Visual Localization in Aerial Imagery using Convolution Neural Network

Abstract

The purpose of this research is to propose a novel approach using deep neural networks to estimate the position and orientation of aerial vehicle using vision sensor only. We will create and contribute an annotated dataset for visual odometry in aerial images. It will provide an alternative solution to locations where GPS is not working accurately. A novel DNN will be proposed and trained on data set that will cover diversity of land areas including cities, villages, forests, deserts, lakes, farms.