Development of a 3D Viewpoint Recognition System Using a Neural Network for Autonomous Flying Drones

Abstract

This paper describes the development of a 3-D object recognition system of a flying drone system for emergency rescue and investigation. In this system, a neural network and new features for 3-D object recognition are key techniques for object recognition from a free viewpoint. Some appearance based characteristics are determined from captured images, and the system uses a neural network to automatically classify automobiles. The proposed system shows that several kinds of automobiles can be recognized from a bird's-eye view. Experimental results showed that the system can effectively recognize four kinds of vehicle objects with high accuracy.