2015 | OriginalPaper | Buchkapitel
Terrain Classification Using Adaboost Algorithm Based on Co-occurrence and Haar-like Features
verfasst von : Ngoc-Hoa Nguyen, Dong-Min Woo
Erschienen in: Advanced Multimedia and Ubiquitous Engineering
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Terrain classification is still a challenging issue in image processing, especially with high resolution satellite images. Specific obstacles include low accuracy in detection of targets, high computation time, and the need for specific algorithms for specific areas. In this paper, we present an approach to classify and detect building footprints, foliage areas, road-grass, and bare ground in a grayscale satellite image (2048x2048). Our contribution is to build a strong classifier using Adaboost based on a combination of co-occurrence and Haar-like features. The Adaboost algorithm selects only critical features and generates an extremely efficient classifier. The combination of two feature extraction decreases the training time and improves the classification accuracy. The accuracy of the proposed method is quite high: over 98.4% for classification and more than 92% for target detection on high resolution images.