Computer Vision Approaches for Big Geo-Spatial Data: Quality Assessment of Raster Tiled Web Maps for Smart City Solutions

Interactive maps are an important component of Smart City solutions. Most of map services rely on Raster Tiled Web Maps (RTWM). Despite of high popularity of interactive maps, the RTWM quality assessment problem remains unsolved. Two main reasons of this situation are as follows. First, commercial companies do not provide access to the source vector layers and restrict massive access to raster tiles. Usually, they allow users to access raster tiled maps only through official APIs. Thus, massive retrieval of web map raster pieces is prohibited. Second, ground-truth reference datasets, in the most cases, are not available and expensive. In this work, an approach to automatic quality assessment of RMWM is proposed. The approach is based on Canny edge detection algorithm. This algorithm enables to extract edges from raster images. Extracted edges allow users to calculate quantity of information. The method was applied for raster tiles of zoom level 19 in a pilot sites. OpenStreetMap and Google Maps tiles were evaluated. 55693 raster files were assessed two times in May 2017 and February 2018. HTML5 facilities provided by modern web browsers and official APIs were used for development. Implemented solutions enable to estimate completeness of information, positional accuracy and timeliness comparably. This approach will be utilized by a number of Smart City solutions based on interactive web maps in tree pilot sites for decision making processes. The results of analysis will be portrayed in a form of interactive web maps available for public access and will be a part of the Geo-Spatial Data Repository (GSDR, https://wgn.gsdr.gq) for quality assessment of open data. GSDR is implemented as part of the WeGovNow (http://www.wegovnow.eu/) web platform developed for a number of Smart City solutions in Europe