An Infrared Target Detection Algorithm for UAVs Based on Lightweight YOLOv8s
HUANG Zedong
JIAO Wenwen
SU Mingxuan
LIU Jinfu
Abstract：In response to the challenges of target detection posed by dense,small targets and complex backgrounds in UAV aerial infrared images,this paper presents an improved algorithm named EUAV_YOLOv8s,which is based on YOLOv8.Firstly,the algorithm incorporates the SEGS module to enhance the extraction of contextual semantic information by embedding the Squeeze-and-Excitation(SE)attention mechanism in the C2f module and integrating GhostNet to reduce model size.Secondly,a tiny object detection head has been added and a lightweight decoupled head structure has been designed to reduce model complexity without compromising detection accuracy.Finally,the Wise Intersection over Union(WIoU)loss function has been incorporated to enhance the performance of bounding box regression.Experimental results demonstrate that the improved algorithm achieves an mAP@50％of 83.3％and a detection speed of 373 FPS,significantly surpassing the performance of mainstream algorithms.
Keywords：YOLOv8slightweighttarget detectionUnmanned Aerial Vehiclesinfrared image
Publication Date：2025-12-30
Online Publishing Date：2026-01-17(First online date of this platform, not the publication date of the document)
Pages：7( 57-63 )