Building detection from high resolution remote sensing imagery

Abstract

Building detection is an important task in very high-resolution remote sensing image analysis. In recent years, availability of very high-resolution images raised new challenges to building detection algorithms. In this report, we use a supervised method to detect buildings from remotely sensed images using spectral-spatial features. The morphological operations (MO), gray level co-occurrence matrix (GLCM) and Variogram techniques are used to extract the spatial features. We concatenated spatial features and spectral features, and then we fed the Support Vector Machines (SVM) classifier with the resulting vector of features. We classified the image data into two classes (Building and Non-Building) using different combinations of features. The simulation results obtained on three different images showed that our approach achieved an acceptable performance in terms of accuracy.