Digital weed management – New trends for weed scoring in sugar beet

Autor/in: Heim, R., S. Streit, D. Koops, M. T. Kuska, S. Paulus
Jahr: 2022
Zeitschrift: Sugar Industry 147(6)
Seite/n: 343-351, doi.org/10.36961/si28804
Stichworte: digital tools, machine learning, precision agriculture, remote sensing, smart farming

Published in Abteilung Sensorik & Datenanalyse

Abstract

Weed scoring is crucial to test the efficacy of herbicides and other weed management methods but has proven to be labor intense and variable across scoring individuals and time. This article provides a comparison of new approaches of weed scoring based on digital tools to conventional visual scoring. The first method collected aerial imagery that was then scored manually on a computer screen. The second method is a machine learning approach, automatically detecting and counting weeds in aerial imagery. The reference method is a conventional visual scoring by human raters. Across all scoring methods, the results show similar patterns, but the total number of scored plants differs. In comparison to the digital approaches, in field scoring by human raters estimated a higher weed infestation. Possible reasons for this, as well as the advantages and disadvantages of each method are discussed to explore new modes of weed scoring. A clear benefit of digital scoring is the potential to automate the procedure and its objective, repeatable nature

FaLang translation system by Faboba