Optical Flow Based Realtime Object Tracking by Active Vision System

Yasushi Mae, Shinya Yamamoto, Yoshiaki Shirai, and Jun Miura
Department of Mechanical Engineering for Computer-Controlled Machinery
Osaka University, 2-1 Yamadaoka, Suita, Osaka 565, JAPAN

Abstract

This paper describes real time object tracking by extracting optical flows from a sequence of images. Optical flows are calculated based on the generalized gradient model. Our method distinguishes a moving object from the flow vectors. Assuming that pixels corresponding to the same object have similar flow vectors, we extract a connected region with similar flow vectors. The direction of a camera is controlled in order to capture the object in the image. Realtime object tracking is realized by an active vision system which consists of an active camera, a realtime image processor, and a personal computer.

Back To Yasushi Mae's Homepage