Automated, Universal Software for Cloud and Cloud Shadow Detection in RS Data

Limitations with existing cloud cover detection (CCD) techniques for large dataset processing and new challenges presented by the increase in the quantity and quality of data in the commercial realm, offer an opportunity for R&D into new and improved methods for the detection of clouds and cloud shadows in acquired imagery. We propose to develop innovative software for automated pixel-based cloud and cloud shadow detection. The novel, iterative, self-guided approach will rely on spectral and spatial information from a limited number of bands (R-G-B or R-G-B-NIR) and will be applicable to large datasets of a wide range of commercial and government space- and air-borne imagery. Our techniques will be refined by algorithmic testing on a variety of image types and by consulting with industry experts on their assessment of the applicability of the algorithms to their data and to the needs of their end users. Further, in order to flesh out the market and requirements for successful commercialization, we propose to research novel methodologies and inputs to: (1) recovery and substitution of cloud and cloud shadow contaminated pixels, (2) on board real-time CCD processing, and (3) cloud cover monitoring, forecasting, and avoidance strategies.