find areas of differences between the selected CAD model and the 2D data. The selected CAL model will be adapted to more closely match the real world 2D data, and the match quality metric will again be used to assure the quality of the adaptation.

Benefits:
The proposed research will allow rapid updating of target knowledge bases that would greatly enhance military automatic target recognition systems. Other commercial applications include assembly line quality assurance systems, robotic handling systems, and medical diagnosis through MRI imaging.

Detailed CAD models of ground targets are currently used to predict images used in missile guidance systems. Images predicted from these models may not agree well with images of specific targets because of variations in target geometry and/or variability in target surface conditions. Model validation by comparison with measured reconnaisance images will reveal the existence of model errors and their location. CAD model updating can then be applied to improve the match of predicted images to those of specific targets. The proposed research will develop a software/hardware system for use in rapidly updating target CAD models from information derived from realtime reconnaisance photographs of potential targets. This system provides functions such as infrared image acquisition, CAD model rendering image preprocessing, edge detection, line feature extraction, automatic image registration, image match quality evaluation, image match error analysis, and CAD model updating.

Benefits:
The proposed research will provide the Government with a direct and efficient method of validating and updating complex CAD models used in target knowledge bases. It has commercial applications as a tool for general CAD model validation using photography.