Machine learning workflows detect ground-dwelling arthropods against heterogenous backgrounds
Journal of Insect Science · 2026
DOI: 10.1093/jisesa/ieag091 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Insect Science, 2026. Verfasst von Robert J Boyd und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Biological Sciences, Kent State University, Kent, OH, USA. Hauptthemen laut MEDLINE: Arthropods, Machine Learning, Image Processing, Computer-Assisted, Photography. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: US National Science Foundation, Greenacres Foundation Ecology & Environment.
Abstract (Originalsprache)
Camera-based monitoring of arthropods is an emerging tool for studying activity, abundance, and behavior. However, many current approaches rely on custom-built camera systems and complex analytical workflows, which limit accessibility and broader adoption. In addition, most arthropod camera-monitoring systems image subjects against uniform backgrounds, conditions that may alter natural activity patterns and reduce ecological realism. Here, we present a camera-monitoring workflow designed to increase accessibility by using commercially available game cameras coupled with a streamlined image-processing pipeline. We evaluated the performance of this approach by imaging arthropods against a complex, heterogeneous background to assess its effectiveness under conditions more representative of field environments. Our object-detection model performed reliably across evaluation metrics and accurately identified arthropods captured in the images despite background complexity. These results demonstrate that low-cost, widely available camera systems can be integrated with simplified analytical workflows to enable effective arthropod monitoring without specialized hardware. This approach provides a scalable and accessible method for studying arthropods in more natural contexts and may facilitate broader adoption of camera-based monitoring in ecological research and biodiversity assessments.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Biological Sciences, Kent State University, Kent, OH, USA.
- Department of Entomology, The Ohio State University, Wooster, OH, USA.
- Kent State University Environmental Science and Design Research Institute, Kent, OH, USA.
ORCID-Profile: Kayla I Perry, Thomas P Franzem, Christie Bahlai
Förderung
- US National Science Foundation
- Greenacres Foundation Ecology & Environment
Zitieren
APA
Boyd, R. J., Perry, K. I., Franzem, T. P., & Bahlai, C. (2026). Machine learning workflows detect ground-dwelling arthropods against heterogenous backgrounds. Journal of Insect Science. https://doi.org/10.1093/jisesa/ieag091
Vancouver
Boyd RJ, Perry KI, Franzem TP, Bahlai C. Machine learning workflows detect ground-dwelling arthropods against heterogenous backgrounds. Journal of Insect Science. 2026. doi:10.1093/jisesa/ieag091
BibTeX
@article{Boyd2026,
author = {Boyd, Robert J and Perry, Kayla I and Franzem, Thomas P and Bahlai, Christie},
title = {Machine learning workflows detect ground-dwelling arthropods against heterogenous backgrounds},
journal = {Journal of Insect Science},
year = {2026},
doi = {10.1093/jisesa/ieag091}
}
EUVIMED