Procedure on Autonomous Sampling and Reductive Learning in Imagery

How to use

1. Stepwise Procedure

• You need R to run the master script: HabitatSampler.r
• Within the master script a step by step procedure is executed, more info in HabitatSampler.md.
• All necessary information is available under the directory: demo

2. R package

• You need R to install the package HaSa that includes all functions and test data
• devtools::install_github("carstennh/HabitatSampler", subdir="R-package", build_vignettes = TRUE)
• Sometimes there are problems, then do 1. devtools:: install_version("velox", version = "0.2.0", repos = "https://cran.uni-muenster.de/")
• For Windows operating systems the Rtools are needed
• library(HaSa) and list datasets: data(package="HaSa") and functions: lsf.str("package:HaSa") or use library(help="HaSa")
• there are information available about programm execution and function behavior in Rmarkdown: HabitatSampler_Usage

Input

• Image File as Raster Layer Stack (e.g. Satellite Time Series, RGB Drone, Orthophoto)
• Reference File (e.g. spectral-temporal profiles or point shape; one profile or point per category)
• Class Names (the categories that are defined to be delineated in imagery)

Output

• Interactive Maps of habitat type probailities

• Classified Image of chosen categories
• Sample Distribution of sampled categories
• Spatial Statistics of categories distribution
• the categories are refferred to as habitat types

Key Features

• the algorithm provides a set of reference samples for each habitat type
• the algorithm provides an ensemble of calibrated machine learning classifiers for each habitat type
• the algorithm provides a map of habitat type probabilities
• the algorithm is optimzed for broad-scale satellite image time series (pixel size > 10m)
• the alogrthm can be applied on variable image categories in complex scenes
• the algorithm is tranferable to variable input imagery

Citation

Neumann, C. (2020): Habitat sampler—A sampling algorithm for habitat type delineation in remote sensing imagery. - Diversity and Distributions, 26 (12), 1752-1766. https://doi.org/10.1111/ddi.13165.

Credits

HaSa was developed by Carsten Neumann (Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences) within the context of the NaTec - KRH project funded by the German Federal Ministry of Education and Research (BMBF) (grant number: 01 LC 1602A).

The test data represent pre-processed Copernicus Sentinel-2 satellite imagery (ESA 2018). Pre-processing was done using GTS2 and AROSICS.

Community version

HaSa will be further developed under a community version located at GitLab's habitat-sampler group.