[1]潘丽丽,瞿栋梁,尹晶晶,等.基于交叉量化和样本校正的自监督遥感图像检索[J].郑州大学学报(工学版),2025,46(02):60-66.[doi:10.13705/j.issn.1671-6833.2024.06.007]
　PAN Lili,QU Dongliang,YIN Jingjing,et al.Self-supervised Remote Sensing Image Retrieval Based on Cross-quantization and Sample Correction[J].Journal of Zhengzhou University (Engineering Science),2025,46(02):60-66.[doi:10.13705/j.issn.1671-6833.2024.06.007]
点击复制

基于交叉量化和样本校正的自监督遥感图像检索()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
46
期数:
2025年02期
页码:
60-66
栏目:
出版日期:
2025-03-10

文章信息/Info

Title:
Self-supervised Remote Sensing Image Retrieval Based on Cross-quantization and Sample Correction
文章编号:
1671-6833(2025)02-0060-07
作者:
潘丽丽1;  瞿栋梁2;  尹晶晶2;  马雪强1
1.中南林业科技大学 计算机与数学学院,湖南 长沙 410000;2.中南林业科技大学 电子信息与物理学院,湖南 长沙 410000
Author(s):
PAN Lili1;  QU Dongliang2;  YIN Jingjing2;  MA Xueqiang1
1.School of Computer Science and Mathematics, Central South University of Forestry and Technology, Changsha 410000, China; 2. School of Electronic Information and Physics, Central South University of Forestry and Technology, Changsha 410000, China
关键词:
遥感图像检索;  对比学习;  乘积量化;  交叉学习
Keywords:
remote sensing image retrieval;  contrastive learning;  product quantization;  cross-learning
分类号:
P407.8TP183
DOI:
10.13705/j.issn.1671-6833.2024.06.007
文献标志码:
A
摘要:
自监督遥感图像检索模型由于标签缺失导致使用错误的样本对进行学习,从而产生抽样偏差,影响图像表征的准确性,为此提出一种基于交叉量化和样本校正的自监督遥感图像检索模型(CQSC)。首先,为了降低数据存储和处理负载,将传统对比学习中的映射层与乘积量化相结合,压缩高维图像数据,提高了检索效率;其次,使用交叉学习策略,最大化检索模型中特征映射前后的交叉相似性,增强模型的特征生成能力和检索精度;最后,针对自监督遥感图像检索中因标签缺失导致的抽样偏差,设计自适应纠错标签,标注训练样本,校正训练过程中存在的错误负样本。在UCMerced和EuroSAT数据集上进行了大量实验,结果表明:与PLSH方法相比,在UCMerced数据集上,所提方法mAP@20平均提升了2.52百分点;在EuroSAT数据集上取64 bits时,所提方法mAP@100提升了3.83百分点。
Abstract:
In view of the fact that the self-supervised remote sensing image retrieval model by using the sample pair for learning due to the lack of labels, resulted in sampling bias and affecting the accuracy of image representation, a self-supervised remote sensing image retrieval model based on cross-quantization and sample correction (CQSC) was proposed. Firstly, in order to reduce the load of data storage and processing, the mapping layer and product quantization in traditional contrast learning were combined to compress the high-dimensional image data and improve the retrieval efficiency. Secondly, the cross-learning strategy was used to maximize the cross-similarity before and after the feature mapping in the retrieval model, and the feature generation ability and retrieval accuracy of the model were enhanced. Finally, design adaptive correction labels to annotate training samples, correct erroneous negative samples during training and address sampling bias caused by missing labels in self-supervised remote sensing image retrieval. Experiments on UCMerced and EuroSAT datasets showed that compared with PLSH, mAP@20 of CQSC improved by 2.52 percentage points on average on UCMerced, and mAP@100 of CQSC improved by 3.83 percentage points on EuroSAT with 64bits.

参考文献/References:

[1]张艺琨, 唐雁, 陈强. 基于多特征融合的三维模型检索[J]. 郑州大学学报(工学版), 2019, 40(1): 1-6. 

ZHANG Y K, TANG Y, CHEN Q. 3D model retrieval method based on multiple feature fusion[J]. Journal of Zhengzhou University (Engineering Science), 2019, 40 (1): 1-6. 
[2]冯孝鑫, 王子健, 吴奇. 基于三元采样图卷积网络的半监督遥感图像检索[J]. 电子与信息学报, 2023, 45 (2): 644-653. 
FENG X X, WANG Z J, WU Q. Semi-supervised learning remote sensing image retrieval method based on triplet sampling graph convolutional network[J]. Journal of Electronics & Information Technology, 2023, 45(2): 644653. 
[3]JING W P, XU Z K, LI L H, et al. Deep unsupervised weighted Hashing for remote sensing image retrieval[J]. Journal of Database Management, 2022, 33(2): 1-19. 
[4]FENG C, PATRAS I. Adaptive soft contrastive learning [C]∥The 26th International Conference on Pattern Recognition (ICPR). Piscataway: IEEE, 2022: 27212727. 
[5]CHUANG C Y, ROBINSON J, YEN-CHEN L, et al. Debiased contrastive learning[C]∥Proceedings of the 34th International Conference on Neural Information Processing Systems. New York: ACM, 2020: 8765-8775. 
[6]AI L F, TAO Y, CHENG H J, et al. Accumulative quantization for approximate nearest neighbor search[J]. Computational Intelligence and Neuroscience, 2022, 2022: 4364252. 
[7]SUN Y X, YE Y M, LI X T, et al. Unsupervised deep Hashing through learning soft pseudo label for remote sensing image retrieval[J]. Knowledge-Based Systems, 2022, 239: 107807. 
[8]CHEN B, FENG Y, DAI T, et al. Adversarial examples generation for deep product quantization networks on image retrieval[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(2): 1388-1404. 
[9]ZHOU H Y, QIN Q B, HOU J K, et al. Deep global semantic structure-preserving Hashing via corrective triplet loss for remote sensing image retrieval[J]. Expert Systems with Applications, 2024, 238: 122105. 
[10] LIANG Y, ZHANG S L, LI K L, et al. Unleashing the full potential of product quantization for large-scale image retrieval[EB/OL]. (2023-10-09)[2024-08-22]. https:∥paperswithcode. com/paper/unleashing-the-fullpotential-of-product.
[11] TANG J Y, WANG D L, TONG X C, et al. Learning from self-supervised features for Hashing-based remote sensing image retrieval[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20: 2501705. 
[12] HUYNH T, KORNBLITH S, WALTER M R, et al. Boosting contrastive self-supervised learning with false negative cancellation[C]∥2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). Piscataway: IEEE, 2022: 986-996. 
[13] DWIBEDI D, AYTAR Y, TOMPSON J, et al. With a little help from my friends: nearest-neighbor contrastive learning of visual representations[EB/OL].(2021-1007)[2024-08-22]. http:∥arxiv.org/abs/2104.14548. 
[14]何悦, 陈广胜, 景维鹏, 等. 基于深度多相似性哈希方法的遥感图像检索[J]. 计算机工程, 2023, 49 (2): 206-212.
HE Y, CHEN G S, JING W P, et al. Remote sensing image retrieval based on deep multi-similarity Hashing method[J]. Computer Engineering, 2023, 49(2): 206212. 
[15]吴刚, 葛芸, 储珺, 等. 面向遥感图像检索的级联池化自注意力研究[J]. 光电工程, 2022, 49 (12): 220029. 
WU G, GE Y, CHU J, et al. Cascade pooling self-attention research for remote sensing image retrieval[J]. Opto-Electronic Engineering, 2022, 49(12): 220029. 
[16] DEMIR B, BRUZZONE L. Hashing-based scalable remote sensing image search and retrieval in large archives [J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(2): 892-904. 
[17] FERNANDEZ-BELTRAN R, DEMIR B, PLA F, et al. Unsupervised remote sensing image retrieval using probabilistic latent semantic Hashing[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 18(2): 256-260. 
[18] LI P, REN P. Partial randomness hashing for large-scale remote sensing image retrieval[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(3): 464-468. 
[19] SLANEY M, CASEY M. Locality-sensitive Hashing for finding nearest neighbors lecture notes[J]. IEEE Signal Processing Magazine, 2008, 25(2): 128-131. 
[20] GONG Y C, LAZEBNIK S, GORDO A, et al. Iterative quantization: a procrustean approach to learning binary codes for large-scale image retrieval[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(12): 2916-2929. 
[21] SHEN Y M, QIN J, CHEN J X, et al. Auto-encoding twin-bottleneck Hashing[C]∥2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2020: 2815-2824.

更新日期/Last Update: 2025-03-13