Overview: This talk will introduce the use of Python and several related modules (GDAL, Shapely, Fiona etc) to build a GIS-based data processing pipeline to retrieve the road network information required to build a rural-urban classification scheme for a region of interest. The speaker will highlight the main concepts used to build the data pipeline and run a simple demo to illustrate these concepts.

Background: A promising source of information for rural-urban classification alternative to population census is remote sensing data, particularly multispectral satellite imagery. A rural-urban classifier based on remote sensing data has the advantage of data consistency and data timeliness. However, from a cost perspective, the proposed methodology has limited practical applicability. The main drawback is that, for geographically diverse countries, the yearly acquisition cost of multispectral satellite images from satellite data provider is prohibitively expensive. Another drawback is that the quality of the satellite images is heavily dependent on the imaging conditions.

In the presence of adverse imaging conditions such as extensive cloud coverage, satellite images of a region may become unclear or unavailable for classification. Even when satellite images are readily available, pre-classification image processing is still necessary and onerous to execute, as the existing classification methodology is sensitive to variations in tone and color of the acquired satellite images.

To address these limitations, a novel solution to the rural-urban classification problem is crafted using data features extracted from road network information. Primarily, rural and urban areas are characterized by distinctive road network profiles. Rural areas generally feature long stretches of road with few intersections and branchings, whereas in urban areas the road networks are often highly inter-connected, resulting in short road segments and a large number of intersections.