Extracting and understanding urban areas of interest using geotagged photos


Our recent paper, "Extracting and understanding urban areas of interest using geotagged photos", has now been accepted and published by the journal Computers, Environment and Urban Systems. In this paper, we propose the concept urban areas of interest (or urban AOI), which refers to the sub-regions in a city that attract the attention of people. Extracting these special areas and understanding their formation can help support the decision making of city planners, transportation analysts, location-based information service providers, as well as others. A number of methods, such as density-based clustering, region construction, and image comparison, have been integrated into a coherent framework, which was then applied to the geotagged Flickr data from 2004 to 2014. The spatiotemporal dynamics of urban AOI as well as some other derived insights have also been discussed.

Journal link of the paper: http://www.sciencedirect.com/science/article/pii/S0198971515300120
Interactive online demo: http://stko-exp.geog.ucsb.edu/urbanAOIs/
Reusable DBSCAN source code: https://github.com/YingjieHu/DBSCAN4LBSN

Selected figures:

Fig. 1. A three-layer framework for extracting and understanding urban AOI from geotagged Flickr data.

Fig. 3. Extracting AOI from photo locations using an example of NYC in 2004–2005; (a) locations of Flickr photos; (b) point clusters detected by DBSCAN; (c) AOI formed using chi-shape algorithm.

Fig. 5. Polygons generated using the chi-shape algorithm under different λP values.

Fig. 14. A historical slideshow of the Eiffel Tower using the preferable photos in three different years; (a) 2004; (b) 2008; (c) 2013.