测绘科学技术学报2025，Vol.41Issue(2)：146-153,8.DOI:10.3969/j.issn.1673-6338.2025.02.006
一种改进YOLOv8的无人机红外影像目标轻量化精确检测方法
The Enhanced YOLOv8 Method for Lightweight and Precise Detection of Targets in UAV Infrared Images
摘要
Abstract
The detection of targets in UAV infrared images has extensive research requirements and application sce-narios in both agriculture and business domains.In order to address the current problem that the detection accuracy of targets in UAV infrared images is low,particularly for small targets,which makes it difficult to effectively detect them,as well as to facilitate model deployment,a lightweight accurate algorithm(YOLOV8-PFAF)for infrared target detection is proposed based on the YOLOv8.An additional prediction head specifically designed for small targets is incorporated into YOLOv8,and the Adaptive Spatial Feature Fusion(ASFF)strategy is introduced to en-hance this prediction head.As a result,it significantly improves the accuracy of infrared target detection with aver-age accuracy mean(when the intersection ratio threshold is 50%)increasing by 1.2%and combined average accu-racy mean(when the intersection ratio threshold from 50%to 95%)increasing by 2%.Furthermore,the C2f_PConv module are designed and integrated to effectively reduce feature map redundancy and minimize model size,which can fully meet the demand for model deployment in real-time detection.关键词
无人机红外影像/目标检测/YOLOv8算法/小目标/自适应空间特征融合策略/轻量化Key words
UAV infrared image/target detection/YOLOv8 algorithm/small target/adaptive spatial feature fu-sion strategy/light wight分类
测绘与仪器引用本文复制引用
郭海涛,张亦弛,陈明岩,朱坤,卢俊,周一..一种改进YOLOv8的无人机红外影像目标轻量化精确检测方法[J].测绘科学技术学报,2025,41(2):146-153,8.基金项目
国家自然科学基金项目(42301464 （）
42201443). （）
