Other

Applied Research, LLC

Bulent Ayhan

Chiman Kwan

A method and apparatus for real-time target recognition within a multispectral image includes generating radiance signatures from reflectance signatures, sensor information and environment information and detecting targets in the multispectral image with a sparsity-driven target recognition algorithm utilizing set of parameters tuned with a deep neural network.

Automatic target recognition system with online machine learning capability

This invention fuses spectral information from multiple imagers of an unmanned ground vehicle (UGV). Since the imagers contain different spectral information for each pixel location, the invention provides highly accurate targeting and guidance information to the UGV. The invention applies a robust 2-step image registration process for alignment of images captured by each of the multiple imagers to automatically identify targets of interest, so that the UGV can move toward the targets. This two-step image registration process can achieve sub-pixel accuracy. After registration, a new multispectral image is formed with pixels containing spectral information from all imagers. The invention further incorporates an accurate anomaly detection algorithm to help detect new targets in the scene, and incorporates advanced composition estimation algorithms to determine the composition of targets. Finally, the invention allows users to interact with the target detection results through a user friendly graphical user interface.

Growing community of inventors

Bethesda, MD, United States of America

Bulent Ayhan