农机化研究2026，Vol.48Issue(6)：140-146,156,8.DOI:10.13427/j.issn.1003-188X.2026.06.018
基于改进YOLOv8s的果园场景下电线杆目标精准检测模型
Accurate Detection Model of Power Pole Target in Orchard Scene Based on Improved YOLOv8s
摘要
Abstract
The orchard environment was complex,and it was difficult to quickly and accurately identify micro-scale obsta-cles such as power lines when the UAV was flying.Through the accurate detection of the power pole,it could effectively avoid the power line obstacles and ensure the operation safety of the orchard drone.In this study,a precise detection model of orchard poles based on improved YOLOv8s(YOLOv8s-pole)was proposed.By using the RepViT-Block mod-ule to replace the main part,the model's ability to capture details in complex scenes was improved.The iRMB attention mechanism was integrated to promote the the model's feature learning ability and deep feature transmission ability.The network structure RepNCSPELAN4 was introduced to improve the feature extraction and fusion method,which enhance the sensitivity of the model to the target features.The improved dynamic convolution(DynamicConv)increased the robustness and context expression ability of the model.When using the YOLOv8s-pole model,the accuracy rate was 93.81%,the recall rate was 80.74%,the F1 score was 0.87 and the mAP@0.5 was 89.33%.Compared with the original YOLOv8 s model,the evaluation indexes of the improved model were increased by 4.92 percentage points,5.06 percentage points,0.05 and 5.06 percentage points,respectively.YOLOv8s-pole was superior to the existing algorithm model in the above four detection dimensions,which improved the detection accuracy of the model while taking into account the detection speed of the model,and could meet the actual use requirements of the detection of the wire rod in the complex environ-ment of the orchard.关键词
果园电线杆/目标检测/YOLOv8s/深度学习Key words
power pole in orchards/object detection/YOLOv8s/deep learning分类
农业科技引用本文复制引用
杨景,张亚莉,卢小阳,杨达成,兰玉彬,王林琳..基于改进YOLOv8s的果园场景下电线杆目标精准检测模型[J].农机化研究,2026,48(6):140-146,156,8.基金项目
国家重点研发计划项目(2023YFD2000202) （2023YFD2000202）
