The Würzburg State Fire Service School is supporting the project as an associated partner.

In October 2025, a test of the forest fire early warning system using drones was conducted in Rotenburg an der Fulda. During the test, the performance of a special drone equipped with thermal imaging and optical sensors was demonstrated. The drone detected a fire being controlled by the volunteer fire department within seconds and transmitted the data to the relevant platform in real time.

Dr. Peter Hofmann from the Deggendorf Institute of Technology explained how the institute supports the project and how eCognition can effectively analyze forest structures and detect potential forest fire hazards. In addition to the recorded forest structure, consideration is given to aspects such as topography, season, weather, and much more. The remote sensing data obtained with the drone can be automatically evaluated using eCognition. The results of this evaluation can then also be used in forest management or for further spatial analyses.

Click here for more information on the KIWA Project.

show less

Let’s meet and talk on September 29 and 30, 2025 at the ‘Annual Meeting AK Remote Sensing 2025’ in Bochum.

show less

Landscape-scale modelling of pesticide exposure

Landscape-scale modelling of pesticide exposure and species effects (risk) requires corresponding geoinformation, however, base data is often not readily available in required resolution (≈10 m or higher)​. Hence, realistically applicable (availability, cost, quality, geographic coverage, collaboration) technology and workflows are needed for a range of risk assessment projects.

Project Goals

The goals of this project were:

1. Extract ‘riparian’ and ‘olives’ vegetation within a 100 m buffer to water bodies​ ​for different catchments in Spain and Greece using free high-resolution satellite data (10 m)
2. Precise extraction of olive field outlines in very high spatial resolution commercial satellite data (3 m)

Tama Group utilized provided catchment data to retrieve appropriate satellite scenes and performed the land cover classification based on time series information, including: cloud masking, atmospheric correction, pan-sharpening, layer calculations, feature importance calculation, object-based classification, accuracy check, and transferability tests. The most promising results were achieved combining object-based image analysis and convolutional neural networks.

show less

Various process steps, such as the purification or drying required in the context of seed breeding or production, rais analytical questions for individual components of a seed mixture: what is the ratio of the individual types of this specific mix, or what is the relationship between the seed sizes of the different varieties?

Many of these questions can be answered qualitatively well by visual inspection, but only a more in-depth quantitative analysis allows for further process optimization and automation for large quantities.

RGB Data for Single Grain Extraktion

To characterize this cognitive-quantitative analysis in the field of agriculture, the Tama Group has made an example of the automatic detection of grains of a lens-sown-seed mixture. The aim of the automatic detection was the creation of individual objects, the calculation of the proportion of dark grains in the total number and the statistical coverage of different grain features. For this purpose, the automated extraction of individual grains on a white background was carried out on the basis of a photo of a digital camera.

Process Optimization through Tama Group Methodoly

A challenge for the automatic grain extraction in this example is mainly dense grain clusters, which make the detection of single grains difficult; nevertheless, meaningful results could be achieved by using high-performance algorithms. Comprehensive statements on the grain characteristics can be made with the help of the combination of grid-based additional information as well as form and neighborhood criteria. A few examples can be found in the adjacent table.

show less

For every aerial imaging service provider, detailed city mapping based on high-resolution imagery may become a very tedious task if image analytics are not highly automated. Tama Group now offers a structured methodology for aerial imagery service providers to increase the level of automation for city mapping projects. The methodology centers on automated extraction of green areas and shadowed areas.

We used for example high resolution airborne orthophotos and a digital surface model of the city Bremen, Germany, to extract green areas and shadowed regions in combination with a building mask. In this example we concentrate on three different study sites, urban, industrial, and rural, characterized by dissimilar amounts of vegetation and impervious surfaces as well as different building types.

Versatility Methodologies

The whole process chain is designed for very fast performance, classification accuracies still range from 95 up to 99%. All the layers created may be mixed and matched with GIS data and other thematic layers. The versatility of this methodologies allows for many different use cases, both for aerial imaging service providers as well as consumers of the information extracted. All steps introduced are fully automated and can be adapted for large scale batch processing.

show less