Cloud Identification Using Genetic Algorithms and Massively Parallel Computation

NASA Grant No. NAG 5-2216

Using the MasPar MP-2/MP-1 at Goddard Space Flight Center, AVHRR images from NOAA and additional images from the Naval Research Lab at Monterey, a massively parallel genetic algorithm was developed that succeeded in identifying clouds at an equal or greater level of speciation and accuracy than most methods in the open literature. The final report (postscript 452K) is now available.

The systems users' manual (tarred and compressed 243K) is also available, but one should first read the README file.

To date, two significant publications have resulted from this research, and both may be obtained by anonymous FTP:

• Dev Prabhu, B.P. Buckles, and F.E. Petry, "Behavior of Interconnected Subpopulations: Genetic Algorithms in a SIMD Environment," submitted to Evolutionary Computation (postscript 326K).
• Dev Prabhu, B.P. Buckles, and F.E. Petry, "Genetic Algorithms for Scene Interpretation from Prototypical Semantic Description," submitted to IEEE Trans. on Systems, Man and Cybernetics (postscript 646K).