본문 바로가기
  • Home

A Real-Time Wildlife Roadkill Detection Approach Using YOLOv8

  • Journal of Internet of Things and Convergence
  • Abbr : JKIOTS
  • 2024, 10(5), pp.185-196
  • Publisher : The Korea Internet of Things Society
  • Research Area : Engineering > Computer Science > Internet Information Processing
  • Received : September 3, 2024
  • Accepted : October 8, 2024
  • Published : October 31, 2024

KimHaeSoung 1 Jihoon Moon 1

1순천향대학교

Accredited

ABSTRACT

Increasing urbanization and road development have led to severe wildlife roadkill incidents, posing a significant social challenge. Especially in Korea, the high road density increases the exposure of wildlife to roadkill risks. This study proposes a wildlife object detection system using the latest object detection technology, YOLOv8, to mitigate domestic wildlife roadkill incidents. Using Roboflow for preprocessing, six mammal species were selected and the YOLOv8 model was trained, achieving high accuracy with mAP50 of 0.986 and mAP50-95 of 0.86. Comparative experiments with YOLOv5 and YOLOv7 demonstrated the superior performance of YOLOv8. The proposed system effectively detects animals even with protective coloration, contributing to the reduction of roadkill incidents and assisting in the design and placement of ecological corridors and guide fences. This research provides an important technological foundation for improving wildlife safety and maintaining ecological balance in urbanized areas.

Citation status

* References for papers published after 2023 are currently being built.