Logo des Repositoriums

Auflistung nach Autor:in "Menz, Patrick"

1 - 3 von 3
  • Konferenzbeitrag
    AI-supported selection procedure for spectral sensors based on technical and economic characteristics
    (INFORMATIK 2024, 2024) Menz, Patrick; Klein, Lauritz; Herzog, Andreas
    This study presents an AI-supported spectral sensor selection process that combines technical and economic criteria to recommend the optimal sensor for specific applications, such as quality control of roasted coffee beans. Using a comprehensive database of spectral sensor characteristics, the SMART algorithm guides decisions that focus on both performance and cost-effectiveness. Our methodology involves simulating spectral responses and using an AI model to evaluate sensor effectiveness in classifying coffee bean types. Initial results highlight the method's ability to optimise sensor selection, effectively balancing performance with budget considerations, and underscore its potential to improve user decision making in technology applications and enhance their digital sovereignty.
  • Konferenzbeitrag
    Drohnenbasiertes Verfahren zur Detektion geschädigter Obstbäume in Obstbaumplantagen
    (INFORMATIK 2023 - Designing Futures: Zukünfte gestalten, 2023) Thielert, Bonito; Menz, Patrick; Warnemünde, Sebastian; Holstein, Katharina; Klein, Lauritz; Kilias, David; Runne, Miriam; Jarausch, Wolfgang; Knauer, Uwe
    Phytoplasmen-induzierte Erkrankungen von Obstbäumen stellen eine große Herausforderung im europäischen Obstanbau dar. Apfeltriebsucht und Birnenverfall zählen zu den wirtschaftlich relevantesten Obstkrankheiten. Im Rahmen des Projekts „Digitaler Obstbau“ wurden verschiedene Technologien zur Diagnose der Krankheiten untersucht und weiterentwickelt. Die drohnenbasierte Bonitur der genannten Krankheiten ermöglicht die flächendeckende räumlich hochauflösende Erfassung von Plantagen und die Symptomerkennung für einzelne Obstbäume. Untersuchungen mit der Hyperspektralkamera Cubert UHD-185 Firefly und dem integrierten Phantom 4 Multispectral Aufnahmesystem zeigen die Durchführbarkeit der drohnenbasierten digitalen Bonitur. Die Daten zeigen mit 76 % eine gute Klassifikationsrate zur Erkennbarkeit der Krankheitssymptome.
  • Konferenzbeitrag
    PhenoTruckAI: On-Site Hyperspectral Measurement for Distinction of Quarantine Grapevine Disease “flavescence dorée” and non-Quarantine Disease “bois noir” in a Mobile Laboratory
    (INFORMATIK 2024, 2024) Thielert, Bonito; Menz, Patrick; Götte, Gesa; Runne, Miriam; Michel, Markus; Wagner, Sylvia; Jarausch, Wolfgang; Warnemünde, Sebastian
    German wine growing regions are threatened by the expected occurrence of the quarantine phytoplasma disease “flavescence dorée (FD)”. As a fast and reliable extension for FD monitoring in the field, hyperspectral imaging using machine learning (ML) based data processing has been assessed for its potential to detect FD and to distinguish it from the less damaging phytoplasma disease “bois noir (BN)”. As FD is not yet present in Germany, the study has been conducted in Northern Italy in a mobile lab. The best models reached a high phytoplasma detection accuracy of 94.9% and 97.8% for the visible to near-infrared (VNIR) and the short-wavelength spectral range (SWIR), respectively. The distinction accuracy to BN reached 79.9% (VNIR) and 79.3% (SWIR). Both, the practicability performing hyperspectral measurements in a sovereign mobile lab and the applicability of hyperspectral sensor systems using ML for detection and distinction of FD and BN phytoplasmas has been shown.