Satellite image processing using AI for identification of large point sources of pollution

Abstract

Comprehensive emission inventory of large point polluting sources is important to government, NGOs and researchers. The ability to track the existing and find new sources of pollution helps the government to set appropriate energy policy (IEA 2007) and allocate limited resources to control pollution (Bollen 2009, Brunekreef 2002, Tong 2018). Though there is a global effort by various organizations like Edgar to track polluting sources, the available emission inventory lacks the pinpointed exact geolocation of all the polluting sources. These existing emission inventory also misses out on newly developed polluting sources due to lack of regular updates. This data gap is hampering governmental efforts to control unnecessary pollution and researcher’s understanding of the cumulative effects of pollution (Junninen 2004).The execution time of our proposed method needs few hours depending on the area of coverage, computational resource and bandwidth availability. The interval lag of collecting new data is determined by the availability of latest satellite imagery. The government or researchers can select their area of interest on map and run a through scan to find all existing polluting sources.After the comprehensive emission inventory is generated, it is against the existing emission database to find the missing and untracked sources of pollution.In contrast to existing database which has either high level generalized pollution data, our emission inventory is an accurate, fine-grained and comprehensive list of all polluting sources.