Automated Building Extraction from High-Resolution Satellite Imagery Using Spectral and Structural Information Based on Artificial Neural Networks

Automatic building extraction remains an open research area in digital photogrammetry. While many algorithms have been proposed for building extraction, none of them solve the problem completely. This paper proposes a system for increasing the degree of automation in extraction of building features with different rooftops from high resolution Multispectral satellite images (e.g., IKONOS and Quickbird) in Middle East countries. Following on, the implementation and functionality of software developed on the basis of neural networks approach are also explained. As known, neural networks have capabilities as pattern recognition and object extraction for remotely sensed data. The software has been designed and developed in C# programming environment and it is rather user friendly due to the fact that little knowledge is required for the users about neural networks theory. The proposed system works in two different phases: the first phase is learning, and the second phase is application. In the first phase, the presented neural network in the system is trained with the aid of test data, and in the second phase, the system will be used for detection and extraction of buildings from satellite images. Fused images from IKONOS sensor (1m resolution) from urban area of Kashan in Iran were selected for testing this system. The building extraction results are compared with manually delineated ones. The comparison illustrates the efficiency of the proposed algorithm in which it can extract approximately 80% of buildings in the image properly. * Corresponding author