Evolving conspicuous point detectors for camera trajectory estimation

Abstract

The interaction between a visual system with its environment is studied in terms of a purposive vision system with the aim of establishing a link between perception and action. A system that performs visuomotor tasks requires a selective perception process in order to execute specific motion actions. This combination is understood as a visual behavior. This paper presents a solution to the process of synthesizing visual behaviors through genetic programming, resulting in specialized visual routines that are used to estimate the trajectory of a camera within a vision based simultaneous localization and map building system. Thus, the experiments were carried out with a real-working system consisting of a robotic manipulator in a hand-eye configuration. The main idea is to evolve a conspicuous point detector based on the concept of an artificial dorsal stream. The results on this paper show that it is in fact possible to find key points in an image through a visual attention process in combination with an evolutionary algorithm to design specialized visual behaviors. © Springer-Verlag Berlin Heidelberg 2013.