Naslov
Object‐Based image analysis for detecting indicators of mine presence to support suspected hazardous area re‐delineation

Autori
Vanhuysse, Sabine ; Hölbling, Daniel ; Friedl, Barbara ; Hanson, Emilie ; Krtalić, Andrija ; Hagenlocher, Michael ; Racetin, Ivan ; Wolff, Eléonore

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
GEOBIA 2014, Advancements, trends and challenges / Gitas, Ioannis ; Mallinis, Giorgios ; Patias, Petros ; Stathakis, Dimitris ; Zalidis, Georgios - Solun : Aristotle University of Thessaloniki, 2014, 525-530

Skup
GEOBIA 2014, 5th Geographic Object-Based Image Analysis Conference

Mjesto i datum
Solun, Grčka, 21.05.2014. - 24.05.2014

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Feature extraction ; Humanitarian Demining ; Image processing ; Remote Sensing

Sažetak
In the framework of Mine Action, the extent of Suspected Hazardous Areas (SHAs) is often overestimated. This study investigates the potential of Object‐Based Image Analysis (OBIA) for extracting Indicators of Mine Presence (IMP) to support a more precise delineation of SHAs, with the aim of ensuring an optimal use of demining resources. The study area is situated in the Svilaja mountain range in Croatia. Using 3K colour aerial photographs, we implemented two approaches for the extraction of dry stone walls located in an area that displays traces of military activities. The first approach uses object‐based class modelling, which describes an iterative process of segmentation and classification. The second approach implements supervised learning techniques based on advanced statistical classification methods, i.e. Support Vector Machines, Random Forests and Recursive Partitioning. The results are compared, the strengths and limitations of both approaches are discussed, and perspectives for further improvements are considered.

Izvorni jezik
Engleski

Znanstvena područja
Geologija, Geodezija

POVEZANOST RADA