Accurate Color Classification and Segmentation for Mobile Robots

Abstract

Visual perception systems are fundamental for robotic systems, as they represent an affordable interface to obtain information on different objects in the environment for a robot, and because they emulate the most commonly used sense in humans for world perception. Many techniques can be used to identify an object within an image. Some of these techniques are color object identification, shape detection and pattern matching. Each one of these techniques has different advantages; however, color based techniques are usually preferred in real-time systems, as they require less computing power than other approaches. Color object identification is composed by two phases: image segmentation, and object identification. The goal of the first phase is to identify all regions of the image that belong to the same object of interest. These regions are analyzed by the second phase in order to extract features of interest from these objects like geometry and relative distances and to infer the presence of a specific object.