Unmanned System Design with Emphasis on Fire Recognition

摘要

Governments around the world are beginning to recognize the value of unmanned machines both in industry and in warfare. Heightened interest in this field has spawned robot competitions such as FIRST, TAVIS, MicroMouse and others. This paper describes the University of Central Florida's entry to the AUVSI Aerial Robotics Competition. We present a novel four-engine air vehicle design that uses no control surfaces. Based on the Hiller Flying Platform [1] and the Dragonflyer [2], this ducted vehicle varies its engine speeds to affect pitch, roll, yaw, and overall thrust. The ground vehicle mechanical design is documented in section 3, including finite element models of two key parts. The focus of this paper is automatic fire detection in video images. Our system uses color and motion information gained from video sequences to locate fire. Unlike previous methods, our algorithm is insensitive to camera motion.