Leveraging Social Network Data to Design Smart Cities
5. July 2019 – The United Nations has estimated that 63% of the world’s population will live in mega-cities by the year 2050. Migration into cities is posing new urban planning challenges, raising such questions as: What is the number of people actually living in a given urban district? What are the infrastructure requirements? How can available space be best utilized? Munich Aerospace fellowship holder Matthias Häberle is collaborating with scientists of the Technical University of Munich and German Aerospace Center to find answers to these questions, utilizing data derived from social networks.
Earth observation is a way to obtain information of great value in efforts to improve urban living. High-resolution satellite cameras deliver views of the Earth’s surface structures in detailed images in which residential buildings are clearly distinguishable from industrial plants. But specific constellations such as whether a shop is operated inside a city centre building are still unrecognisable. In such cases social media could be helpful.
The world of Earth observation may be turned upside down through the unconventional usage of valuable spatial data derived from social media networks. A new technology has been developed to tie other data sources in with data from Earth observation satellites. This allows combining data sets which have a range of different spatial features for improved understanding of a phenomenon. Twitter, which has 330 million users worldwide generating billions of tweet messages with text and images, has particular potential for providing information on urban features to enhance the limited perspective obtained from satellite data.
Munich Aerospace fellowship holder Matthias Häberle is taking a novel approach, utilising select tweets in conventional earth observation. “A lot of tweets are sent inside buildings or out on the street. This is a rich source of data which can be combined with high-resolution satellite images to greatly improve specification of building types,” explains the young scientist. His project could thus add more detailed information to the images we have of urban spaces. City planners for example would be able to better assess local infrastructure requirements based on building types.
The Munich Aerospace research group on Fusion of Remote Sensing and Social Media Data analyses a huge number of tweets using natural language processing technology to mine physical environment data from text messages. Word embedding is used for instance to map text within a multi-dimensional space of features without little loss of syntactic or semantic information. This renders language machine-readable for further processing within deep learning systems.
The phenomenon of urbanization is posing increasing challenges for cities. Utilising data science in this innovative fashion will enable better planning for the limited amount of space available in urban areas, thus realising the vision of smart cities. Deep learning holds great opportunities, as drawing on new data sources affords better definition of building types that yields greater understanding of cities and urban processes.