频域衰减卷积的图像分类网络

doi:10.16451/j.cnki.issn1003-6059.202605003

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (3719 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要针对传统卷积神经网络中卷积核频域响应因缺乏显式约束而导致的高频分量过度强化的问题,提出频域衰减卷积的图像分类网络(Image Classification Network with Frequency Decay Convolution, FDNet).首先,设计频域衰减卷积模块(Frequency Decay Convolution, FDConv),将神经元信号衰减机制引入频域空间,通过对卷积核频域幅度实施衰减调控,在保留完整相位信息的前提下抑制过强频率分量的响应,有效缓解因频域响应失衡引发的过拟合问题.然后,将FDConv嵌入网络浅层特征提取阶段和残差块结构中,使频谱正则化贯穿整个特征提取过程.在此基础上,构建空间方向衰减模块(Directional Spatial Decay, DSD),通过并行的1×3 FDConv和3×1 FDConv对水平方向与垂直方向的特征进行分离提取,实现对方向特征与重要通道的协同增强.最后,将DSD嵌入残差块末端,在残差连接后对特征进行方向分解与通道重标定,使网络聚焦关键判别信息.在CIFAR-10、CIFAR-100、SVHN、STL-10、Ima-genette、Imagewoof图像数据集上的实验表明,FDNet分类性能较优.

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	袁姮
	禚霁坤
	张晟翀

关键词 ：图像分类, 神经元信号衰减, 频域衰减卷积, 空间方向衰减, 残差网络

Abstract：In traditional convolutional neural networks,the lack of explicit constraints on the frequency-domain responses of convolutional kernels leads to excessive enhancement of high-frequency components. To address this issue, an image classification network with frequency decay convolution(FDNet) is proposed. First, the frequency decay convolution(FDConv) module is designed. The neuron signal attenuation mechanism is introduced into the frequency domain. By attenuating the frequency-domain amplitude of the convolutional kernel, the responses of excessively strong frequency components are suppressed with complete phase information retained. Consequently, the overfitting problem caused by imbalance in frequency-domain responses is effectively alleviated. Second, FDConv is embedded into the shallow feature extraction stage and the residual block structure, enabling spectral regularization throughout the entire feature extraction process. Then, the directional spatial decay(DSD) module is constructed. Parallel 1×3 and 3×1 FDConv modules are adopted to separately extract horizontal and vertical directional features, thereby achieving collaborative enhancement of directional features and important channels. Finally, the DSD module is embedded at the end of the residual block. After the residual connection, both directional feature decomposition and channel recalibration are performed, enabling the network to focus on key discriminative information. Experiments on CIFAR-10, CIFAR-100, SVHN, STL-10, Imagenette, and Imagewoof image datasets demonstrate the superior classification performance of FDNet.

Key words： Image Classification Neuron Signal Attenuation Frequency Decay Convolution Directional Spatial Decay Residual Network

收稿日期: 2026-03-31

ZTFLH:

TP391

基金资助:国家自然科学基金项目(No.61601213)、辽宁省自然科学基金项目(No.20170540426)、辽宁省教育厅重点基金项目(No.LJYL049)资助

通讯作者: 袁姮,博士,副教授,主要研究方向为图像处理、模式识别、人工智能.E-mail:lntuyuanheng@163.com.

作者简介: 禚霁坤,硕士研究生,主要研究方向为图像处理、模式识别、人工智能.E-mail:zhuojikun@163.com.
张晟翀,硕士,高级工程师,主要研究方向为数字信号处理、模式识别、人工智能.E-mail:zsc417@126.com.

引用本文:

袁姮, 禚霁坤, 张晟翀. 频域衰减卷积的图像分类网络[J]. 模式识别与人工智能, 2026, 39(5): 415-435. YUAN Heng, ZHUO Jikun, ZHANG Shengchong. Image Classification Network with Frequency Decay Convolution. Pattern Recognition and Artificial Intelligence, 2026, 39(5): 415-435.

链接本文:

http://manu46.magtech.com.cn/Jweb_prai/CN/10.16451/j.cnki.issn1003-6059.202605003 或 http://manu46.magtech.com.cn/Jweb_prai/CN/Y2026/V39/I5/415

[1] 刘鸿达,孙旭辉,李沂滨,等.基于卷积神经网络的图像分类深度学习模型综述.计算机工程与应用, 2025, 61(11): 1-21.
(LIU H D, SUN X H, LI Y B, et al. Review of Deep Learning Models for Image Classification Based on Convolutional Neural Networks. Computer Engineering and Applications, 2025, 61(11): 1-21.)
[2] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-Based Lear-ning Applied to Document Recognition. Proceedings of the IEEE, 1998, 86(11): 2278-2324.
[3] KRIZHEVSKY A, SUTSKEVER I, HINTON G E.ImageNet Classification with Deep Convolutional Neural Networks. Communications of the ACM, 2017, 60(6): 84-90.
[4] SIMONYAN K, ZISSERMAN A.Very Deep Convolutional Networks for Large-Scale Image Recognition[C/OL].[2026-02-27]. https://arxiv.org/pdf/1409.1556.
[5] SZEGEDY C, LIU W, JIA Y Q, et al. Going Deeper with Convolutions // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2015: 1-9.
[6] HE K M, ZHANG X Y, REN S Q, et al. Deep Residual Learning for Image Recognition // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2016: 770-778.
[7] ZAGORUYKO S, KOMODAKIS N. Wide Residual Networks[C/OL].[2026-02-27]. https://arxiv.org/pdf/1605.07146.
[8] HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely Co-nnected Convolutional Networks // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2017: 2261-2269.
[9] HOWARD A G, ZHU M L, CHEN B, et al. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications[C/OL].[2026-02-27]. https://arxiv.org/pdf/1704.04861.
[10] ZHAO C Q, DONG G F, ZHANG S P, et al. Frequency Regularization: Reducing Information Redundancy in Convolutional Neural Networks. IEEE Access, 2023, 11: 106793-106802.
[11] ZHANG H, WU C R, ZHANG Z Y, et al. ResNeSt: Split-Attention Networks // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2022: 2735-2745.
[12] QIN Z Q, ZHANG P Y, WU F, et al. FcaNet: Frequency Cha-nnel Attention Networks // Proc of the IEEE/CVF International Conference on Computer Vision. Washington, USA: IEEE, 2021: 763-772.
[13] LEE J H.HiFAR: High-Frequency-Aware Regularization for Domain Generalization in Convolutional Neural Networks. Master Di-ssertation. Pohang, Korea: Pohang University of Science and Technology, 2025.
[14] LU J H.Frequency Regularization: Unveiling the Spectral Inductive Bias of Deep Neural Networks[C/OL].[2026-02-27].https://arxiv.org/pdf/2512.22192v1.
[15] MIYATO T, KATAOKA T, KOYAMA M, et al. Spectral Normalization for Generative Adversarial Networks[C/OL].[2026-02-27]. https://arxiv.org/pdf/1802.05957.
[16] HU J, SHEN L, SUN G.Squeeze-and-Excitation Networks // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2018: 7132-7141.
[17] WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional Block Attention Module // Proc of the 15th European Conference on Computer Vision. Berlin, Germany: Springer, 2018: 3-19.
[18] WANG Q L, WU B G, ZHU P F, et al. ECA-Net: Efficient Cha-nnel Attention for Deep Convolutional Neural Networks // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2020: 11531-11539.
[19] JIANG W T, YUAN H, LIU W J.Neuron Signal Attenuation Activation Mechanism for Deep Learning. Patterns, 2025, 6(1). DOI: 10.1016/j.patter.2024.101117.
[20] CHOROMANSKI K, LIKHOSHERSTOV V, DOHAN D, et al. Rethinking Attention with Performers[C/OL].[2026-02-27]. https://arxiv.org/pdf/2009.14794.
[21] HAN K, WANG Y H, TIAN Q, et al. GhostNet: More Features from Cheap Operations // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2020: 1577-1586.
[22] LAN H, WANG X H, SHEN H, et al. Couplformer: Rethinking Vision Transformer with Coupling Attention Map // Proc of the IEEE/CVF Winter Conference on Applications of Computer Vision. Washington, USA: IEEE, 2023: 6464-6473.
[23] HASSANI A, WALTON S, SHAH N, et al. Escaping the Big Data Paradigm with Compact Transformers[C/OL].[2026-02-27]. https://arxiv.org/pdf/2104.05704.
[24] 姜文涛,赵琳琳,涂潮.双分支多注意力机制的锐度感知分类网络.模式识别与人工智能, 2023, 36(3): 252-267.
(JIANG W T, ZHAO L L, TU C.Double-Branch Multi-attention Mechanism Based Sharpness-Aware Classification Network. Pattern Recognition and Artificial Intelligence, 2023, 36(3): 252-267.)
[25] LUO Z B, SUN Z T, ZHOU W L, et al. Rethinking ResNets: Improved Stacking Strategies with High-Order Schemes for Image Classification. Complex & Intelligent Systems, 2022, 8(4): 3395-3407.
[26] TSENG C H, LEE S J, FENG J N, et al. UPANets: Learning from the Universal Pixel Attention Networks. Entropy, 2022, 24(9). DOI: 10.3390/e24091243.
[27] TAN M X, LE Q V.EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks // Proc of the 36th International Conference on Machine Learning. San Diego, USA: JMLR, 2019: 6105-6114.
[28] 袁姮,于东琪,高原.面向图像分类的双域特征联合网络.模式识别与人工智能, 2025, 38(4): 325-340.
(YUAN H, YU D Q, GAO Y.Two-Domain Feature Association Networks for Image Classification. Pattern Recognition and Artificial Intelligence, 2025, 38(4): 325-340.)
[29] WU X D, GAO S Q, ZHANG Z Y, et al. Auto-Train-Once: Controller Network Guided Automatic Network Pruning from Scratch // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2024: 16163-16173.
[30] SHIN H, CHOI D W.Teacher as a Lenient Expert: Teacher-Agnostic Data-Free Knowledge Distillation. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(13): 14991-14999.
[31] 姜文涛,王鑫杰,张晟翀.空间约束注意力机制的图像分类网络.智能系统学报, 2025, 20(6): 1444-1460.
(JIANG W T, WANG X J, ZHANG S C.Spatially Constrained Attention Mechanism for Image Classification Network. CAAI Transactions on Intelligent Systems, 2025, 20(6): 1444-1460.)
[32] ZHOU C L, ZHANG H, ZHOU Z K, et al. QKFormer: Hierarchical Spiking Transformer Using Q-K Attention // Proc of the 38th International Conference on Neural Information Processing Systems. Cambridge, USA: MIT Press, 2024: 13074-13098.
[33] MA C X, WU J B, SI C Y, et al. Scaling Supervised Local Lear-ning with Augmented Auxiliary Networks[C/OL].[2026-02-27]. https://arxiv.org/pdf/2402.17318.
[34] ABDI M, NAHAVANDI S.Multi-residual Networks: Improving the Speed and Accuracy of Residual Networks[C/OL].[2026-02-27]. https://arxiv.org/pdf/1609.05672.
[35] 袁姮,霍欣燃,姜文涛.全息梯度差分卷积的图像分类网络.自动化学报, 2025, 51(9): 2106-2130.
(YUAN H, HUO X R, JIANG W T.Image Classification Network of Holographic Gradient Differential Convolution. Acta Automatica Sinica, 2025, 51(9): 2106-2130.)
[36] GUO M H, LU C Z, LIU Z N, et al. Visual Attention Network. Computational Visual Media, 2023, 9(4): 733-752.