APPARATUS AND METHOD FOR ROBOT LOCALIZATION USING VISUAL FEATURE AND GEOMETRIC CONSTRAINTS

摘要

Disclosed is an apparatus to estimate the location of a robot using visual characteristics and geometric information. A depth image including three-dimensional depth data and an RGB color image including color information are inputted to an image input unit. A feature section extracting unit extracts three-dimensional geometric information from the inputted depth image and extracts two-dimensional color information from the color image. A calculating unit calculates an appearance distance from the RGB color image, calculates a geometric distance from the depth image, and calculates a plane normal vector. A refining unit refines a feature section which exists on a plane. A word matching unit matches the feature section and a word by using a visual word database unit stored in advance. A location estimating unit estimates the current location of a robot based on word matching information. The present invention is able to generate an index word in a topological map when estimating the location of a robot by using three-dimensional depth data. Thereby the present invention is able to accurately grasp the location of a robot by using visual characteristics and geometric information.