Clustering of Handheld Thermal Camera Images in Volcanic Areas and Temperature Statistics

Abstract

Thermal camera use is becoming ever more widespread in volcanic and environmental research and monitoring activities. Depending on the scope of an investigation and on the type of thermal camera used, different software for thermal infrared (IR) images analysis is employed. The Osservatorio Vesuviano Sezione in Napoli of the Istituto Nazionale di Geofisica e Vulcanologia (INGV-OV) processes the images acquired during thermal monitoring activities acquired in the Neapolitan areas (Vesuvio, Ischia and Campi Flegrei) with different FLIR software that returns for each image, or for each selected area within the image, a series of parameters (maximum temperature, average temperature, standard deviation, etc.). An operator selects the area of interest and later “manually” inserts the relevant parameters in Excel sheets to generate graphs. Such a tedious, time- and resource-consuming procedure gave reason to implement a software able to automatically analyze sets of thermal images taken with a handheld thermal camera without any manual action. This paper describes the method and the software implemented to “automate” and refine the extrapolation process and the analysis of the relevant information. The employed method clusters thermal images by applying K-MEANS and DBSCAN techniques. After clustering a series of images, the software displays the necessary statistics to highlight possible fluctuations in temperature values. The software, “StaTistical Analysis clusteRed ThErmal Data” (STARTED), is already available. Although it has been developed mostly to support monitoring of the volcanoes in Campania, it is quite versatile and can be used for any activity that implies thermal data analysis. In this paper, we describe the workflow and the dataset used to develop the software, as well as the first result obtained from it.

Keywords

• thermal cameras
• thermal images
• clustering
• ground temperature
• volcanic monitoring
• machine learning