Sensor-Based Characterisation of Plant Populations

Drone image from 10 m height showing grassland and unpacked sensor equipment distributed around a bus

Terrestrial laserscanner in an extensive grassland site invaded by lupine

Examples of sensor data: Spectral, point clouds, objects and texture data

Sensors and platforms: 3D laser, RGB, multispectral, thermal, satellite and hyperspectral sensors and platforms

The research unit “Sensor-Based Characterisation of Plant Populations” utilises spatial data from different sensors to analyse parameters such as yield, forage quality and sward composition of plant populations. Based on studies using ultrasonic sensors for canopy height determination, laser and spectral sensors have been employed to measure agronomically relevant plant traits at the GNR working group since around 15 years ago.

A core theme of the research unit is the analysis of grassland and arable plant populations using airborne (e.g., drone) multispectral and hyperspectral images. 3D models derived from a terrestrial laser scanner and photogrammetric methods are exploited to monitor plant biophysical characteristics (e.g., height, biomass). Additionally, space-borne images are utilised for land-use/land-cover analysis on a regional scale. State of art machine learning (ML) and deep learning (DL) methods are explored for model development, and spatial analyses are conducted using Geographic information systems (GIS).

Fields of application of the methods mentioned above are:

conservation of grassland biodiversity (DBU Rhön)
analysis of spatio-temporal variability of plant populations (SIGNAL, GreenDairy)
land-use/land-cover analysis (FOR2432, SYMOBIO)

Contact person: Dr. Jayan Wijesingha

3D animation of a scanned 8 to 8 m grassland plot in early spring, followed by the same plot in june and a detailed cut out of 1 square meter with high detailed point cloud information