Presentation Information
[A-14-23]A Proposal of an Automated Monitoring System for the Stenotus Rubrovittatus Using Edge AI and SAHI
◎△Kotaro Inoue1, Airi Kokuryo2, Ryuya Itano2, Ayato Enami3, Akihito Kohiga1, Takahiro Koita2 (1. Doshisha Univ., 2. Graduate School of Doshisha Univ., 3. AGRI-PASS Co.,Ltd.)
Keywords:
Edge AI,Pest Monitoring,Object Detection
The Stenotus rubrovittatus and other rice bugs are major pests that feed on rice grains during the ripening stage and cause dark brown spots on brown rice; even a small proportion of spotted grains lowers the rice grade, resulting in economic losses on the order of several billion yen annually across Japan. Currently, monitoring pest occurrence and determining the optimal timing for pest control depend on agricultural advisors patrolling pheromone traps and counting captures visually at intervals of several days. Because of these long intervals, migration peaks are often missed, making it difficult to base decisions on sufficient data. This study proposes a system in which an IoT camera and a Raspberry Pi are retrofitted onto an existing adhesive-net pheromone trap to capture images of the adhesive surface at fixed intervals. The captured images are processed on the edge device using the object detection model YOLO together with SAHI (Slicing Aided Hyper Inference), a method for small object detection, and the detection results are aggregated in the cloud and visualized on a web dashboard. Slicing Aided Fine-tuning (SF) is employed during training. The proposed system enables daily tracking of pest population dynamics and aims to support data-driven decisions on the optimal timing for pest control.