Label Evolution Based on Local Contrast Measure for Single-Point Supervised Infrared Small-Target Detection

被引:5
作者
Yang, Dongning [1 ,2 ]
Zhang, Haopeng [1 ,2 ]
Li, Ying [3 ]
Jiang, Zhiguo [1 ,2 ]
机构
[1] Tianmushan Lab, Hangzhou 311115, Peoples R China
[2] Beihang Univ, Image Proc Ctr, Dept Aerosp Informat Engn, Sch Astronaut, Beijing 102206, Peoples R China
[3] Guide Infrared, Wuhan 430205, Peoples R China
来源
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING | 2024年 / 62卷
关键词
Training; Object detection; Convolutional neural networks; Shape; Noise; Image edge detection; Geoscience and remote sensing; Detection algorithms; Convergence; Deep learning; Convolutional neural networks (CNNs); infrared small target; point-supervision; target detection;
D O I
10.1109/TGRS.2024.3472455
中图分类号
P3 [地球物理学]; P59 [地球化学];
学科分类号
0708 ; 070902 ;
摘要
In recent years, the implementation of infrared small-target detection using convolutional neural networks (CNNs) has garnered widespread attention due to its high performance. Since this issue is often addressed through fully supervised image segmentation networks, the training process requires significant human effort and time to annotate pixel-level mask labels. Consequently, employing single-point supervision as a form of weak supervision for model training has aroused widespread interest in saving annotation costs. However, the class imbalance issue caused by single-point supervision in the early stages of training, along with inaccuracies in pseudo-label updates, and the difficulty in achieving convergence simultaneously, have made it challenging for such methods to attain satisfactory performance. In this article, we introduce a label evolution framework based on local contrast measure (LELCM) to address these issues. Before training, we expand the single-point labels into initial pseudo-labels based on the inherent information of the targets, which mitigates the problem of class imbalance. Furthermore, in the process of updating pseudo-labels, we employ a strategy that utilizes confidence contrast for updates, not only enabling more stable updates of pseudo-labels based on target characteristics but also facilitating adaptive cessation of updates. Our experimental results reveal that our approach not only attains target detection rates (Pd) on par with full supervision models but also achieves 80% of the full supervisory effect in terms of intersection over union (IoU).
引用
收藏
页数:12
相关论文
共 50 条
[1]   Finding Berries: Segmentation and Counting of Cranberries using Point Supervision and Shape Priors [J].
Akiva, Peri ;
Dana, Kristin ;
Oudemans, Peter ;
Mars, Michael .
2020 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW 2020), 2020, :219-228
[2]   What's the Point: Semantic Segmentation with Point Supervision [J].
Bearman, Amy ;
Russakovsky, Olga ;
Ferrari, Vittorio ;
Fei-Fei, Li .
COMPUTER VISION - ECCV 2016, PT VII, 2016, 9911 :549-565
[3]   End-to-End Object Detection with Transformers [J].
Carion, Nicolas ;
Massa, Francisco ;
Synnaeve, Gabriel ;
Usunier, Nicolas ;
Kirillov, Alexander ;
Zagoruyko, Sergey .
COMPUTER VISION - ECCV 2020, PT I, 2020, 12346 :213-229
[4]   A Local Contrast Method for Small Infrared Target Detection [J].
Chen, C. L. Philip ;
Li, Hong ;
Wei, Yantao ;
Xia, Tian ;
Tang, Yuan Yan .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2014, 52 (01) :574-581
[5]   Points as Queries: Weakly Semi-supervised Object Detection by Points [J].
Chen, Liangyu ;
Yang, Tong ;
Zhang, Xiangyu ;
Zhang, Wei ;
Sun, Jian .
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, :8819-8828
[6]   A Multi-Task Framework for Infrared Small Target Detection and Segmentation [J].
Chen, Yuhang ;
Li, Liyuan ;
Liu, Xin ;
Su, Xiaofeng .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2022, 60
[7]   Pointly-Supervised Instance Segmentation [J].
Cheng, Bowen ;
Parkhi, Omkar ;
Kirillov, Alexander .
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022), 2022, :2607-2616
[8]   DBGNet: Dual-Branch Gate-Aware Network for Infrared Small Target Detection [J].
Chi, Weijian ;
Liu, Jiahang ;
Wang, Xiaozhen ;
Feng, Ruilei ;
Cui, Jian .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2023, 61
[9]   Attentional Local Contrast Networks for Infrared Small Target Detection [J].
Dai, Yimian ;
Wu, Yiquan ;
Zhou, Fei ;
Barnard, Kobus .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2021, 59 (11) :9813-9824
[10]   Asymmetric Contextual Modulation for Infrared Small Target Detection [J].
Dai, Yimian ;
Wu, Yiquan ;
Zhou, Fei ;
Barnard, Kobus .
2021 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2021), 2021, :949-958