融入多尺度区域注意力的小样本遥感场景分类

doi:10.3778/j.issn.1002-8331.2404-0337

摘要/Abstract

摘要： 针对基于小样本的研究方法大都关注图像深层特征的提取、对于浅层信息的挖掘不充分，且忽视了遥感图像区域特征重要性的问题，提出一种融入多尺度区域注意力的小样本遥感场景分类方法。为了提取遥感场景包含细粒度及粗粒度信息的多尺度特征，提出了基于ResNet-12的多尺度特征提取模块。由于在样本少的情况下，区域特征在场景分类中起着关键作用，为强调区域特征的重要性，提出区域注意力（patch attention，PAT）机制，构建基于PAT的特征增强模块。为充分利用多尺度特征信息，提出特征融合和分类模块，将来自不同尺度的增强特征进行融合，再通过计算样本间的余弦相似度完成分类任务。实验结果表明，提出的方法能够有效地提升分类准确率和对于新类别的识别能力。

关键词: 小样本学习, 遥感场景分类, 多尺度, 区域注意力

Abstract: In view of the problem that most research methods based on few-shot focus on the extraction of deep features of images, insufficient mining of shallow information, and ignore the importance of regional features in remote sensing images, a few-shot scene classification with multi-scale patch attention mechanism in remote sensing is proposed. In order to extract multi-scale features containing fine-grained and coarse-grained information in remote sensing scenes, a multi-scale feature extraction module based on ResNet-12 is proposed. Since regional features play a key role in scene classifi-cation when there are few samples, in order to emphasize the importance of regional features, a patch attention (PAT) mechanism is proposed to construct a feature enhancement module based on PAT. Finally, to make full use of multi-scale feature information, a feature fusion and classification module is proposed to fuse enhanced features from different scales, and then complete the classification task by calculating the cosine similarity between samples. Experimental results show that the proposed method can effectively improve the classification accuracy and the recognition ability for new categories.

Key words: few-shot learning, remote sensing scene classification, muti-scale, patch attention

周立俭, 赵志昂, 孟庆宇, 王梦圆, 郝思媛, 赵锟. 融入多尺度区域注意力的小样本遥感场景分类[J]. 计算机工程与应用, 2025, 61(15): 278-287.

ZHOU Lijian, ZHAO Zhi’ang, MENG Qingyu, WANG Mengyuan, HAO Siyuan, ZHAO Kun. Few-Shot Scene Classification with Multi-Scale Patch Attention Mechanism in Remote Sensing[J]. Computer Engineering and Applications, 2025, 61(15): 278-287.

参考文献

[1] YAO X W, FENG X X, HAN J W, et al. Automatic weakly supervised object detection from high spatial resolution remote sensing images via dynamic curriculum learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(1): 675-685.
[2] CHEN F H, TSOU J Y. Assessing the effects of convolutional neural network architectural factors on model performance for remote sensing image classification: an in-depth investigation[J]. International Journal of Applied Earth Observation and Geoinformation, 2022, 112: 102865.
[3] PENATTI O A B, NOGUEIRA K, DOS SANTOS J A. Do deep features generalize from everyday objects to remote sensing and aerial scenes domains? [C]//Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops. Piscataway: IEEE, 2015: 44-51.
[4] ZHANG Y, ZHENG X T, LU X Q. Pairwise comparison network for remote-sensing scene classification[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19: 6505105.
[5] WANG Q, HUANG W, XIONG Z T, et al. Looking closer at the scene: multiscale representation learning for remote sensing image scene classification[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 33(4): 1414-1428.
[6] LI L J, HAN J W, YAO X W, et al. DLA-MatchNet for few-shot remote sensing image scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(9): 7844-7853.
[7] CHEN Y Q, LI Y Y, MAO H T, et al. A novel deep nearest neighbor neural network for few-shot remote sensing image scene classification[J]. Remote Sensing, 2023, 15(3): 666.
[8] 张多纳, 赵宏佳, 鲁远耀, 等. 融入注意力机制的小样本遥感图像场景分类[J]. 计算机工程与应用, 2024, 60(4): 173-182.
ZHANG D N, ZHAO H J, LU Y Y, et al. Few-shot scene classification with attention mechanism in remote sensing[J]. Computer Engineering and Applications, 2024, 60(4): 173-182.
[9] WANG J Y, WANG X Q, XING L, et al. Class-shared SparsePCA for few-shot remote sensing scene classification[J]. Remote Sensing, 2022, 14(10): 2304.
[10] YUAN Z W, TANG C, YANG A X, et al. Few-shot remote sensing image scene classification based on metric learning and local descriptors[J]. Remote Sensing, 2023, 15(3): 831.
[11] HUANG W D, YUAN Z W, YANG A X, et al. TAE-net: task-adaptive embedding network for few-shot remote sensing scene classification[J]. Remote Sensing, 2022, 14(1): 111.
[12] 陈杰虎, 汪西莉. 多图卷积网络的遥感图像小样本分类[J]. 遥感学报, 2022, 26(10): 2029-2042.
CHEN J H, WANG X L. Multi-graph convolutional network for a remote sensing image few shot classification[J]. National Remote Sensing Bulletin, 2022, 26(10): 2029-2042.
[13] ZHANG B Q, FENG S S, LI X T, et al. SGMNet: scene graph matching network for few-shot remote sensing scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5628915.
[14] QIAO X J, XING L, HAN A X, et al. Simplified multi-head mechanism for few-shot remote sensing image classification[J]. Neural Processing Letters, 2024, 56(1): 2.
[15] MA Y T, MENG J M, LIU B D, et al. Dictionary learning for few-shot remote sensing scene classification[J]. Remote Sensing, 2023, 15(3): 773.
[16] DONG Z, LIN B J, XIE F. Optimizing few-shot remote sensing scene classification based on an improved data augmentation approach[J]. Remote Sensing, 2024, 16(3): 525.
[17] WANG Y K, XU C M, LIU C, et al. Instance credibility inference for few-shot learning[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 12833-12842.
[18] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 770-778.
[19] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7132-7141.
[20] CHENG G, HAN J W, LU X Q. Remote sensing image scene classification: benchmark and state of the art[J]. Proceedings of the IEEE, 2017, 105(10): 1865-1883.
[21] YANG Y, NEWSAM S. Bag-of-visual-words and spatial extensions for land-use classification[C]//Proceedings of the 18th SIGSPATIAL International Conference on Advances in Geographic Information Systems. New York: ACM, 2010: 270-279.
[22] WANG X L, GIRSHICK R, GUPTA A, et al. Non-local neural networks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7794-7803.
[23] FINN C, ABBEEL P, LEVINE S. Model-agnostic meta-learning for fast adaptation of deep networks[C]//Proceedings of the 34th International Conference on Machine Learning, 2017: 1126-1135.
[24] SNELL J, SWERSKY K, ZEMEL R S. Prototypical networks for few-shot learning[C]//Advances in Neural Information Processing Systems, 2017.
[25] VINYALS O, BLUNDELL C, LILLICRAP T, et al. Matching networks for one shot learning[C]//Advances in Neural Information Processing Systems, 2016.
[26] CHENG G, CAI L M, LANG C B, et al. SPNet: siamese-prototype network for few-shot remote sensing image scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 5608011.
[27] WANG B, WANG Z R, SUN X, et al. TDNet: a novel transductive learning framework with conditional metric embedding for few-shot remote sensing image scene classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 16: 4591-4606.
[28] CHEN X L, ZHU G B, LIU M Q, et al. Few-shot remote sensing image scene classification based on multiscale covariance metric network (MCMNet)[J]. Neural Networks, 2023, 163: 132-145.