Selection and Fusion of Spectral Indices to Improve Water Body Discrimination
| Autor: | Gabriela Calvario Sanchez, Basilio Sierra, Teresa E. Alarcón, Oscar Dalmau, Carmen Hernández |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2018 |
| Předmět: |
Water resources
Index (economics) 010504 meteorology & atmospheric sciences General Computer Science 0211 other engineering and technologies 02 engineering and technology 01 natural sciences Measure (mathematics) remote sensing spectral water index General Materials Science Linear combination Cluster analysis Selection (genetic algorithm) 021101 geological & geomatics engineering 0105 earth and related environmental sciences Mathematics Probabilistic classification business.industry General Engineering Pattern recognition spectral analysis Hierarchical clustering Support vector machine Artificial intelligence lcsh:Electrical engineering. Electronics. Nuclear engineering business optimization lcsh:TK1-9971 |
| Zdroj: | IEEE Access, Vol 6, Pp 72952-72961 (2018) |
| ISSN: | 2169-3536 |
| Popis: | Spectral indices are widely used to emphasize water body information in satellite images. The selection of the appropriate index is one of the tasks that the remote sensing community faces when water bodies are studied. In this paper, we propose an approach for the selecting and fusing of spectral indices, in order to improve water discrimination. First, we compute several spectral indices and analyze their discrimination power, taking into account the accuracy value. Through a hierarchical clustering applied only on indices with accuracy value greater than a certain threshold, we cluster the water indices into different groups. The result of the clustering depends on two factors: the discrimination capacity of the computed indices and the features of the studied water body. Indices in each group are fused by means of a linear combination. Therefore, we obtain an adaptive fusion of different spectral indices. The previous information is used to compute the likelihoods belonging to water and non-water. These values are the inputs for a probabilistic classification framework named Gaussian–Markov measure field. According to our experimental work, the proposed selection and fusion approach improves the discrimination power of the studied indices. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|