Robust feature detection and tracking in thermal-infrared video

Abstract

In this thesis, popular techniques within the area of machine vision: noise reduction, feature detection, edge detection and feature tracking, have been studied. This project is concerned with the use of thermal-infrared cameras which are much less affected by changes in lighting, shadows and out-of-view motion compared to visible cameras. The main research focus of this thesis is how to deal with the low signal-to-noise ratio of thermal-infrared video in developing a novel real-time methodology for robust feature detection and tracking. The thesis first reviews the background of thermal-infrared imagery. It then covers the necessity of a noise reduction filter in thermal-infrared video. Next, it presents a number of existing approaches in edge and feature detection followed by four proposed techniques. Finally, results reveal that the proposed techniques perform well in thermal-infrared video.