A collection of my solutions to the exercises of the master course Advanced Remote Sensing and Geomatics (winter term 2023/24). The exercises cover the full chain from raw geodata to thematic maps: terrain analysis, multispectral and hyperspectral image analysis, machine-learning classification and regression, and airborne / terrestrial LiDAR processing – mostly in R, complemented by QGIS and CloudCompare.
| Area | Methods / tools |
|---|---|
| Raster & vector GIS | terra: cropping, reprojection, zonal statistics, map algebra, raster ↔ vector extraction |
| Terrain analysis | SRTM DEM, slope / aspect, reclassification, hillshade, elevation–temperature regression |
| Multispectral RS | Sentinel-2 L1C/L2A preprocessing (sen2r), spectral signatures, feature spaces, NDVI / EVI / PVI |
| Hyperspectral RS | HyMap & EnMAP data, spectral libraries, SVM and random forest (hyper-parameter tuning, CV) |
| Machine learning | Tree-species classification, above-ground biomass regression, variable importance |
| Airborne LiDAR | lidR: ground classification (CSF), DTM/DSM/CHM, individual tree detection & segmentation, area-based metrics |
| Terrestrial LiDAR | CloudCompare + 3DFin: stem detection, DBH and tree height estimation |
| Desktop GIS | QGIS: digitizing training data on 20 cm true orthophotos, map projects |
| # | Exercise | Topic | Data | Tools |
|---|---|---|---|---|
| 01 | R fundamentals | R basics, data frames, climate normals & t-test | DWD station data | R |
| 02 | DEM terrain analysis | Slope, aspect & elevation classes (Guelb er Richat, Black Forest) | SRTM 1″ | terra |
| 03 | Temperature vs. elevation | Linear regression, raster interpolation, zonal statistics per municipality | SRTM, DWD, ALKIS | terra, QGIS |
| 04 | Sentinel-2 preprocessing | SAFE → GeoTIFF, subsetting, RGB / false-colour composites | Sentinel-2 | sen2r, terra, QGIS |
| 05 | Spectral indices & land cover | Spectral signatures, feature spaces, NDVI / EVI / PVI | Sentinel-2 | terra |
| 06 | Hyperspectral tree species | SVM & random forest classification of 5 tree species | HyMap (125 bands) | e1071, randomForest |
| 07 | EnMAP biomass regression | Random forest regression of above-ground biomass | EnMAP (195 bands) | randomForest, tidyterra |
| 08 | LiDAR DTM / CHM & tree segmentation | DTM, DSM, CHM, local-maximum tree tops, Dalponte segmentation | Berlin ALS | lidR |
| 09 | LiDAR metrics & forest stands | Pixel- and tree-level metrics vs. forest inventory | Berlin ALS, forest stands | lidR, ggplot2 |
| 10 | TLS stem detection | DBH / height from terrestrial laser scanning | TLS pine stand | CloudCompare, 3DFin |
| 11–12 | Land-cover digitizing | Training-data digitizing on RGB / CIR orthophotos | Berlin TrueDOP 20 cm | QGIS |
Rendered reports (no R needed): GitHub Pages
The code is in this repository; the input data (~3.5 GB, mostly point clouds and imagery) is attached to the
data-v1.0 release
and downloaded on demand with checksum verification.
git clone https://github.com/toihr/Advanced-Remote-Sensing-Exercises.git
cd Advanced-Remote-Sensing-Exercises
Rscript install_packages.R # R packages used in the exercises
Rscript download_all.R # all data – or per exercise, see belowPer exercise: open the .Rproj in the exercise folder and run source("get_data.R").
Without R: tools/download_data.sh exercises/<name> (Linux/macOS) or .\tools\download_data.ps1 exercises\<name> (Windows).
exercises/<nn-name>/
├── README.md task, methods, results
├── *.R / *.Rmd / *.qgz scripts, reports, QGIS projects
├── data/ input data (large files come from the release)
├── data-manifest.csv which release assets belong to this exercise
└── get_data.R downloads them
tools/ download helpers (R, bash, PowerShell) and release build script
docs/ rendered HTML reports (GitHub Pages)
Code: MIT. Data belongs to the respective providers – see DATA.md for sources and terms. Exercise tasks were set by the course; the solutions are my own work.