arXiv 2605.207382026-05-20cs.CV

STAR-IOD: 无尺度耦合拓扑对齐与伪标签细化用于遥感增量目标检测

STAR-IOD: Scale-decoupled Topology Alignment with Pseudo-label Refinement for Remote Sensing Incremental Object Detection

School of Computer Science, Northwestern Polytechnical University, Xi’an 710072, China（计算机科学学院，西北工业大学，西安710072，中国）；School of Artificial Intelligence, OPtics and ElectroNics (iOPEN), Northwestern Polytechnical University, Xi’an 710072, China（人工智能学院，光学与电子学（iOPEN），西北工业大学，西安710072，中国）

Yaoteng Zhang, Qing Zhou, Junyu Gao, Qi Wang

备注：: STAR-IOD was accepted by ISPRS Journal of Photogrammetry and Remote Sensing

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出STAR-IOD框架，通过子空间解耦拓扑蒸馏模块和聚类驱动伪标签生成器，解决遥感增量目标检测中类别间拓扑关系对齐和尺度变化导致的表示差异问题，同时通过动态识别类别特定阈值来缓解旧类标注缺失问题，实验表明在DIOR-IOD和DOTA-IOD数据集上，方法在mAP上分别优于现有方法1.7%和2.1%。

摘要：

遥感影像通常以连续数据流的形式出现。传统检测器在学习新类别时往往会遗忘之前学习的类别；因此，研究遥感增量目标检测（RS-IOD）具有重要意义。然而，现有方法大多忽视了遥感场景中普遍存在的类别内尺度变化，这削弱了知识迁移和旧知识保留的有效性。此外，RS-IOD还受到标注缺失的影响，导致模型将旧类实例误分类为背景。为了解决这些挑战，我们提出了一种新的框架STAR-IOD。首先，我们引入了子空间解耦拓扑蒸馏（STD）模块，以转移结构知识，显式对齐类别间拓扑关系，并缓解由尺度变化引起的类别内表示差异。此外，我们引入了聚类驱动伪标签生成器（CPG），这是一个即插即用模块，利用K-Means聚类动态识别类别特定阈值，从而保证真正阳性目标与背景噪声之间的准确区分，并缓解旧类标注缺失问题。我们还构建了两个遥感增量目标检测数据集，DIOR-IOD和DOTA-IOD，以促进RS-IOD的研究。广泛的实验表明，我们的方法在DIOR-IOD和DOTA-IOD数据集上分别以1.7%和2.1%的mAP优于现有方法，有效缓解了灾难性遗忘，同时在基础类和新类上保持了强劲的检测性能。代码和数据集已发布在：https://github.com/zyt95579/STAR-IOD。

英文摘要：

Remote sensing imagery typically arrives in the form of continuous data streams. Traditional detectors often forget previously learned categories when learning new ones; therefore, research on Remote Sensing Incremental Object Detection (RS-IOD) is of great significance. However, existing methods largely overlook the intra-class scale variations prevalent in remote sensing scenes, which undermines the effectiveness of knowledge transfer and old knowledge preservation. Moreover, RS-IOD also suffers from missing annotations, which cause the model to misclassify old-class instances as background. To address these challenges, we propose a novel framework, STAR-IOD. First, we introduce a Subspace-decoupled Topology Distillation (STD) module to transfer structural knowledge, explicitly aligning inter-class topological relationships and mitigating intra-class representation discrepancies induced by scale shifts. Furthermore, we introduce the Clustering-driven Pseudo-label Generator (CPG), a plug-and-play module that leverages K-Means clustering to dynamically identify class-specific thresholds, thereby guaranteeing an accurate distinction between true positive targets and background noise and alleviating the issue of missing annotations for old classes. We also constructed two Remote Sensing Incremental Object Detection datasets, DIOR-IOD and DOTA-IOD to facilitate research on RS-IOD. Extensive experiments demonstrate that our method outperforms state-of-the-art approaches by 1.7% and 2.1% mAP on DIOR-IOD and DOTA-IOD, respectively, effectively alleviating catastrophic forgetting while preserving strong detection performance on both base and novel classes. The code and dataset are released at: https://github.com/zyt95579/STAR-IOD.

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。