Skip to content

Towards Automating Reliable SO_2 Camera Retrievals

Alyssa Heggison 1, Thomas C Wilkes1, Tom D Pering1, Jefersson A dos Santos2

  • Affiliations: 1School of Geography and Planning, University of Sheffield, UK; 2School of Computer Science, University of Sheffield, UK

  • Presentation type: Poster

  • Presentation time: Monday 16:30 - 18:30, Room Poster Hall

  • Poster Board Number: 127

  • Programme No: 3.1.19

  • Theme 3 > Session 1

Abstract

Since the development of the technique in the early 2000s, the use of UV cameras to obtain sulphur dioxide emission rates at volcanoes has become widespread. As robust, low-energy-use permanent systems have been developed, the volume of data available for processing has greatly increased. However, imaging conditions are rarely ideal - rain, cloud, wind and other influences can impact significant proportions of data retrieved. Dealing with imperfect data requires changes to standard processing procedures, and the uncertainty in measurements is significantly increased. Therefore evaluating data quality and adapting processing methods manually is a necessary but time consuming task. Multiple previous works have focussed on automation of processing steps. However, a current limitation is the ability to adapt both processing methods and uncertainty estimations based on the quality of a given retrieval, without user interaction. In this poster we present progress toward detailed, automated quality assessment of UV image data. Computer vision techniques are used to replicate the judgement of an expert human user evaluating the data. This will form a first step toward reliable automation of data processing as a whole, and improve understanding of the uncertainty in flux values derived from this data.