arXiv 2510.005472025-10-01cs.CVcs.AI

Forestpest-YOLO：一种面向小型林业害虫的高性能检测框架

Forestpest-YOLO: A High-Performance Detection Framework for Small Forestry Pests

School of Advanced Manufacturing, Guangdong University of Technology（广东技术大学先进制造学院）；Guangdong Power Grid Co.,Ltd.（广东电网公司）；School of Computer Science and Engineering, Huizhou University（惠州市大学计算机科学与工程学院）；Obstetrics and Gynaecology, Huizhou First Maternal and Child Health Care Hospital（惠州市第一人民医院妇产科）

Aoduo Li, Peikai Lin, Jiancheng Li, Zhen Zhang, Shiting Wu, Zexiao Liang, Zhifa Jiang

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对复杂林业遥感环境下小型害虫检测难的问题，本文提出基于YOLOv8优化的Forestpest-YOLO框架，通过三项创新模块提升检测效果，在自建数据集上达到SOTA性能。

摘要：

利用遥感影像在复杂林业环境中检测农业害虫是生态保护的基础工作，但该任务目前受到诸多实际挑战的严重制约。检测目标通常体积极小、遮挡严重，且与杂乱背景视觉特征高度相似，导致传统目标检测模型因细粒度特征丢失、无法应对极端数据不平衡问题而表现不佳。为克服这些障碍，本文提出Forestpest-YOLO，这是一款针对林业遥感场景特性精心优化的检测框架。该框架以YOLOv8架构为基础，引入了三项协同创新：首先集成无损下采样模块SPD-Conv，确保小型目标的关键高分辨率细节在整个网络中得以保留；其次搭配新型跨阶段特征融合模块CSPOK，在抑制背景噪声的同时动态增强多尺度特征表征；最后采用VarifocalLoss优化训练目标，促使模型聚焦于高质量且难分类的样本。在我们自建的高难度ForestPest数据集上开展的大量实验表明，Forestpest-YOLO取得了当前最优性能，在小型、受遮挡害虫检测方面提升显著，性能大幅优于现有基准模型。

英文摘要：

Detecting agricultural pests in complex forestry environments using remote sensing imagery is fundamental for ecological preservation, yet it is severely hampered by practical challenges. Targets are often minuscule, heavily occluded, and visually similar to the cluttered background, causing conventional object detection models to falter due to the loss of fine-grained features and an inability to handle extreme data imbalance. To overcome these obstacles, this paper introduces Forestpest-YOLO, a detection framework meticulously optimized for the nuances of forestry remote sensing. Building upon the YOLOv8 architecture, our framework introduces a synergistic trio of innovations. We first integrate a lossless downsampling module, SPD-Conv, to ensure that critical high-resolution details of small targets are preserved throughout the network. This is complemented by a novel cross-stage feature fusion block, CSPOK, which dynamically enhances multi-scale feature representation while suppressing background noise. Finally, we employ VarifocalLoss to refine the training objective, compelling the model to focus on high-quality and hard-to-classify samples. Extensive experiments on our challenging, self-constructed ForestPest dataset demonstrate that Forestpest-YOLO achieves state-of-the-art performance, showing marked improvements in detecting small, occluded pests and significantly outperforming established baseline models.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。