Parallel Processor-Based Real-Time Atmospheric Compensation of Hyperspectral Images

Airborne down-looking hyperspectral imaging (HSI) is promising for intelligence, surveillance, and reconnaissance (ISR). HSI systems use numerous spectral bands to distinguish materials, and thus provide target identification and discrimination against background clutter. Down-looking HSI sensors are severely compromised by atmospheric effects such as clouds and aerosols. Spectral Sciences, Inc (SSI) proposes to dramatically increase the speed of atmospheric compensation for ISR. Accordingly, the goal of the proposed effort is a factor 10 to 20 speed-up of VIS/SWIR atmospheric compensation, by modifying the SSI-developed code Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) for implementation on GPU-based hardware. A two-pronged strategy is proposed for the necessary FLAASH speed-up. First is to use a much faster version of the radiation transport code MODTRAN, dubbed kURT (correlated-k-based Ultrafast Radiative Transfer) currently under final stages of development at SSI. Second is to implement both kURT and FLAASH on a GPU (Graphical Processing Unit). This will make FLAASH fast enough to exploit radiosonde data so that atmospheric correction will also be more accurate. The enhanced atmospheric compensation system will be called FLAASH-Parallel kURT (FLAASH-PK). BENEFIT: The anticipated results from successful completion of this project will be parallelization hardware and software to address the difficult problem real-time hyperspectral image enhancement for the remote tracking and assessment of targets. The new FLAASH-PK system will be applied to DoD HSI-based ISR missions from air and space platforms. Subtracting atmospheric effects from hyperspectral data will also enhance HSI, LADAR, and RADAR sensor fusion. Because of the flexibility of the algorithms, lightweight and powerful GPU hardware, and the high information content afforded by hyperspectral imaging, there are commercialization opportunities in remote sensing (such as environmental monitoring, mineral exploration and precision farming), process control (such as in the chemical, petroleum, and pharmaceutical industries), public safety, and homeland security.