Entropy-Based re-sampling method on SAR class imbalance target detection

被引:1
作者
Zhang, Chong-Qi [1 ]
Deng, Yao [1 ]
Chong, Ming-Zhe [1 ]
Zhang, Zi-Wen [1 ]
Tan, Yun-Hua [1 ]
机构
[1] Peking Univ, Sch Elect, Beijing 100871, Peoples R China
基金
中国国家自然科学基金;
关键词
SAR target detection; Class imbalance problem; Re-sampling strategy; Variance-weighted information entropy; Count smoothing; IMAGES; NETWORK;
D O I
10.1016/j.isprsjprs.2024.02.019
中图分类号
P9 [自然地理学];
学科分类号
0705 ; 070501 ;
摘要
Detection tasks based on Synthetic aperture radar (SAR) images have been studied widely but severely constrained by the quality of datasets. Meanwhile, both the unperceived category imbalance problem and SAR image discrepancy of multi-class SAR datasets are not fully considered. Researchers usually care about the foregroundbackground imbalance more than the class imbalance for SAR images. To solve these problems, an entropy-based re-sampling method is proposed in which category imbalance and quality discrepancies of SAR images are both considered. Initially, the relationship is established between variance-weighted image entropy and factors affecting SAR image quality, such as noise, resolution, and density, thereby validating entropy as a robust metric for image quality assessment. Subsequently, quantity penalty scores of categories and difficulty penalty scores of each image are calculated separately to capture the inter-class and intra-class disparities. Next, logarithmic smoothing is employed to avoid overestimation of image difficulty due to the margin effect. Finally, all these scores are combined to generate normalized scores representing the final distribution of the dataset to guide the training process. The proposed approach serves as a plug-and-play strategy for general SAR detection tasks, and experimental results indicate a significant performance improvement. Specifically, detection accuracy in terms of AP for airplanes and bridges (minority classes) in the MSAR dataset is improved by 16.6 % and 13.0 %, respectively, compared to the YOLOv5 baseline, with only a minimal 1.6 % sacrifice in ship detection (majority class). The datasets and the codes can be found at https://www.radars.ac.cn/web/data/getData?dataType=MSAR, https://github.com/Phoenix0qi/Yolov5-entropy-balance/tree/master.
引用
收藏
页码:432 / 447
页数:16
相关论文
共 62 条
[1]   Multiplicative Noise Removal Using Variable Splitting and Constrained Optimization [J].
Bioucas-Dias, Jose M. ;
Figueiredo, Mario A. T. .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2010, 19 (07) :1720-1730
[2]  
Cao KD, 2019, ADV NEUR IN, V32
[3]   Fast target detection method for high-resolution SAR images based on variance weighted information entropy [J].
Cao, Zongjie ;
Ge, Yuchen ;
Feng, Jilan .
EURASIP JOURNAL ON ADVANCES IN SIGNAL PROCESSING, 2014,
[4]   Entropy-Based Non-Local Means Filter for Single-Look SAR Speckle Reduction [J].
Chan, Debora ;
Gambini, Juliana ;
Frery, Alejandro C. .
REMOTE SENSING, 2022, 14 (03)
[5]  
Chierchia G, 2017, INT GEOSCI REMOTE SE, P5438, DOI 10.1109/IGARSS.2017.8128234
[6]   Large Scale Fine-Grained Categorization and Domain-Specific Transfer Learning [J].
Cui, Yin ;
Song, Yang ;
Sun, Chen ;
Howard, Andrew ;
Belongie, Serge .
2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2018, :4109-4118
[7]   Image denoising with block-matching and 3D filtering [J].
Dabov, Kostadin ;
Foi, Alessandro ;
Katkovnik, Vladimir ;
Egiazarian, Karen .
IMAGE PROCESSING: ALGORITHMS AND SYSTEMS, NEURAL NETWORKS, AND MACHINE LEARNING, 2006, 6064
[8]   Learning Deep Ship Detector in SAR Images From Scratch [J].
Deng, Zhipeng ;
Sun, Hao ;
Zhou, Shilin ;
Zhao, Juanping .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2019, 57 (06) :4021-4039
[9]   Target detection in synthetic aperture radar imagery: a state-of-the-art survey [J].
El-Darymli, Khalid ;
McGuire, Peter ;
Power, Desmond ;
Moloneyb, Cecilia .
JOURNAL OF APPLIED REMOTE SENSING, 2013, 7
[10]   Exploring Classification Equilibrium in Long-Tailed Object Detection [J].
Feng, Chengjian ;
Zhong, Yujie ;
Huang, Weilin .
2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021), 2021, :3397-3406