Dual-coding space-frequency mixing method for infrared small target detection

doi:10.11772/j.issn.1001-9081.2025010078

Abstract

Abstract:

InfraRed Small Target Detection （IRSTD） aims to find targets from infrared images with low Signal-to-Clutter Ratio （SCR） accurately， and has been used in many fields widely. However， due to the weak target features and severe background interference， the existing methods struggle to extract structural information of the target effectively， which leads to issues such as incomplete target segmentation and low detection accuracy. Moreover， these models usually have a large number of parameters. To solve the above problems， a dual-coding space-frequency mixing IRSTD method was proposed. Firstly， with using U-Net3+ as the basic framework， a dual-coding structure combining Multi-Shape Context Aware （MSCA） module and Frequency-Domain Interactive Attention （FDIA） module was proposed to extract space-frequency mixing features in the coding stage. Secondly， in the decoding stage， a Cross-Layer Feature Guide （CLFG） module was designed to fuse multi-scale feature maps. Finally， the proposed method was experimentally verified on the NUAA-SIRST and IRSTD-1k datasets. The results show that the proposed method has the number of parameters of 0.86×10⁶， and the Intersection over Union （IoU） reached 78.11% and 69.08% respectively. Compared with Attention Multiscale feature Fusion U-Net （AMFUNet） method， the number of parameters is reduced by 1.31×10⁶， and the IoUs are increased by 2.25 percentage points and 1.23 percentage points respectively. It can be seen that the proposed method has high detection performance while retaining fewer parameters.

Key words: deep learning, Infrared Small Target Detection (IRSTD), dual-coding, space-frequency mixing, cross-layer guidance

摘要：

红外小目标检测（IRSTD）旨在从低信杂比（SCR）的红外图像中精准找到目标，在多个领域获得了非常广泛的应用。然而，现有方法因目标特征微弱和背景干扰严重，难以有效提取目标结构性信息，进而导致目标分割不完整和检测精度低等问题，并且模型参数量较大。为了解决以上问题，提出双编码空频混合的IRSTD方法。首先，采用U-Net3+作为基本框架，并在编码阶段提出一种多形状上下文感知（MSCA）模块和频域交互注意力（FDIA）模块相结合的双编码结构来提取空频混合特征；其次，在解码阶段设计跨层特征引导（CLFG）模块来融合多尺度下的特征图；最后，所提方法分别在NUAA-SIRST和IRSTD-1k数据集上进行了实验验证。结果表明，所提方法的参数量为0.86×10⁶，交并比（IoU）分别达到了78.11%和69.08%，与注意力多尺度特征融合U型网络（AMFUNet）方法相比，参数量减少了1.31×10⁶， IoU分别提升了2.25和1.23个百分点。可见，所提方法在保留较少参数量的同时具有较高的检测性能。

关键词: 深度学习, 红外小目标检测, 双编码, 空频混合, 跨层引导

CLC Number:

TP391.4

Xiaoyong BIAN, Peiyang YUAN, Qiren HU. Dual-coding space-frequency mixing method for infrared small target detection[J]. Journal of Computer Applications, 2026, 46(1): 252-259.

边小勇, 袁培洋, 胡其仁. 双编码空频混合的红外小目标检测方法[J]. 《计算机应用》唯一官方网站, 2026, 46(1): 252-259.

Figures/Tables 12

References 23

[1]	BAI X， ZHOU F. Analysis of new top-hat transformation and the application for infrared dim small target detection ［J］. Pattern Recognition， 2010， 43（6）： 2145-2156.
[2]	CHEN C L P， LI H， WEI Y， et al. A local contrast method for small infrared target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2014， 52（1）： 574-581.
[3]	GAO C， MENG D， YANG Y， et al. Infrared patch-image model for small target detection in a single image ［J］. IEEE Transactions on Image Processing， 2013， 22（12）： 4996-5009.
[4]	DAI Y， WU Y. Reweighted infrared patch-tensor model with both nonlocal and local priors for single-frame small target detection ［J］. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing， 2017， 10（8）： 3752-3767.
[5]	DAI Y， LI X， ZHOU F， et al. One-stage cascade refinement networks for infrared small target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2023， 61： No.5000917.
[6]	McINTOSH B， VENKATARAMANAN S， MAHALANOBIS A. Infrared target detection in cluttered environments by maximization of a Target to Clutter Ratio （TCR） metric using a convolutional neural network ［J］. IEEE Transactions on Aerospace and Electronic Systems， 2021， 57（1）： 485-496.
[7]	WANG H， ZHOU L， WANG L. Miss detection vs. false alarm： adversarial learning for small object segmentation in infrared images ［C］// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2019： 8508-8517.
[8]	DAI Y， WU Y， ZHOU F， et al. Attentional local contrast networks for infrared small target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2021， 59（11）： 9813-9824.
[9]	LI B， XIAO C， WANG L， et al. Dense nested attention network for infrared small target detection ［J］. IEEE Transactions on Image Processing， 2023， 32： 1745-1758.
[10]	WU X， HONG D， CHANUSSOT J. UIU-Net： U-Net in U-Net for infrared small object detection ［J］. IEEE Transactions on Image Processing， 2023， 32： 364-376.
[11]	CHUNG W Y， LEE I H， PARK C G. Lightweight infrared small target detection network using full-scale skip connection U-Net ［J］. IEEE Geoscience and Remote Sensing Letters， 2023， 20： No.7000705.
[12]	ZHANG T， LI L， CAO S， et al. Attention-guided pyramid context networks for detecting infrared small target under complex background ［J］. IEEE Transactions on Aerospace and Electronic Systems， 2023， 59（4）： 4250-4261.
[13]	ZHANG M， ZHANG R， ZHANG J， et al. Dim2Clear network for infrared small target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2023， 61： No.5001714.
[14]	ZHANG M， ZHANG R， YANG Y， et al. ISNet： shape matters for infrared small target detection ［C］// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2022： 867-876.
[15]	刘奎，唐慧萍，苏本跃.门控卷积和高频特征融合的红外小目标检测［J］.计算机工程与应用， 2025， 61（7）： 306-314.
	LIU K， TANG H P， SU B Y. Gated convolution and high-frequency feature fusion for infrared small target detection ［J］. Computer Engineering and Applications， 2025， 61（7）： 306-314.
[16]	ZHANG M， YANG H， GUO J， et al. IRPruneDet： efficient infrared small target detection via wavelet structure-regularized soft channel pruning ［C］// Proceedings of the 38th AAAI Conference on Artificial Intelligence. Palo Alto： AAAI Press， 2024： 7224-7232.
[17]	代少升，刘科生，黄炼，等.基于视觉Transformer和双解码器的红外小目标检测方法［J］.红外技术， 2024， 46（9）： 1070-1080.
	DAI S S， LIU K S， HUANG L， et al. Infrared small target detection method with Vision Transformer and dual decoder ［J］. Infrared Technology， 2024， 46（9）： 1070-1080.
[18]	FAN L， WANG Y， HU G， et al. Diffusion-based continuous feature representation for infrared small-dim target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2024， 62： No.5003617.
[19]	YANG H， MU T， DONG Z， et al. PBT： progressive background-aware transformer for infrared small target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2024， 62： No.5004513.
[20]	HUANG Y， ZHI X， HU J， et al. FDDBA-NET： frequency domain decoupling bidirectional interactive attention network for infrared small target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2024， 62： No.5004416.
[21]	LIU Q， LIU R， ZHENG B， et al. Infrared small target detection with scale and location sensitivity ［C］// Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2024： 17490-17499.
[22]	GUO T， ZHOU B， LUO F， et al. DMFNet： dual-encoder multistage feature fusion network for infrared small target detection ［J］. IEEE Transactions on Geoscience and Remote Sensing， 2024， 62： No.5614214.
[23]	PENG S， JI L， CHEN S， et al. Moving infrared dim and small target detection by mixed spatio-temporal encoding ［J］. Engineering Applications of Artificial Intelligence， 2025， 144： No.110100.

方法	NUAA-SIRST				IRSTD-1k				参数量/10⁶	浮点运算量/GFLOPs
方法	IoU/%	nIoU/%	Pd/%	Fa/10^-6	IoU/%	nIoU/%	Pd/%	Fa/10^-6	参数量/10⁶	浮点运算量/GFLOPs
IPI^［3］	25.67	24.57	85.55	11.47	27.92	20.46	81.37	16.18	—	—
RIPT^［4］	11.05	10.15	79.08	22.61	14.11	8.09	77.55	28.31	—	—
MDvsFA^［7］	60.30	58.26	89.35	56.35	49.50	47.41	82.11	80.33	3.59	230.14
ALCNet^［8］	74.31	73.12	97.34	20.21	62.05	59.58	92.19	31.56	8.56	14.52
DNANet^［9］	77.54	76.06	98.10	2.51	67.54	66.27	91.92	8.81	4.69	56.34
AMFUNet^［11］	75.86	73.74	100.00	5.86	67.85	64.98	89.90	15.98	2.17	23.80
AGPCNet^［12］	69.09	73.37	93.94	50.00	62.39	61.80	90.91	21.29	11.79	40.22
Dim2Clear［^13］	77.20	75.20	99.10	6.72	66.30	64.20	93.70	20.90	—	—
DCFR-Net^［18］	76.23	74.69	99.08	6.52	65.41	65.45	93.60	7.35	—	—
MSHNet^［21］	76.93	74.08	99.10	5.86	68.57	66.03	91.91	17.63	4.07	24.43
本文方法	78.11	75.44	100.00	9.58	69.08	66.34	93.94	13.55	0.86	17.71

方法	NUAA-SIRST				IRSTD-1k				参数量/10⁶	浮点运算量/GFLOPs
方法	IoU/%	nIoU/%	Pd/%	Fa/10^-6	IoU/%	nIoU/%	Pd/%	Fa/10^-6	参数量/10⁶	浮点运算量/GFLOPs
IPI^［3］	25.67	24.57	85.55	11.47	27.92	20.46	81.37	16.18	—	—
RIPT^［4］	11.05	10.15	79.08	22.61	14.11	8.09	77.55	28.31	—	—
MDvsFA^［7］	60.30	58.26	89.35	56.35	49.50	47.41	82.11	80.33	3.59	230.14
ALCNet^［8］	74.31	73.12	97.34	20.21	62.05	59.58	92.19	31.56	8.56	14.52
DNANet^［9］	77.54	76.06	98.10	2.51	67.54	66.27	91.92	8.81	4.69	56.34
AMFUNet^［11］	75.86	73.74	100.00	5.86	67.85	64.98	89.90	15.98	2.17	23.80
AGPCNet^［12］	69.09	73.37	93.94	50.00	62.39	61.80	90.91	21.29	11.79	40.22
Dim2Clear［^13］	77.20	75.20	99.10	6.72	66.30	64.20	93.70	20.90	—	—
DCFR-Net^［18］	76.23	74.69	99.08	6.52	65.41	65.45	93.60	7.35	—	—
MSHNet^［21］	76.93	74.08	99.10	5.86	68.57	66.03	91.91	17.63	4.07	24.43
本文方法	78.11	75.44	100.00	9.58	69.08	66.34	93.94	13.55	0.86	17.71

方法				NUAA-SIRST				IRSTD-1k
Baseline	CLFG	MSCA	FDIA	IoU/%	nIoU/%	Pd/%	Fa/10^-6	IoU/%	nIoU/%	Pd/%	Fa/10^-6
√				74.05	72.37	99.08	23.24	65.29	62.16	90.90	19.81
√	√			75.24	72.46	99.08	18.99	66.08	65.18	91.25	16.36
√	√	√		77.60	74.82	100.00	10.47	68.41	64.23	93.27	17.50
√	√	√	√	78.11	75.44	100.00	9.58	69.08	66.34	93.94	13.55

方法				NUAA-SIRST				IRSTD-1k
Baseline	CLFG	MSCA	FDIA	IoU/%	nIoU/%	Pd/%	Fa/10^-6	IoU/%	nIoU/%	Pd/%	Fa/10^-6
√				74.05	72.37	99.08	23.24	65.29	62.16	90.90	19.81
√	√			75.24	72.46	99.08	18.99	66.08	65.18	91.25	16.36
√	√	√		77.60	74.82	100.00	10.47	68.41	64.23	93.27	17.50
√	√	√	√	78.11	75.44	100.00	9.58	69.08	66.34	93.94	13.55

方法	NUAA-SIRST训练+IRSTD-1k测试				IRSTD-1k训练+NUAA-SIRST测试
方法	IoU/%	nIoU/%	Pd/%	Fa/10^-6	IoU/%	nIoU/%	Pd/%	Fa/10^-6
DNANet^［9］	51.38	58.91	89.23	99.20	60.89	71.85	95.41	88.18
AMFUNet^［11］	57.58	57.76	90.24	29.65	62.24	67.87	90.83	61.92
MSHNet^［21］	55.89	59.60	90.24	29.95	65.67	70.51	94.50	31.05
本文方法	58.96	60.86	90.91	21.71	70.42	69.96	93.58	22.00