Shadow detection on multispectral images


Thesis Type: Postgraduate

Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Graduate School of Informatics, Information Systems, Turkey

Approval Date: 2015

Student: HAZAN DAĞLAYAN SEVİM

Supervisor: YASEMİN ÇETİN

Abstract:

Shadows caused by clouds, mountains or high human-made structures pose challenges for identification of objects from satellite or aerial images since they deteriorate and mask true spectral properties of the objects. Therefore, they should be detected to accurately classify objects. Moreover, in change detection problems, shadows deteriorate performance because mere spectral changes act as actual object changes. Shadows are also useful as cues for estimating the height of an object or determining the time of recording. This study is dedicated to developing novel algorithms for detecting shadows on multispectral images using different color spaces. The proposed methods are compared with the methods in the literature.