Infrared Physics & Technology最新文献_第9页

A novel encoded line laser array method of scanning infrared thermography nondestructive testing for CFRP defect 一种新型编码线激光阵列扫描红外热成像CFRP缺陷无损检测方法

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-14 DOI: 10.1016/j.infrared.2026.106397

Rui Li , Di Wu , Yunsheng An , Yucai Xie , Hongpeng Zhang , Jizhe Wang , Wei Li , Chenzhao Bai , Chenyong Wang , Li Sun

This paper presents a novel encoded line laser array (ELLA) method for scanning infrared thermography nondestructive testing (IRT‑NDT) of carbon fiber reinforced plastics (CFRP). By spatially arranging multiple line lasers, ELLA generates Lock‑in like, frequency‑modulation like, and 13‑bit Barker‑coded pulse like (Barker13‑like) excitation waveforms, extending modulated thermography from static to scanning systems. Numerical simulations confirmed that ELLA heating profiles closely match static excitations. To process dynamic image sequences, pseudo‑static matrix reconstruction (PSMR) converts them into spatially static datasets, enabling established algorithms such as fast Fourier transform, thermal signal reconstruction, and matched filtering (MF). The algorithmic results demonstrated that, compared with single line laser scanning, the combination of ELLA with PSMR and post‑processing effectively improves defect detectability, even under added salt‑and‑pepper and Gaussian noise. Both the signal‑to‑clutter ratio and signal‑to‑noise ratio (SNR) evaluations confirm the effectiveness of ELLA, with overall values being improved by post‑processing algorithms. Barker13‑like excitation achieves consistently high SNR across most algorithms, and the moving average model of MF reached the maximum SNR of 12.91. Overall, ELLA integrates the rapid coverage of scanning IRT‑NDT with the sensitivity of modulated excitation, offering a feasible and efficient approach for large‑scale CFRP defect detection.

提出了一种用于碳纤维增强塑料（CFRP）扫描红外热成像无损检测（IRT - NDT）的新型编码线激光阵列（ELLA）方法。通过在空间上安排多行激光器，ELLA产生类似锁相、频率调制和13位巴克编码脉冲（类似巴克13）的激励波形，将调制热成像从静态扩展到扫描系统。数值模拟证实了ELLA的加热分布与静态激励密切匹配。为了处理动态图像序列，伪静态矩阵重构（PSMR）将其转换为空间静态数据集，从而实现快速傅立叶变换、热信号重构和匹配滤波（MF）等已建立的算法。算法结果表明，与单线激光扫描相比，ELLA与PSMR和后处理相结合，即使在添加盐和胡椒和高斯噪声的情况下，也能有效提高缺陷的可检测性。信杂波比和信噪比（SNR）评估都证实了ELLA的有效性，并且通过后处理算法提高了总体值。Barker13 - like激励在大多数算法中都获得了一致的高信噪比，并且MF的移动平均模型达到了12.91的最大信噪比。总的来说，ELLA结合了扫描IRT - NDT的快速覆盖和调制激励的灵敏度，为大规模CFRP缺陷检测提供了一种可行而有效的方法。

{"title":"A novel encoded line laser array method of scanning infrared thermography nondestructive testing for CFRP defect","authors":"Rui Li , Di Wu , Yunsheng An , Yucai Xie , Hongpeng Zhang , Jizhe Wang , Wei Li , Chenzhao Bai , Chenyong Wang , Li Sun","doi":"10.1016/j.infrared.2026.106397","DOIUrl":"10.1016/j.infrared.2026.106397","url":null,"abstract":"<div><div>This paper presents a novel encoded line laser array (ELLA) method for scanning infrared thermography nondestructive testing (IRT‑NDT) of carbon fiber reinforced plastics (CFRP). By spatially arranging multiple line lasers, ELLA generates Lock‑in like, frequency‑modulation like, and 13‑bit Barker‑coded pulse like (Barker13‑like) excitation waveforms, extending modulated thermography from static to scanning systems. Numerical simulations confirmed that ELLA heating profiles closely match static excitations. To process dynamic image sequences, pseudo‑static matrix reconstruction (PSMR) converts them into spatially static datasets, enabling established algorithms such as fast Fourier transform, thermal signal reconstruction, and matched filtering (MF). The algorithmic results demonstrated that, compared with single line laser scanning, the combination of ELLA with PSMR and post‑processing effectively improves defect detectability, even under added salt‑and‑pepper and Gaussian noise. Both the signal‑to‑clutter ratio and signal‑to‑noise ratio (SNR) evaluations confirm the effectiveness of ELLA, with overall values being improved by post‑processing algorithms. Barker13‑like excitation achieves consistently high SNR across most algorithms, and the moving average model of MF reached the maximum SNR of 12.91. Overall, ELLA integrates the rapid coverage of scanning IRT‑NDT with the sensitivity of modulated excitation, offering a feasible and efficient approach for large‑scale CFRP defect detection.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106397"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146074068","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

DWPCNNFusion: Deep pulse-coupled neural networks incorporating Weber’s law for efficient infrared and visible image fusion DWPCNNFusion：采用韦伯定律的深度脉冲耦合神经网络，用于有效的红外和可见光图像融合

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-29 DOI: 10.1016/j.infrared.2026.106417

Jia Zhao , Sirui Jia , Jing Di , Jing Lian , Yide Ma , Yuelan Xin , Jisheng Dang , Jizhao Liu

Infrared and visible image fusion is a key task in computer vision, aiming to combine complementary multimodal information to generate a salient and texture-rich image. However, existing deep learning-based fusion methods typically rely on increasing network depth to enhance performance, often overlooking the significant computational resources required, which leads to inefficiency. To address this, we propose a novel brain-inspired, end-to-end trainable infrared and visible image fusion method (DWPCNNFusion). Specifically, in the feature extraction stage, we design a deep pulse-coupled neural networks based on Weber’s law (DWPCNN) , where the coupling weight matrix is treated as a learnable parameter, enabling the network to flexibly adapt to varying data characteristics. Additionally, linking strength coefficients are set according to Weber’s law, simulating the nonlinear perception of brightness in the human visual system, which effectively mitigates detail loss in low-light environments. To accommodate dynamic changes in input data over time, a time adaptive batch normalization method is proposed, and temporal information is integrated via a rate encoding scheme, allowing DWPCNN to be efficiently incorporated into existing deep learning frameworks. Furthermore, pulse convolutional dense blocks (PCDB) are employed to extract high-level semantic features, further enhancing the model’s feature representation capability. Experimental results on the TNO and MSRS datasets, compared with 15 representative methods using both objective and subjective metrics, demonstrate that the proposed method excels in detail preservation while achieving a better balance between computational efficiency and fusion performance.

红外图像与可见光图像融合是计算机视觉中的一项关键任务，其目的是将互补的多模态信息结合在一起，生成显著且纹理丰富的图像。然而，现有的基于深度学习的融合方法通常依赖于增加网络深度来提高性能，往往忽略了所需的大量计算资源，从而导致效率低下。为了解决这个问题，我们提出了一种新颖的脑启发，端到端可训练的红外和可见光图像融合方法（DWPCNNFusion）。具体而言，在特征提取阶段，我们设计了基于韦伯定律的深度脉冲耦合神经网络（DWPCNN），将耦合权矩阵作为可学习参数，使网络能够灵活地适应不同的数据特征。此外，根据韦伯定律设置连接强度系数，模拟人类视觉系统对亮度的非线性感知，有效减轻了弱光环境下的细节损失。为了适应输入数据随时间的动态变化，提出了一种时间自适应批归一化方法，并通过速率编码方案集成时间信息，使DWPCNN能够有效地融入现有的深度学习框架。此外，采用脉冲卷积密集块（PCDB）提取高级语义特征，进一步增强了模型的特征表示能力。在TNO和MSRS数据集上的实验结果表明，该方法在计算效率和融合性能之间取得了更好的平衡，并与15种具有代表性的客观和主观度量方法进行了比较。

{"title":"DWPCNNFusion: Deep pulse-coupled neural networks incorporating Weber’s law for efficient infrared and visible image fusion","authors":"Jia Zhao , Sirui Jia , Jing Di , Jing Lian , Yide Ma , Yuelan Xin , Jisheng Dang , Jizhao Liu","doi":"10.1016/j.infrared.2026.106417","DOIUrl":"10.1016/j.infrared.2026.106417","url":null,"abstract":"<div><div>Infrared and visible image fusion is a key task in computer vision, aiming to combine complementary multimodal information to generate a salient and texture-rich image. However, existing deep learning-based fusion methods typically rely on increasing network depth to enhance performance, often overlooking the significant computational resources required, which leads to inefficiency. To address this, we propose a novel brain-inspired, end-to-end trainable infrared and visible image fusion method (DWPCNNFusion). Specifically, in the feature extraction stage, we design a deep pulse-coupled neural networks based on Weber’s law (DWPCNN) , where the coupling weight matrix is treated as a learnable parameter, enabling the network to flexibly adapt to varying data characteristics. Additionally, linking strength coefficients are set according to Weber’s law, simulating the nonlinear perception of brightness in the human visual system, which effectively mitigates detail loss in low-light environments. To accommodate dynamic changes in input data over time, a time adaptive batch normalization method is proposed, and temporal information is integrated via a rate encoding scheme, allowing DWPCNN to be efficiently incorporated into existing deep learning frameworks. Furthermore, pulse convolutional dense blocks (PCDB) are employed to extract high-level semantic features, further enhancing the model’s feature representation capability. Experimental results on the TNO and MSRS datasets, compared with 15 representative methods using both objective and subjective metrics, demonstrate that the proposed method excels in detail preservation while achieving a better balance between computational efficiency and fusion performance.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106417"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146074070","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Noise influence on defect depth estimation in CFRP by One-Dimensional convolution neural network in Non-Linear frequency modulated thermal wave Imaging 非线性调频热波成像中一维卷积神经网络对CFRP缺陷深度估计的噪声影响

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-19 DOI: 10.1016/j.infrared.2026.106398

Naga Prasanthi Yerneni , S.S. Banda , G.T. Vesala , V.S. Ghali , Fei Wang , Junyan Liu , A. Sridhar , Ravibabu Mulaveesala

Quantitative depth estimation of defects in composite structures using active infrared thermography is a challenging task and recent advancements promote deep learning (DL) techniques to achieve this in a quick and automated manner. However, noise present in thermal data at extreme inspection conditions degrades the performance of the DL models. This paper introduces a one-dimensional convolution neural network (1D-CNN) and analyzes its performance in different levels of noise for automatic depth estimation in Logarithmic Frequency Modulated Thermal Wave Imaging. The experimentation is conducted over two carbon fiber reinforced polymer (CFRP) specimens of different thickness and varying sizes and depths of flat-bottom hole defects. Including the original thermal data of one CFRP specimen, five datasets are prepared by adding additive white Gaussian noise of four levels: 5 dB, 10 dB, 15 dB and 20 dB and the proposed 1D-CNN model, named as M1 to M5 for each case, is individually trained and tested. Further, transfer learning is applied for identifying depths in the second CFRP structure. The results, along with performance metrics, indicate that the 1D-CNN presents more than 98 % and 95 % accuracy for training from scratch and transfer learning cases over original data, whereas it degrades as the noise level increases. Performance metrics such as accuracy, F-score and mean intersection of union demonstrate the defect depth estimation performance, including the defect boundary preservation under various noise conditions that are consistent with the accuracy of the model. In addition, the proposed 1D-CNN achieves the best quantitative performance, with consistently higher accuracy (≈95–98 %), F-score (≈0.85–0.95), and mean IoU (≈0.45–0.75) across all defect depths and noise levels, outperforming conventional DT, SVM, ANN, and 1D-ResNet, whose metrics drop sharply, for deeper defects and higher noise conditions.

利用主动红外热成像技术对复合材料结构中的缺陷进行深度定量估计是一项具有挑战性的任务，最近的进展促进了深度学习（DL）技术以快速和自动化的方式实现这一目标。然而，在极端的检测条件下，热数据中的噪声会降低深度学习模型的性能。介绍了一维卷积神经网络（1D-CNN），分析了其在不同噪声水平下用于对数调频热波成像深度自动估计的性能。在两个不同厚度的碳纤维增强聚合物（CFRP）试件上进行了不同尺寸和深度的平底孔缺陷试验。包括一个CFRP试样的原始热数据，通过添加5 dB、10 dB、15 dB和20 dB四个级别的高斯白噪声制备5个数据集，并对每种情况下提出的1D-CNN模型分别进行训练和测试，命名为M1 ~ M5。进一步，将迁移学习应用于第二CFRP结构的深度识别。结果以及性能指标表明，1D-CNN在原始数据的从头训练和迁移学习案例中呈现出超过98%和95%的准确率，而随着噪声水平的增加，准确率会下降。精度、F-score和平均交集等性能指标展示了缺陷深度估计的性能，包括在各种噪声条件下的缺陷边界保持，与模型的精度相一致。此外，本文提出的1D-CNN具有最佳的定量性能，在所有缺陷深度和噪声水平上都具有较高的精度（≈95 - 98%）、f分数（≈0.85-0.95）和平均IoU(≈0.45-0.75)，优于传统的DT、SVM、ANN和1D-ResNet，后者在缺陷深度和噪声水平较高的情况下指标急剧下降。

{"title":"Noise influence on defect depth estimation in CFRP by One-Dimensional convolution neural network in Non-Linear frequency modulated thermal wave Imaging","authors":"Naga Prasanthi Yerneni , S.S. Banda , G.T. Vesala , V.S. Ghali , Fei Wang , Junyan Liu , A. Sridhar , Ravibabu Mulaveesala","doi":"10.1016/j.infrared.2026.106398","DOIUrl":"10.1016/j.infrared.2026.106398","url":null,"abstract":"<div><div>Quantitative depth estimation of defects in composite structures using active infrared thermography is a challenging task and recent advancements promote deep learning (DL) techniques to achieve this in a quick and automated manner. However, noise present in thermal data at extreme inspection conditions degrades the performance of the DL models. This paper introduces a one-dimensional convolution neural network (1D-CNN) and analyzes its performance in different levels of noise for automatic depth estimation in Logarithmic Frequency Modulated Thermal Wave Imaging. The experimentation is conducted over two carbon fiber reinforced polymer (CFRP) specimens of different thickness and varying sizes and depths of flat-bottom hole defects. Including the original thermal data of one CFRP specimen, five datasets are prepared by adding additive white Gaussian noise of four levels: 5 dB, 10 dB, 15 dB and 20 dB and the proposed 1D-CNN model, named as M1 to M5 for each case, is individually trained and tested. Further, transfer learning is applied for identifying depths in the second CFRP structure. The results, along with performance metrics, indicate that the 1D-CNN presents more than 98 % and 95 % accuracy for training from scratch and transfer learning cases over original data, whereas it degrades as the noise level increases. Performance metrics such as accuracy, F-score and mean intersection of union demonstrate the defect depth estimation performance, including the defect boundary preservation under various noise conditions that are consistent with the accuracy of the model. In addition, the proposed 1D-CNN achieves the best quantitative performance, with consistently higher accuracy (≈95–98 %), F-score (≈0.85–0.95), and mean IoU (≈0.45–0.75) across all defect depths and noise levels, outperforming conventional DT, SVM, ANN, and 1D-ResNet, whose metrics drop sharply, for deeper defects and higher noise conditions.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106398"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147397478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Infrared small target detection based on multiscale low-rankness and firm thresholding function 基于多尺度低秩稳健阈值函数的红外小目标检测

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-02-06 DOI: 10.1016/j.infrared.2026.106439

Xiaorun Li, Yuan Luo, Shuhan Chen

Recently, infrared (IR) small target detection based on low-rank and sparse decomposition (LRSD) has gained increasing attention among civilian and military fields. Whereas, the LRSD-based methods face the challenges of accurately mathematically describing the intrinsic characteristics of each component and effectively separating targets from non-targets. This paper proposes a target detection model with multiscale low-rankness and firm thresholding function (MSR-FT). Specifically, a high-pass filtering spatial–temporal tensor corresponding to the original IR sequence is constructed. Secondly, we define a firm thresholding function-based norm for target characterization, leading to more accurate sparse target estimation. Meanwhile, using a multiscale low-rank background decomposition technique, we introduce a multiscale Log-based tensor nuclear norm, which ensures that the estimated background fully considers the global and local low-rank properties from different scales. Furthermore, with a posterior information feedback strategy, we propose a target detection method called MSR-FT. Through an optimization scheme based on the alternating direction method of multipliers (ADMM), it proves that MSR-FT exceeds seventeen competitive IR small target detection methods on six IR sequences from the perspectives of target detectability (TD), background suppressibility (BS), and overall performance.

近年来，基于低秩稀疏分解（LRSD）的红外小目标检测越来越受到民用和军事领域的关注。然而，基于lrsd的方法面临着精确数学描述各成分内在特征和有效分离目标与非目标的挑战。提出了一种多尺度低秩稳健阈值函数（MSR-FT）目标检测模型。具体而言，构建了一个与原始红外序列相对应的高通滤波时空张量。其次，我们定义了一个基于坚定阈值函数的目标表征范数，使得稀疏目标估计更加准确。同时，采用多尺度低秩背景分解技术，引入基于对数的多尺度张量核范数，保证了估计的背景充分考虑了不同尺度的全局和局部低秩特性。在此基础上，利用后验信息反馈策略，提出了一种称为MSR-FT的目标检测方法。通过一种基于乘子交替方向法（ADMM）的优化方案，从目标可探测性（TD）、背景抑制性（BS）和综合性能三个方面证明了MSR-FT在6个红外序列上优于17种竞争红外小目标检测方法。

{"title":"Infrared small target detection based on multiscale low-rankness and firm thresholding function","authors":"Xiaorun Li, Yuan Luo, Shuhan Chen","doi":"10.1016/j.infrared.2026.106439","DOIUrl":"10.1016/j.infrared.2026.106439","url":null,"abstract":"<div><div>Recently, infrared (IR) small target detection based on low-rank and sparse decomposition (LRSD) has gained increasing attention among civilian and military fields. Whereas, the LRSD-based methods face the challenges of accurately mathematically describing the intrinsic characteristics of each component and effectively separating targets from non-targets. This paper proposes a target detection model with multiscale low-rankness and firm thresholding function (MSR-FT). Specifically, a high-pass filtering spatial–temporal tensor corresponding to the original IR sequence is constructed. Secondly, we define a firm thresholding function-based norm for target characterization, leading to more accurate sparse target estimation. Meanwhile, using a multiscale low-rank background decomposition technique, we introduce a multiscale Log-based tensor nuclear norm, which ensures that the estimated background fully considers the global and local low-rank properties from different scales. Furthermore, with a posterior information feedback strategy, we propose a target detection method called MSR-FT. Through an optimization scheme based on the alternating direction method of multipliers (ADMM), it proves that MSR-FT exceeds seventeen competitive IR small target detection methods on six IR sequences from the perspectives of target detectability (TD), background suppressibility (BS), and overall performance.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106439"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147397483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Early detection of tobacco leaf mildew using multi-attention enhanced 3D residual convolutional Neural network with hyperspectral imaging 基于多关注增强三维残差卷积神经网络的高光谱成像烟草叶霉病早期检测

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-06 DOI: 10.1016/j.infrared.2026.106375

Lei Zhang , Jinsong Du , Jiakang Li , Chengyuan Li , Jiandong Zhang , Lianlian Wu

Early detection of mildew in tobacco leaves is essential for maintaining product quality. While hyperspectral imaging (HSI) offers a non-destructive alternative with rich spectral–spatial information, but the high dimensionality of HSI and complex characteristics of early mildew pose significant challenges for conventional deep learning approach. In this article, we propose a novel multi-attention enhanced 3D Residual Convolutional Neural Network (3D-ResCNN) for early mildew detection of tobacco leaves using HSI data. First, the model employs 3D convolutions to simultaneously extract spatial and spectral features, while residual connections mitigate the vanishing gradient problem in deep networks. To improve mildew localization and spectral discrimination, a spatial–spectral attention module is integrated to selectively emphasize mildew-sensitive spatial regions and identify key spectral bands. Subsequently, a channel attention mechanism is introduced to adaptively reweight feature channels, thereby suppressing redundancy and emphasizing the most discriminative feature maps. Extensive experiments conducted on a real-world HSI tobacco dataset demonstrate that the proposed method achieves superior performance over traditional deep learning models in terms of accuracy and early-stage detection sensitivity, which validate the model’s effectiveness and superiority.

早期发现烟草叶片霉变对保持产品质量至关重要。虽然高光谱成像（HSI）提供了一种具有丰富光谱空间信息的非破坏性替代方法，但高光谱成像的高维性和早期霉菌的复杂特征给传统的深度学习方法带来了重大挑战。在本文中，我们提出了一种新的多注意力增强的3D残差卷积神经网络（3D- rescnn），用于利用HSI数据进行烟草叶片的早期霉变检测。首先，该模型采用三维卷积同时提取空间和光谱特征，残差连接缓解了深度网络中的梯度消失问题。为了提高霉菌的定位和光谱识别能力，集成了空间-光谱关注模块，选择性地强调霉菌敏感的空间区域，识别关键的光谱波段。随后，引入通道注意机制自适应地重加权特征通道，从而抑制冗余并强调最具判别性的特征映射。在真实HSI烟草数据集上进行的大量实验表明，该方法在准确性和早期检测灵敏度方面优于传统深度学习模型，验证了该模型的有效性和优越性。

{"title":"Early detection of tobacco leaf mildew using multi-attention enhanced 3D residual convolutional Neural network with hyperspectral imaging","authors":"Lei Zhang , Jinsong Du , Jiakang Li , Chengyuan Li , Jiandong Zhang , Lianlian Wu","doi":"10.1016/j.infrared.2026.106375","DOIUrl":"10.1016/j.infrared.2026.106375","url":null,"abstract":"<div><div>Early detection of mildew in tobacco leaves is essential for maintaining product quality. While hyperspectral imaging (HSI) offers a non-destructive alternative with rich spectral–spatial information, but the high dimensionality of HSI and complex characteristics of early mildew pose significant challenges for conventional deep learning approach. In this article, we propose a novel multi-attention enhanced 3D Residual Convolutional Neural Network (3D-ResCNN) for early mildew detection of tobacco leaves using HSI data. First, the model employs 3D convolutions to simultaneously extract spatial and spectral features, while residual connections mitigate the vanishing gradient problem in deep networks. To improve mildew localization and spectral discrimination, a spatial–spectral attention module is integrated to selectively emphasize mildew-sensitive spatial regions and identify key spectral bands. Subsequently, a channel attention mechanism is introduced to adaptively reweight feature channels, thereby suppressing redundancy and emphasizing the most discriminative feature maps. Extensive experiments conducted on a real-world HSI tobacco dataset demonstrate that the proposed method achieves superior performance over traditional deep learning models in terms of accuracy and early-stage detection sensitivity, which validate the model’s effectiveness and superiority.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106375"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923453","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Corrigendum to “Infrared small target detection fusion network based on singular value decomposition and rayleigh feature amplifier”. [Infrared Phys. Technol. 153(2026) 106353, ISSN 1350–4495, https://doi.org/10.1016/j.infrared.2025.106353] 基于奇异值分解和瑞利特征放大器的红外小目标检测融合网络勘误表。(红外物理。科技进展，153(2026)106353,ISSN 1350-4495, https://doi.org/10.1016/j.infrared.2025.106353]

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-11 DOI: 10.1016/j.infrared.2026.106390

Tianlei Ma , Liang Fu , Jinzhu Peng , Fang-Lue Zhang , Heng Zhang , Xiangbo Feng , Teng Wang , Lu Xin

引用次数: 0

Research on measurement accuracy correction for TDLAS-based methane leakage monitoring under environmental variations 环境变化下基于tlas的甲烷泄漏监测测量精度校正研究

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-05 DOI: 10.1016/j.infrared.2026.106371

Jinyu Li, Feng He, Xiaokang Liu, Fangting Liu, Junhui Li

Tunable diode laser absorption spectroscopy (TDLAS), owing to its high selectivity, sensitivity, and fast response, has been widely employed for monitoring methane leakage in urban natural gas pipelines. However, variations in ambient temperature and pressure can alter the absorption spectral lines, thereby reducing the accuracy of concentration measurements. To address this issue and improve measurement reliability, a 1654 nm DFB laser was adopted as the light source, and methane (CH₄) at different concentrations was used as the target gas for testing under conditions of 263–323 K and 0.6–1.1 atm. To handle temperature effects, we developed two corrections: one for direct absorption spectroscopy (DAS) that integrates line-strength variation with a system-error compensation coefficient, and another for wavelength modulation spectroscopy (WMS) based on dual-peak combined intensity, while pressure effects were mitigated via a least-squares correction. The temperature correction reduced the maximum relative errors of DAS and WMS from about 30 % and 20 % to around 2 %, respectively, while the pressure correction decreased the maximum relative error from 3.69 % to 1.05 %. Allan deviation analysis indicated that the sensor achieved a minimum detection limit (MDL) of 4.41 ppm at an integration time of 30 s. In a 24-hour continuous monitoring test conducted under fluctuating temperature conditions, the maximum relative errors for measuring 1 × 10⁴ ppm CH₄, after applying the correction formulas, were reduced to 1.92 % for DAS and 0.84 % for WMS. This study provides a novel and effective approach to enhancing gas concentration measurement accuracy in urban natural gas pipeline leakage detection and related industrial applications.

可调谐二极管激光吸收光谱（TDLAS）以其高选择性、高灵敏度和快速响应等优点，在城市天然气管道甲烷泄漏监测中得到了广泛应用。然而，环境温度和压力的变化会改变吸收谱线，从而降低浓度测量的准确性。为了解决这一问题，提高测量的可靠性，采用1654 nm DFB激光器作为光源，以不同浓度的甲烷（CH4）作为目标气体，在263 ~ 323 K、0.6 ~ 1.1 atm条件下进行测试。为了处理温度影响，我们开发了两种校正方法：一种用于直接吸收光谱（DAS），将线强度变化与系统误差补偿系数相结合；另一种用于波长调制光谱（WMS），基于双峰组合强度，同时通过最小二乘校正减轻压力影响。温度校正将DAS和WMS的最大相对误差分别从30%和20%减小到2%左右，压力校正将最大相对误差从3.69%减小到1.05%。Allan偏差分析表明，传感器在30秒的集成时间内实现了4.41 ppm的最小检测限（MDL）。在波动温度条件下进行的24小时连续监测试验中，应用校正公式后，DAS测量1 × 104 ppm CH4的最大相对误差降至1.92%，WMS测量的最大相对误差降至0.84%。本研究为提高城市天然气管道泄漏检测及相关工业应用中气体浓度测量精度提供了一种新颖有效的方法。

{"title":"Research on measurement accuracy correction for TDLAS-based methane leakage monitoring under environmental variations","authors":"Jinyu Li, Feng He, Xiaokang Liu, Fangting Liu, Junhui Li","doi":"10.1016/j.infrared.2026.106371","DOIUrl":"10.1016/j.infrared.2026.106371","url":null,"abstract":"<div><div>Tunable diode laser absorption spectroscopy (TDLAS), owing to its high selectivity, sensitivity, and fast response, has been widely employed for monitoring methane leakage in urban natural gas pipelines. However, variations in ambient temperature and pressure can alter the absorption spectral lines, thereby reducing the accuracy of concentration measurements. To address this issue and improve measurement reliability, a 1654 nm DFB laser was adopted as the light source, and methane (CH<sub>4</sub>) at different concentrations was used as the target gas for testing under conditions of 263–323 K and 0.6–1.1 atm. To handle temperature effects, we developed two corrections: one for direct absorption spectroscopy (DAS) that integrates line-strength variation with a system-error compensation coefficient, and another for wavelength modulation spectroscopy (WMS) based on dual-peak combined intensity, while pressure effects were mitigated via a least-squares correction. The temperature correction reduced the maximum relative errors of DAS and WMS from about 30 % and 20 % to around 2 %, respectively, while the pressure correction decreased the maximum relative error from 3.69 % to 1.05 %. Allan deviation analysis indicated that the sensor achieved a minimum detection limit (MDL) of 4.41 ppm at an integration time of 30 s. In a 24-hour continuous monitoring test conducted under fluctuating temperature conditions, the maximum relative errors for measuring 1 × 10<sup>4</sup> ppm CH<sub>4</sub>, after applying the correction formulas, were reduced to 1.92 % for DAS and 0.84 % for WMS. This study provides a novel and effective approach to enhancing gas concentration measurement accuracy in urban natural gas pipeline leakage detection and related industrial applications.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106371"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Self-Q-switching laser performance of Nd:ASL crystals at 1.3 μm 1.3 μm Nd:ASL晶体的自调q激光性能

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2025-12-29 DOI: 10.1016/j.infrared.2025.106360

Wenfang Lin , Conghui Huang , Shulong Zhang , Min Xu , Siliang Tao , Shanming Li , Chengchun Zhao , Qiannan Fang , Xisheng Ye , Yin Hang

The self-Q-switching (SQS) laser performance on Nd-doped crystal at 1.3 μm has been reported for the first time, as far as is known. On Sr_0.7Nd_0.05La_0.25Mg_0.3Al_11.7O₁₉ (Nd:ASL) disorder crystal, a SQS dual-wavelength laser at 1339.9 and 1370.3 nm with output power up to 1.65 W was obtained under an absorbed pump power of 10.13 W with slope and optical-to-optical efficiencies of 22.3 % and 16.3 %, respectively. Furthermore, an on-surface optical axis quartz birefringent filter (BRF) was inserted in the V-folded cavity to tune the laser wavelength. Lasers at 1306.4, and approximately 1340, 1370, or 1391 nm were obtained. The experimental results indicated that σ polarization direction Nd:ASL is capable of producing dual-wavelength lasers at 1339.9 and 1370.3 nm, which was potential to be employed as the source of THz radiation. Besides, Nd:ASL crystals are enable to generate tunable lasers near 1370 and 1391 nm.

1.3 μm掺杂nd晶体上的自调q （SQS）激光性能是目前所知的首次报道。在sr0.7 nd0.05 la0.25 mg0.3 al11.70 o19 （Nd:ASL）无序晶体上，在吸收泵浦功率为10.13 W的条件下，获得了波长为1339.9和1370.3 nm、输出功率为1.65 W的SQS双波长激光器，其斜率和光效率分别为22.3%和16.3%。此外，在v型折叠腔中插入表面光轴石英双折射滤光片（BRF）来调节激光波长。获得了波长为1306.4 nm、1340 nm、1370 nm或1391 nm的激光。实验结果表明，σ偏振方向Nd:ASL能够产生1339.9和1370.3 nm的双波长激光器，具有作为太赫兹辐射源的潜力。此外，Nd:ASL晶体能够产生1370和1391 nm附近的可调谐激光。

{"title":"Self-Q-switching laser performance of Nd:ASL crystals at 1.3 μm","authors":"Wenfang Lin , Conghui Huang , Shulong Zhang , Min Xu , Siliang Tao , Shanming Li , Chengchun Zhao , Qiannan Fang , Xisheng Ye , Yin Hang","doi":"10.1016/j.infrared.2025.106360","DOIUrl":"10.1016/j.infrared.2025.106360","url":null,"abstract":"<div><div>The self-Q-switching (SQS) laser performance on Nd-doped crystal at 1.3 μm has been reported for the first time, as far as is known. On Sr<sub>0.7</sub>Nd<sub>0.05</sub>La<sub>0.25</sub>Mg<sub>0.3</sub>Al<sub>11.7</sub>O<sub>19</sub> (Nd:ASL) disorder crystal, a SQS dual-wavelength laser at 1339.9 and 1370.3 nm with output power up to 1.65 W was obtained under an absorbed pump power of 10.13 W with slope and optical-to-optical efficiencies of 22.3 % and 16.3 %, respectively. Furthermore, an on-surface optical axis quartz birefringent filter (BRF) was inserted in the V-folded cavity to tune the laser wavelength. Lasers at 1306.4, and approximately 1340, 1370, or 1391 nm were obtained. The experimental results indicated that σ polarization direction Nd:ASL is capable of producing dual-wavelength lasers at 1339.9 and 1370.3 nm, which was potential to be employed as the source of THz radiation. Besides, Nd:ASL crystals are enable to generate tunable lasers near 1370 and 1391 nm.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106360"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923448","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

DSHANet: Dual-path sampling and hybrid attention network for infrared image destriping DSHANet：用于红外图像去条纹的双路径采样和混合关注网络

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-23 DOI: 10.1016/j.infrared.2026.106421

Xue Li , Hongying Zhang , Lijun Yang , Xi Yang , Song Liu

Infrared images are often severely degraded by stripe noise, which significantly hinders subsequent image analysis and applications. To address the limitations of existing destriping methods in distinguishing noise from image details and modeling cross-scale feature correlations, this paper proposes a dual-path sampling and hybrid attention-based approach for infrared image destriping. The method implicitly splits feature branches through the designed residual dual-path downsampling module. One branch uses adaptive pooling to suppress stripe noise, while the other retains image edge details via grouped strided convolution. These two branches are fused using dynamic weights. Additionally, a hybrid attention module is proposed to separately capture noise patterns and structural features via 1 × 3 convolution and vertical strip attention, respectively, with a self-calibration branch adaptively modulating feature responses to suppress stripe noise while enhancing target integrity. Experiments demonstrate that the proposed method outperforms existing approaches on the INFRARED, ICSRN, CVC09, BSD68, and SIDD benchmark datasets, as well as real data. Specifically, it achieves an average Peak Signal-to-Noise Ratio of 37.96 dB across four typical stripe noise scenarios, surpassing the state-of-the-art method by 0.34 dB while effectively suppressing stripe noise.

红外图像经常受到条纹噪声的严重影响，严重阻碍了后续的图像分析和应用。针对现有去条纹方法在图像细节噪声区分和跨尺度特征相关性建模方面的局限性，提出了一种基于双路径采样和混合注意的红外图像去条纹方法。该方法通过设计的残差双径下采样模块隐式分割特征分支。一个分支使用自适应池来抑制条纹噪声，而另一个分支通过分组跨行卷积来保留图像边缘细节。这两个分支使用动态权值进行融合。此外，提出了一种混合注意模块，分别通过1 × 3卷积和垂直条形注意分别捕获噪声模式和结构特征，并通过自校准分支自适应调制特征响应来抑制条形噪声，同时增强目标完整性。实验表明，该方法在红外、ICSRN、CVC09、BSD68和SIDD基准数据集以及实际数据上都优于现有方法。具体来说，在四种典型条纹噪声情况下，该方法的平均峰值信噪比为37.96 dB，在有效抑制条纹噪声的同时，比目前最先进的方法高出0.34 dB。

{"title":"DSHANet: Dual-path sampling and hybrid attention network for infrared image destriping","authors":"Xue Li , Hongying Zhang , Lijun Yang , Xi Yang , Song Liu","doi":"10.1016/j.infrared.2026.106421","DOIUrl":"10.1016/j.infrared.2026.106421","url":null,"abstract":"<div><div>Infrared images are often severely degraded by stripe noise, which significantly hinders subsequent image analysis and applications. To address the limitations of existing destriping methods in distinguishing noise from image details and modeling cross-scale feature correlations, this paper proposes a dual-path sampling and hybrid attention-based approach for infrared image destriping. The method implicitly splits feature branches through the designed residual dual-path downsampling module. One branch uses adaptive pooling to suppress stripe noise, while the other retains image edge details via grouped strided convolution. These two branches are fused using dynamic weights. Additionally, a hybrid attention module is proposed to separately capture noise patterns and structural features via 1 × 3 convolution and vertical strip attention, respectively, with a self-calibration branch adaptively modulating feature responses to suppress stripe noise while enhancing target integrity. Experiments demonstrate that the proposed method outperforms existing approaches on the INFRARED, ICSRN, CVC09, BSD68, and SIDD benchmark datasets, as well as real data. Specifically, it achieves an average Peak Signal-to-Noise Ratio of 37.96 dB across four typical stripe noise scenarios, surpassing the state-of-the-art method by 0.34 dB while effectively suppressing stripe noise.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106421"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146073973","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Spectral clustering dimensionality reduction in wheat quality detection based on hyperspectral data 基于高光谱数据的小麦品质检测光谱聚类降维

IF 3.4 3区物理与天体物理 Q2 INSTRUMENTS & INSTRUMENTATION

Infrared Physics & Technology

Pub Date : 2026-03-01 Epub Date: 2026-01-14 DOI: 10.1016/j.infrared.2026.106400

Huawei Jiang , Yiduo Zhu , Wanbao Sheng , Ruomeng Hu , Wenqiang Pi , Zhen Yang , Like Zhao

As one of the most important food crops worldwide, the accurate quality detection of wheat is a key link in safeguarding food security and food safety. Hyperspectral technology, as an effective method for quality detection, however, faces challenges in accurately determining critical quality indicators such as wheat deterioration degree due to the presence of massive redundant information. To address this issue, this study proposes a Spectral Clustering Dimensionality Reduction (SCDR) algorithm that integrates spectral angle similarity and spatial distance. First, the differences and similarities among various spectral features are quantitatively analyzed to construct the feature relationships between different bands. Second, based on these feature relationships, high-dimensional features are partitioned via clustering to generate feature clusters with dimensions far lower than those of the original data. Finally, weights are assigned according to the intra-cluster feature differences and similarities to calculate the representative feature values, thereby achieving dimensionality reduction. The experimental results demonstrate that the wheat quality detection model established based on the SCDR algorithm achieves an accuracy, precision, recall and F1-score of 0.9821, 0.9818, 0.9822 and 0.9818, respectively, on the test set, and its performance is significantly superior to that of other comparative models.

小麦作为世界上最重要的粮食作物之一，准确的质量检测是保障粮食安全和食品安全的关键环节。然而，高光谱技术作为一种有效的品质检测方法，由于存在大量冗余信息，在准确确定小麦变质程度等关键品质指标方面面临挑战。为了解决这一问题，本研究提出了一种融合光谱角相似度和空间距离的光谱聚类降维算法。首先，定量分析各种光谱特征之间的异同，构建不同波段之间的特征关系；其次，基于这些特征关系，对高维特征进行聚类分割，生成远低于原始数据维数的特征聚类；最后，根据聚类内特征的相似度和差异分配权重，计算具有代表性的特征值，从而实现降维。实验结果表明，基于SCDR算法建立的小麦品质检测模型在测试集上的准确率、精密度、召回率和f1分数分别为0.9821、0.9818、0.9822和0.9818，其性能明显优于其他比较模型。

{"title":"Spectral clustering dimensionality reduction in wheat quality detection based on hyperspectral data","authors":"Huawei Jiang , Yiduo Zhu , Wanbao Sheng , Ruomeng Hu , Wenqiang Pi , Zhen Yang , Like Zhao","doi":"10.1016/j.infrared.2026.106400","DOIUrl":"10.1016/j.infrared.2026.106400","url":null,"abstract":"<div><div>As one of the most important food crops worldwide, the accurate quality detection of wheat is a key link in safeguarding food security and food safety. Hyperspectral technology, as an effective method for quality detection, however, faces challenges in accurately determining critical quality indicators such as wheat deterioration degree due to the presence of massive redundant information. To address this issue, this study proposes a Spectral Clustering Dimensionality Reduction (SCDR) algorithm that integrates spectral angle similarity and spatial distance. First, the differences and similarities among various spectral features are quantitatively analyzed to construct the feature relationships between different bands. Second, based on these feature relationships, high-dimensional features are partitioned via clustering to generate feature clusters with dimensions far lower than those of the original data. Finally, weights are assigned according to the intra-cluster feature differences and similarities to calculate the representative feature values, thereby achieving dimensionality reduction. The experimental results demonstrate that the wheat quality detection model established based on the SCDR algorithm achieves an accuracy, precision, recall and F1-score of 0.9821, 0.9818, 0.9822 and 0.9818, respectively, on the test set, and its performance is significantly superior to that of other comparative models.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106400"},"PeriodicalIF":3.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146023733","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0