江苏农业学报2025，Vol.41Issue(2)：313-322,10.DOI:10.3969/j.issn.1000-4440.2025.02.012
基于轻量化改进YOLOv8模型和边缘计算的玉米病虫害检测系统
Maize pest and disease detection system based on lightweight improved YOLOv8 model and edge computing
摘要
Abstract
In order to achieve in-situ accurate detection and identification of maize pests and diseases,this study de-signed an intelligent detection system for maize pests and diseases based on edge computing.The system was improved based on the YOLOv8 model with specific improvement methods,including adopting the Efficient Vision Transformer(EfficientViT)as the backbone network to reduce the com-putational load,introducing Ghost Convolution(Ghost-Conv)into the feature fusion network to further reduce the computational burden,introducing Spatial-Channel Convo-lution(SCConv)into the C2f module to enhance the feature extraction performance,and replacing the loss function with the Wise Intersection over Union(WIoU)loss function that had a dynamic non-monotonic focusing mechanism to improve the recognition accuracy of the model.At the same time,this study designed the upper and lower computer architectures of the pest and disease detection system based on edge computing and deployed this lightweight model to the Jetson Orin Nano edge computing device.The system used Pyside6 to develop a visual interface.In addition to the recognition and training functions,it also integrated an AI expert library based on large-model technology,which could realize intelligent diagnosis of pests and diseases.The improved model YOLOv8-EGCW was tested using a self-built maize pest and disease dataset.The results showed that compared with the original YOLOv8m model,the precision,recall rate,and mean average precision of the improved model YOLOv8-EGCW increased by 0.4 percentage points,1.6 percentage points,and 1.2 percentage points,respectively.The number of parameters and the model size were greatly reduced,and the detection time for a single image was shortened.The test results of the established corn pest and disease detection system indicated that the accuracy rate reached 93.4%and the detection speed reached 25 frames per second.These results indicated that the system could meet the requirements of in-situ detection of maize pests and diseases in the edge computing environment.关键词
玉米/病虫害检测系统/YOLOv8模型/轻量化改进/边缘计算Key words
corn/pest and disease detection system/YOLOv8 model/lightweight improved model/edge computing分类
农业科技引用本文复制引用
施杰,熊凯祥,李志,陈立畅,唐秀英,杨琳琳..基于轻量化改进YOLOv8模型和边缘计算的玉米病虫害检测系统[J].江苏农业学报,2025,41(2):313-322,10.基金项目
云南省重大科技专项(202302AE090020) （202302AE090020）
云南省农业基础研究联合专项(202401BD070001-069) （202401BD070001-069）
云南省作物生产与智慧农业重点实验室开放课题 （）
