OPTICAL-FLOW ESTIMATION USING SIMPLE GENETIC ALGORITHM FOR ENVIRONMENTAL MAP GENERATING SYSTEM

Makoto Ohki, Kousuke Miyake, Masaaki Ohkita

Abstract

A technique to automatically configure an environment map using image processing is proposed in this paper. An autonomous mobile robot, which we are developing travels in indoor environment while referring to the environment map. This technique does not need a great cost in a measurement and can support a change of more indoor environment. Taking the indoor environment by a moving CCD camera, a continuation image is acquired. For each frame image of the continuation image, optical-flow estimation is computed. Straight line elements are extracted from an individual frame image. The straight line elements are chased between frames using an optical-flow estimation. By repeating such an operation, correspondence of straight line elements between the initial frame and the last frame is decided. By applying 3-D stereography to these two frames, 3-D information of the indoor environment is detected. In this paper, several techniques and new technique of the optical-flow estimation are explained. By an experiment, the effectiveness of the new technique is shown.

Keywords

optical-flow estimation, genetic algorithm, image processing

Download

IMEKO-TC7-2004-052.pdf

DOI

-

IMEKO TC

TC7 - Measurement Science

Back to the proceedings