Roads Feature Extraction from High Resolution Satellite Images Using Decision Tree Classifier

Document Type : Original Article

Authors

Egyptian Armed Forces.

Abstract

Extracting roads network data, from satellite images, is important for urban planning, infrastructure development, navigation applications, military purposes and updating topographic GIS databases. Very high spatial resolution remote sensing images are useful source for extracting roads information. In this paper a novel algorithm for roads feature extraction of sub-urban and rural areas from high resolution images is developed. The developed algorithm is based on Decision Tree classifier (DTC). The developed DTC models are trained and applied on corrected Multispectral images acquired by Worldview – II (WV-2) satellite. The experimental work is performed using MATLAB with Graphic User Interface (GUI) for designing and managing the training data .The result of DTC model were evaluated using three of quality measures. The results are compared with a digitized reference roads layer. The obtained results show the possibility of using the developed algorithm for automatic roads feature extraction from high resolution images. The results provide about 82% accuracy of roads extraction.

Keywords

Files

Share

How to cite

Statistics