The paper proposes a multi-criteria and data driven fusion approach whose semantics can be explained in terms of attitude towards decisions. It is exemplified to assess environmental status from remote sensing images in order to identify hot spot of critical situations and anomalies induced by wildfires, floods, desertification, erosion etc. by fusing multiple factors defined by experts knowledge. The fusion function is an Ordered Weighted Averaging (OWA) operator, whose behaviour is here characterized by degrees of pessimism and democracy. The paper proposes to explain the semantics of the fusion function learnt from few ground truth data available, i.e., the OWA operator, by computing its degrees of pessimism/optimism and democracy/monarchy, which are defined as semantic interpretations of both orness and dispersions respectively. Pessimism indicates if the fused map is more prone to commission (overestimation) or omission (underestimation) errors, while democracy indicates how many factors contribute to the generation of the map. The approach is exemplified to map the flooded areas from remote sensing by considering different models based on distinct spectral indexes and domain experts.
Explainable Multi-Criteria Data-Driven Environmental Status Assessment from Remote Sensing
Stroppiana D;Boschetti M;Brivio PA;Bordogna G
2022
Abstract
The paper proposes a multi-criteria and data driven fusion approach whose semantics can be explained in terms of attitude towards decisions. It is exemplified to assess environmental status from remote sensing images in order to identify hot spot of critical situations and anomalies induced by wildfires, floods, desertification, erosion etc. by fusing multiple factors defined by experts knowledge. The fusion function is an Ordered Weighted Averaging (OWA) operator, whose behaviour is here characterized by degrees of pessimism and democracy. The paper proposes to explain the semantics of the fusion function learnt from few ground truth data available, i.e., the OWA operator, by computing its degrees of pessimism/optimism and democracy/monarchy, which are defined as semantic interpretations of both orness and dispersions respectively. Pessimism indicates if the fused map is more prone to commission (overestimation) or omission (underestimation) errors, while democracy indicates how many factors contribute to the generation of the map. The approach is exemplified to map the flooded areas from remote sensing by considering different models based on distinct spectral indexes and domain experts.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.