Advanced Algorithms for Exploitation of Space-Based Optical Spectral Imagery

We will develop a software product ATMOD for the automatic generation of optimized atmospheric models. ATMOD will combine a MODTRAN-based physical model, image-based estimates, and standard forecast products. A physics-based atmospheric perturbation stage will support the modeling of fine-grained temporal and spatial variation in the atmosphere. The techniques will be derived to support real-time performance. ATMOD will consider three spectral ranges: the visible through short-wave infrared (VNIR/SWIR), the mid-wave infrared (MWIR), and the long-wave infrared (LWIR). We will characterize the ATMOD software using a large amount of real hyperspectral imagery. The results of each stage in the atmospheric modeling process will be assessed by comparison with hyperspectral sensor data. We will also demonstrate the utility of ATMOD for improving the performance of subpixel target and plume detection algorithms for space-based hyperspectral imagery. We will consider the utility of the new models and algorithms for various military and commercial applications.