Integrating an AI-based insect detection and identification system to optimize grain facility management

Abstract

This study introduces a novel AI-driven insect detection and classification mechanism for grain storage and food facilities. The system was developed to efficiently capture and analyze insect images, addressing the growing issue of insect infestations that compromise the quality of stored grains. It harnesses the Jetson Nano's power-efficient single-board computer's capabilities in conjunction with a manual-focus camera. Central to this innovation is a methodically trained CNN that proficiently discerns and categorizes distinct stored grain insect pests, emphasizing adult warehouse and cigarette beetles. The methodology adopted for this study involved rigorous calibration and validation phases. A defining feature of this research remains the exhaustive evaluation under varied lighting conditions, specifically white Light Emitting Diode (LED), yellow LED, and total darkness, which underscores the system's resilience and adaptability. Comparative studies were conducted to benchmark the model against traditional detection methods. When assessed using F1 scores, the system demonstrated superior detection and species classification precision. This innovation, while notably affordable due to the Jetson Nano's cost-effectiveness, has the potential to redefine the paradigms of insect management across stored product facilities. Its real-time analytics capability empowers industry stakeholders, providing swift decision-making tools to reduce pest management costs, improve operational efficiency, and preserve stored produce quality. Furthermore, the scalable nature of this system holds promise for broader applications, signaling a potential shift in how the larger agriculture sector approaches pest surveillance and control measures.