A Class-Incremental Object Detection Method for Remote Sensing Images Based on Dynamic Multiprototype Matching

被引:0
|
作者
Hu, Mingtao [1 ,2 ,3 ,4 ]
Yin, Wenxin [1 ,2 ]
Diao, Wenhui [1 ,2 ]
Gao, Xin [1 ,2 ]
Sun, Xian [1 ,2 ,3 ,4 ]
机构
[1] Chinese Acad Sci, Aerosp Informat Res Inst, Beijing 100190, Peoples R China
[2] Chinese Acad Sci, Aerosp Informat Res Inst, Key Lab Network Informat Syst Technol NIST, Beijing 100190, Peoples R China
[3] Univ Chinese Acad Sci, Beijing 100190, Peoples R China
[4] Univ Chinese Acad Sci, Sch Elect Elect & Commun Engn, Beijing 100190, Peoples R China
关键词
Prototypes; Object detection; Remote sensing; Incremental learning; Data models; Adaptation models; Training; Semantics; Computational modeling; Iron; Deep learning; incremental learning; object detection; remote sensing;
D O I
10.1109/JSTARS.2024.3520831
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
In object detection task, incremental learning method enables the previously trained model better adapt to the new task using either a small amount of old data or none at all. In the incremental training processes of complex remote sensing scenes, the newly arrived data only includes new classes annotations. These new classes may exhibit spatial overlap and shape similarity with old classes or may have been labeled as background in earlier tasks, leading to a unique challenge called class semantic confusion. To address this issue, this article dynamically generate multiple representative prototypes of the different categories for refined matching the objects. To improve the matching accuracy, prototype contrastive learning is employed for expanding the distance between dissimilar prototypes and reducing the distance between similar prototypes. Meanwhile, a category perception enhancement module is proposed to enhance the aware of old categories to mitigate catastrophic forgetting. Comprehensive experimental results demonstrate that our proposed method outperforms the current state-of-the-art class-incremental object detection methods in most experimental settings on DIOR and FAIR1M datasets.
引用
收藏
页码:5157 / 5171
页数:15
相关论文
共 50 条
  • [1] A Class-Incremental Detection Method of Remote Sensing Images Based on Selective Distillation
    Ruan, Hang
    Peng, Jian
    Chen, Ye
    He, Silu
    Zhang, Zhenshi
    Li, Haifeng
    SYMMETRY-BASEL, 2022, 14 (10):
  • [2] CLASS-INCREMENTAL LEARNING FOR REMOTE SENSING IMAGES BASED ON KNOWLEDGE DISTILLATION
    Song, Jingduo
    Jia, Hecheng
    Xu, Feng
    IGARSS 2023 - 2023 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, 2023, : 5026 - 5028
  • [3] Class-incremental object detection
    Dong, Na
    Zhang, Yongqiang
    Ding, Mingli
    Bai, Yancheng
    PATTERN RECOGNITION, 2023, 139
  • [4] Class-Incremental Recognition of Objects in Remote Sensing Images With Dynamic Hybrid Exemplar Selection
    Fu, Yimin
    Liu, Zhunga
    Wu, Changyuan
    Wu, Feiyan
    Liu, Meiqin
    IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS, 2024, 60 (03) : 3468 - 3481
  • [5] Enhancing class-incremental object detection in remote sensing through instance-aware distillation
    Feng, Hangtao
    Zhang, Lu
    Yang, Xu
    Liu, Zhiyong
    NEUROCOMPUTING, 2024, 583
  • [6] Historical Information-Guided Class-Incremental Semantic Segmentation in Remote Sensing Images
    Rong, Xuee
    Sun, Xian
    Diao, Wenhui
    Wang, Peijin
    Yuan, Zhiqiang
    Wang, Hongqi
    IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2022, 60
  • [7] Adaptive scale matching for remote sensing object detection based on aerial images
    Han, Lu
    Li, Nan
    Zhong, Zeyuan
    Niu, Dong
    Gao, Bingbing
    IMAGE AND VISION COMPUTING, 2025, 157
  • [8] Object Detection in Remote Sensing Images based on One-class Classification
    Wang, Tian
    Chen, Yang
    Qiao, Meina
    Zhu, Aichun
    Wang, Ziyu
    Snoussi, Hichem
    2017 CHINESE AUTOMATION CONGRESS (CAC), 2017, : 6583 - 6587
  • [9] Physical-Simulation-Based Dynamic Template Matching Method for Remote Sensing Small Object Detection
    Cao, Yaming
    Guo, Lei
    Xiong, Fengguang
    Kuang, Liqun
    Han, Xie
    IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2024, 62 : 1 - 14
  • [10] Class-Incremental Learning Based on Anomaly Detection
    Zhang, Lijuan
    Yang, Xiaokang
    Zhang, Kai
    Li, Yong
    Li, Fu
    Li, Jun
    Li, Dongming
    IEEE ACCESS, 2023, 11 : 69423 - 69438