Surface observers on monitoring meteorological stations regularly classify and register sky patterns based on human cognitive training. Substitute those observers by automatic systems is highly desirable to reduce subjective analysis. Most automatic systems use thresholding methods and ignores other physical phenomena existent in atmosphere obtaining only binary results (clouds and sky). The current work proposes a method that will establish the correspondence between observations percept from those patterns and theory using Bayesian approach embedded in Intelligent Agents in the following steps. Model the cognitive patterns on color space obtained from surface cameras and analysis. Validation and refinement of system results could be done using sun photometers and LIDAR systems.