Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Pays Bas → Evaluating high resolution GeoEye - 1 satellite imagery for mapping wildlife in open savannahs

International Institute for Geo-Information Science and Earth Observation (ITC) 2012

Evaluating high resolution GeoEye - 1 satellite imagery for mapping wildlife in open savannahs

Yang, Zheng

Titre : Evaluating high resolution GeoEye - 1 satellite imagery for mapping wildlife in open savannahs

Auteur : Yang, Zheng

Etablissement de soutenance : University of Twente International Institute for Geo-Information Science and Earth Observation (ITC)

Grade : Master of Science in Geo-Information Science and Earth Observation 2012

Résumé _ Traditional methods of counting wildlife (e.g., ground or aerial survey) not only disturb the animals but also can be exceedingly time consuming, labour intensive and costly. In contrast, high resolution satellite images, which have the advantage of covering large areas, short cycling time, silence, and demanding less manpower, offer an alternative method to support animal surveys in the in open savannahs. However, the question about whether current high resolution sate llite images are capable of capturing medium to large- sized herd animals such as wildebeest, zebra and buffalo still remain unknown. The present study was conducted in an attempt to map an d estimate the population sizes of wildlife from the high resolution GeoEye-1 satellite images in the Masai Mara Nationa l Reserve. Two different image classification approaches, i.e. pixel-based approach and object-based approach have been se lected to evaluate the capacity of GeoEye-1 imagery to detect animals. Besides, we analysed the factors affecting the performance of the two approaches and compared the performance of pixel-based and object-based approaches in the areas with different animal distribution. We successfully accomplished the objective of mapping wildlife in open savannahs. Specifically, we managed to detect the presence of wildlife in the pilot study area and give an estimation of the population size. Both image classification approaches produced satisfactory results in automated recognition of animals in the pilot study area. The producer’s accuracy ranged from 84.2% to 88.2% and the user’s accuracy ranged from 85.0% to 92.5% by using pixel-based and object-based approaches respectively. Statistical analysis indicates that the density of animals did not affect the accuracy of both pixel-based and object-bas ed approach, and there is no significant difference between the performance of pixel- based and object-based approaches in user’s accuracy and population estimation. However, the pixel-based approach was proved having better producer’s accuracy than object- based. We conclude that high resolution GeoEye-1 satellite imagery is suitable for mapping medium to large-sized wildlife as well as population estimation in open savannahs, but not ready for counting absolute number of animals. In addition to the conclusion, we also make reco mmendations regarding the application fields of the pixel-based and object-based approaches for mapping wildlife in open savannahs applied in this study

Version intégrale (ITC)

Page publiée le 8 février 2018