A system to detect houses and residential street networks in multispectral satellite images

摘要: Maps are vital tools for most government agencies and consumers. However, their manual generation updating is tedious, time consuming, expensive. To address these concerns, we developing automated techniques. In this paper, restrict our attention to residential regions. These regions provide a challenge, testing the current limits of image analysis. Such also typically areas rapid growth development and, therefore, interest from applications perspective. previous studies, introduced statistical measures extract kinds satellite images [in: Proceedings International Conference on Pattern Recognition, vol. 1, 2002, p. 127, IEEE Trans. GeoRS (2003), PAMI]. As next step toward automatic map generation, here introduce novel system detect houses street networks in IKONOS multispectral images. have one meter panchromatic resolution with 4 m spectral bands. Our consists four major components: analysis cultural activity, segmentation possible human activity (based surface material), decomposition segmented images, graph theoretical algorithms over decompositions network houses. We tested large diverse data set. results indicate usefulness detecting networks, hence generating maps.