The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLIII-B4-2021
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B4-2021, 167–174, 2021
https://doi.org/10.5194/isprs-archives-XLIII-B4-2021-167-2021
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B4-2021, 167–174, 2021
https://doi.org/10.5194/isprs-archives-XLIII-B4-2021-167-2021

  30 Jun 2021

30 Jun 2021

AI IN SUPPORT TO WATER QUALITY MONITORING

C. A. Biraghi1, M. Lotfian1,2, D. Carrion1, and M. A. Brovelli1 C. A. Biraghi et al.
  • 1Department of Civil and Environmental Engineering, Politecnico di Milano – Lecco Campus, Via Gaetano previati 1/c, 23900 Lecco, Italy
  • 2University of Applied Sciences and Arts Western Switzerland, Institute INSIT, 1400, Yverdon-les-Bains, Switzerland

Keywords: Artificial Intelligence, Convolutional Neural Networks, Citizen Science, Water monitoring

Abstract. This study explores the possibility of using Artificial Intelligence (AI) as a means to support water monitoring. More precisely, it addresses the issue of the quality and reliability of Citizen Science data. The paper addresses the tools and data of the SIMILE (Informative System for the Integrated Monitoring of Insubric Lakes and their Ecosystems) project in order to develop an open pre-filtering system for Volunteer Geographic Information (VGI) of lake water monitoring at the global scale. The goal is to automatically determine the presence of harmful phenomena (algae and foams) in the images uploaded by citizen scientists to reduce the time required for a manual check of the contributions. The task is challenging because of the heterogeneity of the data that consist in geotagged pictures taken without specific instructions. For this purpose, different tools and deep learning techniques have been tested (Clarifai platform, a Convolutional Neural Network (CNN), and an object detection algorithm called faster Region-based CNN (R-CNN). The original dataset composed by the observations of SIMILE – Lake Monitoring application, has been integrated with the results of both keyword and image searches on web engines (Google, Bing, etc) and crawling Flickr data. The performances of the different algorithms are presented for their capability of detecting the presence and correctly labelling the phenomenon together with some possible strategies to improving them in the future.