Research on Intelligent Detection Method of Mold and Insect Pests in Grain Storage Based on YOLOv5 and Large Model Fusion

DOI: https://doi.org/10.62517/jbdc.202501209

Author(s)

Zhiduo Bo, Cancan Zhou, Mingbo Yang

Affiliation(s)

School of Artificial Intelligence and Big Data Henan University of Technology Zhengzhou, Henan, China

Abstract

In order to solve the problem that the detection of moldy insect pests in grain storage relies on manual work, is inefficient and has insufficient generalization ability, this study proposes an intelligent detection method based on the fusion of YOLOv5 and a large model in order to achieve high-precision real-time detection and early warning of moldy insect pests in grain storage. In this study, firstly, the small target detection capability is improved by constructing a stored grain dataset containing a total of 3,564 images of three categories, namely, insect, moldy, and deteriorated, and adopting data enhancement and adaptive anchor frame optimization strategies. Based on the YOLOv5s model for training, combined with CIoU loss function and multi-scale feature fusion mechanism, The model achieved 92.3% on the test set mAP@0.5, the accuracy rate reaches 89.7%. Further decision optimization is carried out by Starfire large model (Max-32K), using role-setting stereotyped Prompt with parameter tuning (temperature=0.3, top_k=3), and experiments show that the comprehensive score of prevention and control decision generated by this method is significantly better than that of the traditional method. This study provides an efficient and reliable solution for the intelligent management of grain silos, which has practical application value.

Keywords

Yolov5; Large Model Fusion; Stored Grain Mold and Insect Pests; Intelligent Detection; Parameter Optimization

References

[1]World Health Organization. The State of Food Security and Nutrition in the World 2023: Urbanization, agrifood systems transformation and healthy diets across the rural–urban continuum. Food & Agriculture Org., 2023. [2]Lipinski B, et al. "SDG Target 12.3 on Food Loss and Waste: 2019 Progress Report." World Resources Institute, 2019. [3]Jie SUN, Qing-he XU, Xiao-yue LIU. Design of Remote Grain Monitoring System Based on NB-IoTTechnology, Automation & Instrumentation, 2020, 35(4): 86-89, 93. [4]FENG Chao, XU Yan-lei. Design of Remote Monitoring System for Granary Based on Internet of Things, Research of Agricultural Modernization, 2017, 38(2): 328-334. [5]Debapriya M, Soyeb N, Manu M, Deepak P, et al. YOLO-Pose: Enhancing YOLO for Multi Person Pose Estimation Using Object Keypoint Similarity Loss, Computer Vision and Pattern Recognition, 2022. [6]Tang Shiyi, Zhang Shu, Fang Yini.HIC-YOLOv5: Improved YOLOv5 for Small Object Detection, IEEE International Conference on Robotics and Automation, 2024. [7]Zhu Shucheng, Huo Hongying, Wang Weikang, et al. Automatic optimization and evaluation of cue fairness based on large language model itself. Computer Science,1-14[2025-04-15].http://kns.cnki.net/kcms/detail/50.1075.TP.20241122.0938.002.html. [8]Special issue on “Decision Support System Technology in the Artificial Intelligence Era”. International Transactions in Operational Research, 2025, 32(5): 3202-3202. [9]He Y, Feng A, Gao Z, et al. LIPT: Improving Prompt Tuning with Late Inception Reparameterization. Electronics, 2020.13(23):4741-4741. [10]Qin Donghong, Li Zhengtao, Bai Fengbo, et al. Computer Engineering and Application,1-30[2025-4-22].http://kns.cnki.net/kcms/detail/11.2121.Tp.20250326.1554.025.html.