A region and contour based technique for automatic detection of tomato roots in minirhizotron images
Year: 2025
Type: article
Abstract: The use of minirhizotrons is a common tool to investigate fine roots turn-over in a variety of ecosystems. Manual analysis of the resulting images is very time consuming, however. An improved approach for automatic root detection combining region and contour based techniques was developed. The first step is aimed at detecting easily recognizable parts of the root (optimally at least one part of each root), wich will be called a R-SEED subsequently in this paper. In a second phase each R-SEED is expanded with a more sophisticated algorithm to find the boundaries of the complete root. 14 tome series with 3 at 8 images of tomato roots were digitized, processed and used to validate the approach. As the R-SEED detection influences the overall performance strongly the approach benefits from the improved algorithm. In four out of five images from one time series examined in dtail the root area detected is nearly correct with analysis relative error between -8.8% and +7.3%. in one image the detection fails substantially due to decomposed roots. In general, the root area detected is systematically lower than the area marked manually.
Source: Angewandte Botanik
Authors: Erz, Gregor, Veste, Maik, Anlauf, Heiko, Breckle, Siegmar-Walter, Posch, Stefan
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: diamond
OpenAlex ID: W2490174744