Related Experiment Video
Updated: Jul 8, 2026
08:16
Collecting and Processing Drone-based Remotely Sensed Data for Use in Forest Recovery Monitoring
Published on: October 24, 2025
Semi-Automated Method for Mapping and Classifying Boreal Coastal Wetland Plant Communities using Drone and Ground
Vineta Gailite1, Raul Sampaio de Lima2, Kaupo Kokamägi2
1Institute of Agriculture and Environmental Sciences, Estonian University of Life Sciences; vineta.gailite@emu.ee.
Journal of Visualized Experiments : Jove
|July 6, 2026
Summary
Semi-automated mapping using Unmanned Aerial Vehicle (UAV) surveys and random forest classification accurately identifies coastal wetland plant communities. RGB cameras offer superior accuracy for this ecological monitoring task.
Area of Science:
- Ecology
- Remote Sensing
- Artificial Intelligence
Background:
- Coastal wetlands require detailed monitoring for effective management and restoration.
- Traditional ecological surveys are time-consuming and lack the spatial resolution for fine-scale mapping.
- Unmanned Aerial Vehicle (UAV) photogrammetry offers a flexible and high-resolution alternative for ecosystem monitoring.
Purpose of the Study:
- To develop and validate a semi-automated protocol for mapping coastal wetland plant communities.
- To compare the effectiveness of multispectral and RGB sensors for UAV-based vegetation mapping.
- To integrate ground-truth data with AI classification for accurate ecological assessments.
Main Methods:
- A four-phase protocol involving UAV aerial surveys, stratified quadrat ground sampling, vegetation index calculation, and random forest (RF) classification.
- Utilized R packages for data processing, including terra, sp, sf, rgdal, raster, rsample, MLmetrics, and randomForest.
- Tested both multispectral and RGB sensors, generating numerous vegetation indices and a digital surface model (DSM).
Main Results:
- The RGB dataset achieved over 98% accuracy with a 1.14% out-of-bag (OOB) error, outperforming the multispectral dataset (92.3% accuracy, 7.75% OOB error).
- Both sensor types proved suitable for mapping plant communities, with the RGB system demonstrating higher performance.
- Generated georeferenced datasets suitable for integration into a Geographical Information System (GIS) project.
Conclusions:
- Semi-automated mapping using UAVs and RF classification is effective for detailed coastal wetland plant community assessment.
- RGB sensors provide a highly accurate and efficient option for this type of ecological monitoring.
- The developed protocol and resulting GIS project support ecosystem research, management, restoration planning, and historical land cover tracking.