Introduction of a Transformer-based Future Prediction Method for Tracking Pollinating Insect Black Bumblebee Using Object Detection

カテゴリ：部門大会

論文No：SS1-8

グループ名：【C】2025年電気学会電子・情報・システム部門大会

発行日：2025/8/20

タイトル(英語)：Introduction of a Transformer-based Future Prediction Method for Tracking Pollinating Insect Black Bumblebee Using Object Detection

著者名：田中 友哉（千葉大学）,中間 公啓（千葉大学）,小圷 成一（千葉大学）

著者名(英語)： Tomoya Tanaka (Chiba University),Kimihiro Nakama (Chiba University),Seiichi Koakutsu (Chiba University)

キーワード：人工光型植物工場,クロマルハナバチ,物体検出,物体追跡,YOLOTransformer,Plant factry with artificial light,Black bumblebee,Object detection,Mutiple object tracking,YOLOTransformer

要約(日本語)：In artificial-light facilities, lettuce dominates because fruiting crops require pollination and are harder to cultivate. Many operations use honeybees, but they can be aggressive, sting workers, and slow down under low light. Black bumblebees, in contrast, are docile and highly photosensitive, remaining active in weak illumination. However, excessive flower visits damage blooms and reduce yield. Growers currently count bee traffic manually during multi-month harvests, but automating hive entry control could save labor.
Prior studies applied YOLO for detection and StrongSORT or DeepSORT for tracking, confirming feasibility but revealing that ID-switch errors account for most measurement inaccuracies. We propose a Transformer-based tracking method to reduce ID switches.

本誌掲載ページ：1779-1781p

原稿種別：英語

PDFファイルサイズ：320Kバイト