Sensors searching for interesting things: Extremum seeking control on entropy maps

This paper presents a novel approach to increasing the information content in sensor measurements, with special applications in images or video. The entropy of a signal gives a measurement of the information content. In the case of images, entropy is low when large parts of an image are uniformly colored or shaded. This can occur due to poor camera settings, poor lighting conditions, or the camera is facing a scene with little interest or activity. Minor camera motions can often alleviate these problems. We employ methods of extremum seeking control to find a local maximum in the entropy map surrounding the camera. Entropy maps often have local maxima that do not correspond to a global maximum. Therefore we combine the global properties of simplex optimization methods with the local search properties and dynamic response of extremum seeking control to create a novel algorithm that is more likely to find a global maximum than conventional extremum seeking control. Simulations and experiments are presented to show the strength of this approach.