A statistical analysis of the relationship between landscape heterogeneity and the quantization of remote sensing data
This report addresses the investigation of the relationship between the landscape heterogeneity and the sequencing of remote sensing imagery for the purpose of better understanding the parameters of the Symbolic Machine Learning developed within the Global Human Settlement Layer project. To address this issue statistical regression analysis was conducted between the sequences derived from the Landsat satellite data and different landscape metrics derived from land cover maps. The results show that only the Relative Patch Richness influences the number of sequences for different levels of image reduction levels. The Shannon Landscape Diversity Index seems to be related to the Number of Sequences in the image until a certain Level of Quantization that may be an indicator of the optimal parameter for the sequencing of the input satellite data. These results represent a good step forward in the attempt to automatize the parameters set of the Symbolic Machine Learning classifier.
DEGEN Antonia;
CORBAN Christina;
PESARESI Martino;
KEMPER Thomas;
2018-12-14
Publications Office of the European Union
JRC111818
978-92-79-92990-8 (online),
1831-9424 (online),
EUR 29340 EN,
OP KJ-NA-29340-EN-N (online),
https://publications.jrc.ec.europa.eu/repository/handle/JRC111818,
10.2760/731774 (online),
Additional supporting files
| File name | Description | File type | |
|---|
