哈尔滨商业大学学报（自然科学版）2025，Vol.41Issue(6)：651-660,10.
基于MPC-YOLOv8算法的遥感影像目标检测
Object detection of remote sensing imagery based on MPC-YOLOv8 algorithm
张凯 1于瓅1
作者信息
- 1. 安徽理工大学 计算机科学与工程学院,安徽 淮南 232001
- 折叠
摘要
Abstract
To address the challenges of significant scale variations and noise interference in optical remote sensing image target detection,this paper proposed an enhanced YOLOv8 algorithm.The backbone network integrated an improved PKIBlock by removing the feed-forward network(FFN)to optimize computational efficiency and redesigned contextual anchor attention(CAA)for comprehensive dependency modeling through joint long-and short-range context.The neck network incorporated median-enhanced channel attention to suppress noise amplification during feature fusion,along with a dynamic multi-scale fusion module for adaptive weighted integration of denoised features.Evaluations on DIOR and NWPU VHR-10 datasets demonstrated mAP50 improvements of 1.5%and 4.9%,and mAP50～95 gains of 2.0%and 1.8%,respectively,confirming enhanced detection accuracy in noisy multi-scale scenarios.关键词
深度学习/YOLOv8/多尺度融合/中值池化/图像去噪/遥感影像Key words
deep learning/YOLOv8/multi-scale fusion/median pooling/image denoising/remote sensing imagery分类
信息技术与安全科学引用本文复制引用
张凯,于瓅..基于MPC-YOLOv8算法的遥感影像目标检测[J].哈尔滨商业大学学报（自然科学版）,2025,41(6):651-660,10.
