The steep increase in urban population and the urban growth associated with it poses challenges to public authorities and utility companies, especially in developing and emerging countries. They need to ensure sufficient living space, good public transport and provide adequate infrastructure such as schools, sanitary systems, electricity, and waste management.
However, infrastructure planning takes years or decades whereas population influx for fast growing urban areas is in the ten to hundred thousands per year. To estimate the urban population and to derive infrastructure needs at earlier stages, constant monitoring of the urban area is necessary. Moreover, urban growth and settlement movements can be mapped to certain events in retrospective, to understand the underlying causalities.
The automated monitoring of urban change in terms of land spread, build-up height and density can be achieved by processing satellite image data. For this, a machine learning model needs training with labelled satellite data to recognize patterns. These could be changes in land cover and land use, the road network, or sealing.
However, the spatial resolution of open access satellite data such as Landsat-8, Sentinel-2 and Sentinel-1 may be too coarse to extract fine details, which could result in weak predictions. This is especially true for the height estimation of urban areas. To mitigate this, several open data satellite sources can be combined for training. They provide data in the visible and radar spectrum, which both can be used to analyse urban growth.
The solution for monitoring urban growth and change with satellite data consists of two tasks: horizontal segmentation and vertical estimation. For image segmentation tasks, certain convolutional neural networks such as the U-Net or the Mask R-CNN algorithms are commonly used.
For the height estimation of satellite imagery, the approach Im2Height can be used, whose architecture is composed of a convolutional sub-network and a deconvolutional sub-network. This method is particularly prone to low resolution data, because the height difference between pixels can be very high, so the right choice of the data source is important.
Also, the data sources should offer bands of different wavelengths, since these can express different features. For example, height information can be derived from the shadows of buildings. These are only observable in the visible spectrum, not in radar data. Radar data in turn offers different methods of height estimation.
Receive news about Machine Learning and news around dida.
Successfully signed up.
Valid email address required.
Something went wrong. Please try again.