One of the major problems confronting future spaceborne Earth observation missions is the tremendous volume of data that must be handled. The communication links will become overloaded if all data collected are relayed to ground-based stations for processing. Less than 1 percent of the collected data are analyzed. This is due to cloud cover, unwanted data, data taken at the wrong tine, etc. including a "smart" sensor on board that could recognize clouds and generic Earth-features would permit collection of only desirable data. This would eliminate overloading the communications links and considerably reduce the ground-based data analysis costs. This paper discusses developments in remote sensing Earth-feature classification that can help solve the data-deluge problem.