A principal component analysis based object detection for thermal infra-red images

Autonomous vehicles are increasingly used for transportation of supply and goods. This is done mainly indoors. In outdoor scenarios, a reliable vision system is crucial for the overall system performance. The restriction of the reliability of this vision system is caused by light changes. To overcom...

Full description

Saved in:

Bibliographic Details
Published in:	Proceedings ELMAR-2013 pp. 357 - 360
Main Authors:	Woeber, Wilfried, Szuegyi, Daniel, Kubinger, Wilfried, Mehnen, Lars
Format:	Conference Proceeding
Language:	English
Published:	Croatian Society Electronics in Marine - ELMAR 01-09-2013
Subjects:	Machine Learning Machine learning algorithms Materials Object detection Principal component analysis Thermal Infra-Red Imaging Training Vectors Vehicles
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Autonomous vehicles are increasingly used for transportation of supply and goods. This is done mainly indoors. In outdoor scenarios, a reliable vision system is crucial for the overall system performance. The restriction of the reliability of this vision system is caused by light changes. To overcome the problem of varying lightning conditions, thermal infra-red cameras are often used. This paper discusses an object detection approach for thermal infra-red images. This object detection approach uses principal component analysis (PCA) based machine learning techniques for image classification. Multiple Supervised machine learning algorithms and an unsupervised machine learning algorithm are analysed, evaluated and compared. Based on the experimental data of several tests, a PCA based Gaussian classifier and a Mahalanobis distance based classifier are the best choice for detection and tracking.
ISBN:	9537044149 9789537044145
ISSN:	1334-2630