摘要：This article deal with cellular automata (CA) technique applied to multi-temporal remote sensing images. Using simple rules, based on parameters formed by local relationships, CA try to predict possible trends in deforested areas, pastures and forests. Basic segmentation information for the project was derived from multi-temporal satellite images; HRV-SPOT and TM-LANDSAT, from the same study area. The binary maps were formed by pixels that contained information about the existence or inexistence of the cartographic feature named by forest. The forest at the region of Ticoporo, Venezuela was chosen as study case. It is highlight that from 1975 to 1994 this feature showed temporally the tendency to regress the forested area. A tool called “evolution map” was developed to assess forest progression, regression or stability over time. This map was used to specify transition rules for the CA model. The CA model was then applied to predict future dynamics. After subtracting the actual image for the year 1994 (reference image) with the simulated to this year (predicted image) the results showed a similarity of 88.6% between the two images.
