广西科技大学学报2025，Vol.36Issue(3)：24-32,9.DOI:10.16375/j.cnki.cn45-1395/t.2025.03.004
Dense-YOLO:一种用于监测复杂场景灰飞虱虫害的检测算法
Dense-YOLO:a detection algorithm for monitoring the small brown planthopper in complex scenes
摘要
Abstract
The small brown planthopper,a typical field pest,not only jeopardizes the development of crops and causes significant economic losses to farmers,but also affects the normal food supply.The deep learning-based target detection algorithm can replace the tedious traditional manual observation and counting while maintaining low operational costs.In this paper,Dense-YOLO,a novel small target detection algorithm based on YOLOv5,was developed for accurately detecting and counting the small brown planthopper dispersed in complex fields with different species.CSPDenseNet was integrated into the YOLOv5 backbone network to facilitate the reuse of shallow features to improve the network's ability to detect small targets.The experimental results indicate that the detection of Dense-YOLO AP@0.5∶0.95,AR,and F1 scores can reach 52.4%,61.5%,and 56.6%respectively,which are 2.2,0.7,and 1.6 percentage points higher than YOLOv5.With an average inference latency of 58.82 ms,the model balances high precision with real-time processing capabilities,fulfilling requirements for dynamic pest monitoring in agricultural settings.Furthermore,this methodology exhibits transferability across different crop pest detection scenarios,highlighting its broader applicability in smart agricultural surveillance systems.关键词
灰飞虱/虫害监测/YOLO/密集单元/残差网络Key words
small brown planthopper/pest monitoring/YOLO/dense unit/residual network分类
植物保护学引用本文复制引用
李善达,王虎奇,丛佩超,冯浩,胥羽涛,李添恒..Dense-YOLO:一种用于监测复杂场景灰飞虱虫害的检测算法[J].广西科技大学学报,2025,36(3):24-32,9.基金项目
中央引导地方科技发展专项资金项目(桂科ZY19183003) （桂科ZY19183003）
广西重点研发计划项目(桂科AB20058001)资助 （桂科AB20058001）
