The ongoing development of open data policies for satellite imagery leads to new opportunities in the urban remote sensing field, such as global mapping or near-real-time monitoring. However, supervised classification that has been proved to be one of the most efficient methods to extract built-up information, happens to be inapplicable in such contexts, since the training data collection step is difficult to automate. This study explores the use of another open data project, OpenStreetMap, to collect built-up training data. In the context of Ouagadougou (Burkina Faso), we investigate the most relevant features to use and the optimal pre-processing procedures to consider. Experimental results show that we can expect similar accuracies with OSM-based training data than with the hand-digitalized ones, provided that the necessary pre-processing operations are carried out.
Forget, Y., Linard, C., & Gilbert, M. (2017). Automated supervised classification of Ouagadougou built-up areas in Landsat scenes using OpenStreetMap. In 2017 Joint Urban Remote Sensing Event, JURSE 2017. Institute of Electrical and Electronics Engineers Inc. https://doi.org/10.1109/JURSE.2017.7924571