Earth resources data processing is presently in a research phase. Scientific spacecraft collect data and telemeter it to ground stations for processing. It is becoming apparent that significant amounts of useful information can-be extracted from the multispectral scanner images by ERTS-1.

It appears that in the 1980-1990 decade operational spacecraft will be launched to continuously monitor crop production, severe weather systems, earth and water pollutants, etc., in order to provide operational information to the users of this information on a near-real time basis. It appears that it may be more cost effective to process the data on-board the satellite and transmit the results directly to the users of the information as opposed to transmitting down the raw data, processing it on the ground, and then disseminating the results via another satellite system.

In this paper we consider the feasibility of processing earth resources data on-board the satellite. We first determine some likely user application areas and from these the required data processing algorithms. These algorithms are then analyzed with respect to the computational load they impose on the on-board data processor. The on-board processor must also be matched to the format of the data as it is produced by the imaging system. Given the data input format constraints and the computational requirements required by the algorithms we investigate some computer architectures for the on-board processor. Since the multispectral scanner produces parallel data streams with high data rates and because of the many computations required by the algorithms some sort of multiprocessor organization seems appropriate.

Date of this Version