Conservation-minded evolution of shape

Most natural and artificial systems rely heavily on vision to recognize, manipulate, and navigate within a world of objects. Although shape is a key element in this process, its representation and analysis have proved to be a difficult, multifaceted problem. A framework, based on conservation laws, which gives rise to computational elements for shape parts, protrusions, and bends, is proposed. The computation takes place in the context of a reaction-diffusion space and is highly robust. This scheme is ideally suited to object recognition and has applications in areas ranging from robotics to the psychology and the physiology of form.<
>