Cognitive Cellular Automata for Image Segmentation: A Social Learning Metaphor

摘要

Cognitive agents have the ability to perceive their environment and act on it according to models of reality built through memory, intelligence and language. Interacting cognitive agents interchange information about their models in order to build a collective knowledge of their reality (social learning). In this paper we use this distributed cognitive system paradigm to solve a segmentation problem in image processing from the complex systems engineering approach. We build a cognitive cellular automata where each pixel in the image is a cognitive agent. Social learning is achieved by stigmergic and direct communication among agents. Our results outperform typical segmentation methodologies for granular material. Our social cognitive learning metaphor exemplifies a complex systems engineering approach for more general applications.