Cross-instrument variability remains a key barrier to the scalable application of near-infrared (NIR) spectroscopy in agri-food quality monitoring. This study introduces two Transformer-based calibration transfer models, Transpec and TPDS, designed to enhance spectral alignment across different instruments.
By combining global attention with localized spectral modeling, the proposed methods reduce reliance on extensive preprocessing and large paired transfer sets. Compared with classical techniques, Transpec and TPDS achieve higher predictive consistency across forward and backward transfers and demonstrate strong performance across multiple flour quality indicators. Their robustness and computational efficiency highlight their potential for real-time deployment in industrial multi-instrument environments. This work establishes a scalable framework for cross-device NIR modeling and contributes to the development of intelligent quality control systems in agricultural processing.
{"title":"Beyond preprocessing and directional bias: Transformer models for robust and efficient cross-instrument NIR calibration in wheat flour analysis","authors":"Jing Liang, Hailong Feng, Yu Xue, Mingyue Huang, Bin Wang, Xiaoxuan Xu, Jing Xu","doi":"10.1016/j.infrared.2026.106448","DOIUrl":"10.1016/j.infrared.2026.106448","url":null,"abstract":"<div><div>Cross-instrument variability remains a key barrier to the scalable application of near-infrared (NIR) spectroscopy in agri-food quality monitoring. This study introduces two Transformer-based calibration transfer models, Transpec and TPDS, designed to enhance spectral alignment across different instruments.</div><div>By combining global attention with localized spectral modeling, the proposed methods reduce reliance on extensive preprocessing and large paired transfer sets. Compared with classical techniques, Transpec and TPDS achieve higher predictive consistency across forward and backward transfers and demonstrate strong performance across multiple flour quality indicators. Their robustness and computational efficiency highlight their potential for real-time deployment in industrial multi-instrument environments. This work establishes a scalable framework for cross-device NIR modeling and contributes to the development of intelligent quality control systems in agricultural processing.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106448"},"PeriodicalIF":3.4,"publicationDate":"2026-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172808","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}
Pub Date : 2026-02-06DOI: 10.1016/j.infrared.2026.106437
Kent Nagumo , Kosuke Oiwa , Akio Nozawa
This study presents a preliminary, exploratory longitudinal analysis of facial skin temperature (FST) spatial distributions using facial thermal imagery (FTI) acquired over six months from summer to winter. We aimed to quantify the typical spatial distribution of FST and assess intra- and inter-individual variability under controlled laboratory conditions. As an initial step toward defining a reference distribution, we compared FST spatial patterns measured during baseline sessions with those obtained under an experimentally induced, non-clinically validated abnormal condition. The results showed relatively small inter-individual variability in FST spatial distributions within the sampled population, suggesting a consistent pattern across participants. In contrast, the abnormal condition produced measurable deviations from the baseline pattern, particularly when distributions were expressed using Z-score normalization. Because this study did not include clinical validation, external control groups, or real-world testing, the findings should be interpreted as suggestive rather than definitive. Future work should include clinical trials and broader participant cohorts to validate the proposed reference distribution, evaluate additional confounders (e.g., circadian effects and environmental variability), and test robustness in real-world settings to support translational applications such as health monitoring.
{"title":"Standard spatial distribution of facial skin temperature: A preliminary study","authors":"Kent Nagumo , Kosuke Oiwa , Akio Nozawa","doi":"10.1016/j.infrared.2026.106437","DOIUrl":"10.1016/j.infrared.2026.106437","url":null,"abstract":"<div><div>This study presents a preliminary, exploratory longitudinal analysis of facial skin temperature (FST) spatial distributions using facial thermal imagery (FTI) acquired over six months from summer to winter. We aimed to quantify the typical spatial distribution of FST and assess intra- and inter-individual variability under controlled laboratory conditions. As an initial step toward defining a reference distribution, we compared FST spatial patterns measured during baseline sessions with those obtained under an experimentally induced, non-clinically validated abnormal condition. The results showed relatively small inter-individual variability in FST spatial distributions within the sampled population, suggesting a consistent pattern across participants. In contrast, the abnormal condition produced measurable deviations from the baseline pattern, particularly when distributions were expressed using Z-score normalization. Because this study did not include clinical validation, external control groups, or real-world testing, the findings should be interpreted as suggestive rather than definitive. Future work should include clinical trials and broader participant cohorts to validate the proposed reference distribution, evaluate additional confounders (e.g., circadian effects and environmental variability), and test robustness in real-world settings to support translational applications such as health monitoring.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106437"},"PeriodicalIF":3.4,"publicationDate":"2026-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146135652","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}
Pub Date : 2026-02-05DOI: 10.1016/j.infrared.2026.106450
Tao Jin, Jianyu Huang
Infrared imaging finds extensive applications in security surveillance, remote sensing, interstellar exploration, and other fields where visible light imaging is limited. However, due to sensor limitations and atmospheric interference, infrared images often suffer from low resolution, severe noise, and poor contrast. To address these challenges, we propose a Memory-Driven Wavelet Network (MDWN) for lightweight infrared image super-resolution. First, we design a Parallel Wavelet Feature Extractor (PWFE) that decomposes input features into multiscale frequency components via wavelet transform, constructing dual path representations that capture complementary low and high frequency details under distinct receptive fields. Second, we propose a Memory-Driven Feature Integration Block (MDFIB), which incorporates a hierarchical memory bank with a progressively increasing number of learnable tokens across network stages. This design enables shallow layers to capture local structural priors, while deeper layers model global semantic representations. The memory tokens act as anchors for cross-region attention, effectively fusing fine-grained local details with long-range contextual information, without resorting to computationally expensive dense pairwise attention. Extensive experiments on multiple infrared benchmark datasets demonstrate that our Memory-Driven Wavelet Network (MDWN) achieves state-of-the-art performance with significantly fewer parameters and lower computational overhead.
{"title":"Memory-Driven Wavelet Network for lightweight infrared image super-resolution","authors":"Tao Jin, Jianyu Huang","doi":"10.1016/j.infrared.2026.106450","DOIUrl":"10.1016/j.infrared.2026.106450","url":null,"abstract":"<div><div>Infrared imaging finds extensive applications in security surveillance, remote sensing, interstellar exploration, and other fields where visible light imaging is limited. However, due to sensor limitations and atmospheric interference, infrared images often suffer from low resolution, severe noise, and poor contrast. To address these challenges, we propose a Memory-Driven Wavelet Network (MDWN) for lightweight infrared image super-resolution. First, we design a Parallel Wavelet Feature Extractor (PWFE) that decomposes input features into multiscale frequency components via wavelet transform, constructing dual path representations that capture complementary low and high frequency details under distinct receptive fields. Second, we propose a Memory-Driven Feature Integration Block (MDFIB), which incorporates a hierarchical memory bank with a progressively increasing number of learnable tokens across network stages. This design enables shallow layers to capture local structural priors, while deeper layers model global semantic representations. The memory tokens act as anchors for cross-region attention, effectively fusing fine-grained local details with long-range contextual information, without resorting to computationally expensive dense pairwise attention. Extensive experiments on multiple infrared benchmark datasets demonstrate that our Memory-Driven Wavelet Network (MDWN) achieves state-of-the-art performance with significantly fewer parameters and lower computational overhead.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106450"},"PeriodicalIF":3.4,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172884","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}
Pub Date : 2026-02-05DOI: 10.1016/j.infrared.2026.106446
Yujun Chen, Peng Cai, Yujia Li, Hang Ming, Minnan Wu, Lei Chen, Ligang Huang, Lei Gao, Tao Zhu
Noise-like pulses (NLPs) with intense randomness play a crucial role in low-coherence spectroscopic measurements, supercontinuum generation, random laser imaging, and chaotic laser sensing. However, systematic analysis addressing both the spectral intensity randomness and the polarization characteristics of NLPs have not yet been explored. Here, we use the dispersive Fourier transform technique and the high-speed wavelength-resolved polarization measurement technique to investigate the spectral randomness and polarization characteristics of NLPs generated by nonlinear polarization rotation mode-locking in a net-normal dispersive cavity with varying lengths of highly nonlinear fiber (HNLF). Through experimental and statistical methods (correlation, Pearson coefficient, mutual information), we demonstrated that HNLF enhances spectral intensity randomness in NLPs. The incorporation of HNLF not only increases the spectral correlation decay rate but also reduces the full width at half maximum of the Pearson coefficient curve corresponding to the peak wavelength from 0.38 nm to below 0.1 nm. Accompanied by the polarization filtering effects of nonlinear polarization rotation, the wavelength-resolved states of polarization (SOP) for the NLPs exhibit partial randomness on the Poincaré sphere. We quantify polarization distribution characteristics and randomness using the relative distance (r) of SOP projection points and approximate entropy (ApEn). HNLF enhances both polarization distribution range and randomness, increasing r_max from 2 to 3 and improving ApEn values for polar/azimuthal angles at different wavelengths. The simulation result is consistent with experiments. Our work provides a systematic routine for investigating the randomness in NLPs, and also offers a new approach for generating low-cost, highly random ultrafast light sources.
{"title":"Randomness enhancement of noise-like pulses in nonlinear polarization rotation fiber cavity","authors":"Yujun Chen, Peng Cai, Yujia Li, Hang Ming, Minnan Wu, Lei Chen, Ligang Huang, Lei Gao, Tao Zhu","doi":"10.1016/j.infrared.2026.106446","DOIUrl":"10.1016/j.infrared.2026.106446","url":null,"abstract":"<div><div>Noise-like pulses (NLPs) with intense randomness play a crucial role in low-coherence spectroscopic measurements, supercontinuum generation, random laser imaging, and chaotic laser sensing. However, systematic analysis addressing both the spectral intensity randomness and the polarization characteristics of NLPs have not yet been explored. Here, we use the dispersive Fourier transform technique and the high-speed wavelength-resolved polarization measurement technique to investigate the spectral randomness and polarization characteristics of NLPs generated by nonlinear polarization rotation mode-locking in a net-normal dispersive cavity with varying lengths of highly nonlinear fiber (HNLF). Through experimental and statistical methods (correlation, Pearson coefficient, mutual information), we demonstrated that HNLF enhances spectral intensity randomness in NLPs. The incorporation of HNLF not only increases the spectral correlation decay rate but also reduces the full width at half maximum of the Pearson coefficient curve corresponding to the peak wavelength from 0.38 nm to below 0.1 nm. Accompanied by the polarization filtering effects of nonlinear polarization rotation, the wavelength-resolved states of polarization (SOP) for the NLPs exhibit partial randomness on the Poincaré sphere. We quantify polarization distribution characteristics and randomness using the relative distance (<em>r</em>) of SOP projection points and approximate entropy (ApEn). HNLF enhances both polarization distribution range and randomness, increasing <em>r_max</em> from 2 to 3 and improving ApEn values for polar/azimuthal angles at different wavelengths. The simulation result is consistent with experiments. Our work provides a systematic routine for investigating the randomness in NLPs, and also offers a new approach for generating low-cost, highly random ultrafast light sources.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106446"},"PeriodicalIF":3.4,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172813","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}
Pub Date : 2026-02-03DOI: 10.1016/j.infrared.2026.106445
Hang Ren , Ying Yang , He Cao , Zhan Hong Lip , Fanlong Dong , Jun Guo , Jiachen Wang , Jinzhang Wang , Tengfei Wu , Chunyu Guo , Shuangchen Ruan
In this paper, to the best of our knowledge, we demonstrate the generation of a dual-wavelength pulse from a mid-infrared spatiotemporally mode-locking (STML) large-mode-area Er: ZBLAN fiber laser based on the nonlinear polarization rotation technology for the first time. Under a pump power of 4.04 W, a dual-wavelength spectrum with a pulse duration of 42 ps is achieved, centered at 2792 nm and 2796 nm, and delivering a pulse energy of 9.94 nJ. Here, we conducted numerical simulations by solving the generalized multimode nonlinear Schrödinger equation. The experimental and simulation results indicate that both long-wavelength and short-wavelength components can independently achieve STML operation, thereby confirming the realization of dual-wavelength STML operation. This work can enhance the understanding of pulse dynamics in mid-infrared multi-wavelength STML fiber lasers.
{"title":"Dual-wavelength pulse generation from a 2.8 μm spatiotemporally mode-locking fiber laser based on nonlinear polarization rotation","authors":"Hang Ren , Ying Yang , He Cao , Zhan Hong Lip , Fanlong Dong , Jun Guo , Jiachen Wang , Jinzhang Wang , Tengfei Wu , Chunyu Guo , Shuangchen Ruan","doi":"10.1016/j.infrared.2026.106445","DOIUrl":"10.1016/j.infrared.2026.106445","url":null,"abstract":"<div><div>In this paper, to the best of our knowledge, we demonstrate the generation of a dual-wavelength pulse from a mid-infrared spatiotemporally mode-locking (STML) large-mode-area Er: ZBLAN fiber laser based on the nonlinear polarization rotation technology for the first time. Under a pump power of 4.04 W, a dual-wavelength spectrum with a pulse duration of 42 ps is achieved, centered at 2792 nm and 2796 nm, and delivering a pulse energy of 9.94 nJ. Here, we conducted numerical simulations by solving the generalized multimode nonlinear Schrödinger equation. The experimental and simulation results indicate that both long-wavelength and short-wavelength components can independently achieve STML operation, thereby confirming the realization of dual-wavelength STML operation. This work can enhance the understanding of pulse dynamics in mid-infrared multi-wavelength STML fiber lasers.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106445"},"PeriodicalIF":3.4,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172810","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}
Pub Date : 2026-02-03DOI: 10.1016/j.infrared.2026.106431
Kanglin Jin, Mengtong Guo, Minghao Piao
Infrared and visible image fusion aims to generate a fused image that highlights salient targets while preserving fine textures. Existing deep learning-based methods predominantly rely on spatial-domain representations, which fail to fully capture the modality-specific frequency characteristics, leading to suboptimal texture preservation and detail enhancement. Since infrared and visible images exhibit distinct frequency distributions, relying solely on spatial-domain methods is insufficient for achieving high-quality fusion. To overcome this limitation, we propose a novel High-order Spatial-Frequency Interaction and Detail Compensation Network (HSFIDCNet), which jointly exploits spatial and frequency representations for more effective feature fusion. Specifically, the High-order Spatial-Frequency Interaction (HSFI) module enhances cross-domain feature integration, achieving a balanced fusion of global structures and local details, while the Detail Compensation (DC) module strengthens texture representation and highlights salient objects. Extensive experiments on three benchmark datasets (M3FD, LLVIP, and MSRS) against twelve state-of-the-art methods demonstrate that our approach consistently outperforms existing methods, producing fused images with higher contrast and richer textures. In particular, our method achieves the best performance across all three datasets in CC (0.5298, 0.7134, 0.6180), (0.7102, 0.7326, 0.7025), MS-SSIM (0.9573, 0.9696, 0.9778), and (478.6155, 267.7829, 203.6782), highlighting its robust and generalizable fusion capability. Code is available at https://github.com/sdat-max/HSFIDCNet.
{"title":"High-order Spatial-Frequency Interaction and Detail Compensation Network for infrared and visible image fusion","authors":"Kanglin Jin, Mengtong Guo, Minghao Piao","doi":"10.1016/j.infrared.2026.106431","DOIUrl":"10.1016/j.infrared.2026.106431","url":null,"abstract":"<div><div>Infrared and visible image fusion aims to generate a fused image that highlights salient targets while preserving fine textures. Existing deep learning-based methods predominantly rely on spatial-domain representations, which fail to fully capture the modality-specific frequency characteristics, leading to suboptimal texture preservation and detail enhancement. Since infrared and visible images exhibit distinct frequency distributions, relying solely on spatial-domain methods is insufficient for achieving high-quality fusion. To overcome this limitation, we propose a novel High-order Spatial-Frequency Interaction and Detail Compensation Network (HSFIDCNet), which jointly exploits spatial and frequency representations for more effective feature fusion. Specifically, the High-order Spatial-Frequency Interaction (HSFI) module enhances cross-domain feature integration, achieving a balanced fusion of global structures and local details, while the Detail Compensation (DC) module strengthens texture representation and highlights salient objects. Extensive experiments on three benchmark datasets (M<sup>3</sup>FD, LLVIP, and MSRS) against twelve state-of-the-art methods demonstrate that our approach consistently outperforms existing methods, producing fused images with higher contrast and richer textures. In particular, our method achieves the best performance across all three datasets in CC (0.5298, 0.7134, 0.6180), <span><math><msup><mrow><mi>Q</mi></mrow><mrow><mi>A</mi><mi>B</mi><mo>/</mo><mi>F</mi></mrow></msup></math></span> (0.7102, 0.7326, 0.7025), MS-SSIM (0.9573, 0.9696, 0.9778), and <span><math><msub><mrow><mi>Q</mi></mrow><mrow><mi>C</mi><mi>V</mi></mrow></msub></math></span> (478.6155, 267.7829, 203.6782), highlighting its robust and generalizable fusion capability. Code is available at <span><span>https://github.com/sdat-max/HSFIDCNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106431"},"PeriodicalIF":3.4,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172883","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}
Pub Date : 2026-02-02DOI: 10.1016/j.infrared.2026.106442
Hao Song , Longkun He , Shan Zhao, Lijin Gao, Chengjiang Zhou
The infrared and visible image fusion is a technique that extracts and combines information from both images to enhance image quality and target recognition. The image fusion serves the high-level vision task. Therefore, existing fusion algorithms prioritize the prerequisites of these tasks, making it difficult to balance visual quality with task performance. Meanwhile, since the distribution of features is disordered and contains redundant information, the ability to directly capture features without constraints and targeting is limited. To address these issues, we propose an infrared and visible image fusion network based on multi-frequency feature decomposition, termed MFDFuse. Specifically, we build a channel-based collaborative feature extraction branch in the encoder to continuously capture and convey the correlations between channels, thus maintaining the coherence of channel modeling across the network to preserve the channel information of the source image. Secondly, we utilize element-wise multiplication-based single-level and multi-level mappings to integrate high-frequency information and semantic information, achieving a balance between detail preservation and semantic enhancement. Then, we propose a feature separation loss based on feature distance to separate low-high frequency features of different modalities to constrain the results of feature extraction. Finally, we fully consider the feature information lost during the feature extraction process to achieve effective compensation. The experiments performed on publicly accessible datasets, along with downstream tasks, have shown that MFDFuse surpasses state-of-the-art (SOTA) methods in both quantitative and qualitative analysis. The code is available at https://github.com/SunsHine0816/MFDFuse.
{"title":"MFDFuse: A multi-frequency decomposition-based network for infrared and visible image fusion","authors":"Hao Song , Longkun He , Shan Zhao, Lijin Gao, Chengjiang Zhou","doi":"10.1016/j.infrared.2026.106442","DOIUrl":"10.1016/j.infrared.2026.106442","url":null,"abstract":"<div><div>The infrared and visible image fusion is a technique that extracts and combines information from both images to enhance image quality and target recognition. The image fusion serves the high-level vision task. Therefore, existing fusion algorithms prioritize the prerequisites of these tasks, making it difficult to balance visual quality with task performance. Meanwhile, since the distribution of features is disordered and contains redundant information, the ability to directly capture features without constraints and targeting is limited. To address these issues, we propose an infrared and visible image fusion network based on multi-frequency feature decomposition, termed MFDFuse. Specifically, we build a channel-based collaborative feature extraction branch in the encoder to continuously capture and convey the correlations between channels, thus maintaining the coherence of channel modeling across the network to preserve the channel information of the source image. Secondly, we utilize element-wise multiplication-based single-level and multi-level mappings to integrate high-frequency information and semantic information, achieving a balance between detail preservation and semantic enhancement. Then, we propose a feature separation loss based on feature distance to separate low-high frequency features of different modalities to constrain the results of feature extraction. Finally, we fully consider the feature information lost during the feature extraction process to achieve effective compensation. The experiments performed on publicly accessible datasets, along with downstream tasks, have shown that MFDFuse surpasses state-of-the-art (SOTA) methods in both quantitative and qualitative analysis. The code is available at <span><span>https://github.com/SunsHine0816/MFDFuse.</span><svg><path></path></svg></span></div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106442"},"PeriodicalIF":3.4,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172873","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}
Pub Date : 2026-01-31DOI: 10.1016/j.infrared.2026.106440
Jun Huang, Yichao Liu, Zefeng Guo, Tianyu Ren, Yujie Ji, Chengkun Cai, Hua Guan
This study systematically investigated the influence of different oxygen-containing oxidants (NaNO3, KNO3, Ba(NO3) 2) on the combustion performance and radiation characteristics of MgB2/PTFE infrared radiation agents under low-pressure conditions. Through thermal analysis (TG-DTA), combustion product characterization (XRD, SEM), and low-pressure combustion experiments (5–––101 kPa), it was found that the decomposition temperature of the oxidants significantly affected the combustion stability and reaction pathway of the agent in the low-pressure environment. NaNO3 and KNO3, due to their lower decomposition temperatures (374.5 ℃ and 408 ℃ respectively), could still promote the efficient reaction of MgB2 at 5 kPa low pressure, significantly improving the combustion stability and radiation area; while Ba(NO3) 2 had a higher decomposition temperature (577.8 ℃), its system could not burn stably at 5 kPa, but by generating high-emissivity condensed-phase products (such as BaO, BaB6), it effectively enhanced the infrared radiation intensity in the α (1.3–––3 μm), β (3–––5 μm), and γ (8–––14 μm) wavelength bands under low-pressure conditions. The study further revealed the reaction mechanisms of each system: NaNO3/KNO3 mainly produced alkali metal borates (such as Na2B4O7·H2O, K3.67B4O5 (OH) 5), while the Ba(NO3) 2 system formed BaB6 through complex phase changes. The results showed that by regulating the type of oxidant, the combustion efficiency and radiation performance of MgB2/PTFE agents under low pressure could be optimized, providing key theoretical and experimental basis for the design of high-altitude infrared decoy agents.
{"title":"Study on the influence Laws of different Oxygen-Containing oxidants on the Low-Pressure combustion and radiation performance of MgB2/PTFE type infrared radiation agents","authors":"Jun Huang, Yichao Liu, Zefeng Guo, Tianyu Ren, Yujie Ji, Chengkun Cai, Hua Guan","doi":"10.1016/j.infrared.2026.106440","DOIUrl":"10.1016/j.infrared.2026.106440","url":null,"abstract":"<div><div>This study systematically investigated the influence of different oxygen-containing oxidants (NaNO<sub>3</sub>, KNO<sub>3</sub>, Ba(NO<sub>3</sub>) <sub>2</sub>) on the combustion performance and radiation characteristics of MgB<sub>2</sub>/PTFE infrared radiation agents under low-pressure conditions. Through thermal analysis (TG-DTA), combustion product characterization (XRD, SEM), and low-pressure combustion experiments (5–––101 kPa), it was found that the decomposition temperature of the oxidants significantly affected the combustion stability and reaction pathway of the agent in the low-pressure environment. NaNO<sub>3</sub> and KNO<sub>3</sub>, due to their lower decomposition temperatures (374.5 ℃ and 408 ℃ respectively), could still promote the efficient reaction of MgB<sub>2</sub> at 5 kPa low pressure, significantly improving the combustion stability and radiation area; while Ba(NO<sub>3</sub>) <sub>2</sub> had a higher decomposition temperature (577.8 ℃), its system could not burn stably at 5 kPa, but by generating high-emissivity condensed-phase products (such as BaO, BaB<sub>6</sub>), it effectively enhanced the infrared radiation intensity in the α (1.3–––3 μm), β (3–––5 μm), and γ (8–––14 μm) wavelength bands under low-pressure conditions. The study further revealed the reaction mechanisms of each system: NaNO<sub>3</sub>/KNO<sub>3</sub> mainly produced alkali metal borates (such as Na<sub>2</sub>B<sub>4</sub>O<sub>7</sub>·H<sub>2</sub>O, K<sub>3.67</sub>B<sub>4</sub>O<sub>5</sub> (OH) <sub>5</sub>), while the Ba(NO<sub>3</sub>) <sub>2</sub> system formed BaB<sub>6</sub> through complex phase changes. The results showed that by regulating the type of oxidant, the combustion efficiency and radiation performance of MgB<sub>2</sub>/PTFE agents under low pressure could be optimized, providing key theoretical and experimental basis for the design of high-altitude infrared decoy agents.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106440"},"PeriodicalIF":3.4,"publicationDate":"2026-01-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172875","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}
Pub Date : 2026-01-30DOI: 10.1016/j.infrared.2026.106426
Xingwei Yan , Kun Liu , Ji Li , Yan Zhang , Yaxiu Zhang , Chenchen Zhang
With the rapid proliferation of unmanned aerial vehicles, the issue of their security has gradually become a focal point of research. In infrared target detection tasks, due to the small target size, complex backgrounds, and low contrast, existing methods often rely solely on the internal features of a single modality, lacking the ability to interact with external information, which limits detection performance. To address this issue, this paper proposes a novel multi-modal image detection method, R2TNet, which can directly process misaligned RGB-T images, effectively avoiding the complexity of traditional manual registration. To achieve efficient modality alignment and fusion, this paper designs a supervised bottom-up multimodal alignment module, which adopts a coarse-to-fine layer-wise registration strategy. This effectively alleviates the modality misalignment issue in multimodal images, thereby achieving precise alignment between RGB and infrared features. On this basis, a semantic-guided module is further employed to optimize cross-modal feature fusion using high-level semantic information, significantly improving the accuracy and robustness of target detection. At the same time, a multi-scale gated dynamic fusion module is incorporated to realize fine-grained fusion of multimodal features, further enhancing the model’s adaptability in complex scenarios. Experimental results demonstrate that the proposed R2TNet significantly outperforms existing state-of-the-art bimodal detection methods across multiple evaluation metrics, including Em, Sm, Fm, and MAE, and exhibits stronger robustness and generalization capability in complex backgrounds and small target detection tasks. Moreover, comparative results with unimodal infrared detection methods further validate the advantages of the proposed method in cross-modal fusion detection.
{"title":"Drone detection network based on RGB-thermal imaging multimodal fusion","authors":"Xingwei Yan , Kun Liu , Ji Li , Yan Zhang , Yaxiu Zhang , Chenchen Zhang","doi":"10.1016/j.infrared.2026.106426","DOIUrl":"10.1016/j.infrared.2026.106426","url":null,"abstract":"<div><div>With the rapid proliferation of unmanned aerial vehicles, the issue of their security has gradually become a focal point of research. In infrared target detection tasks, due to the small target size, complex backgrounds, and low contrast, existing methods often rely solely on the internal features of a single modality, lacking the ability to interact with external information, which limits detection performance. To address this issue, this paper proposes a novel multi-modal image detection method, R2TNet, which can directly process misaligned RGB-T images, effectively avoiding the complexity of traditional manual registration. To achieve efficient modality alignment and fusion, this paper designs a supervised bottom-up multimodal alignment module, which adopts a coarse-to-fine layer-wise registration strategy. This effectively alleviates the modality misalignment issue in multimodal images, thereby achieving precise alignment between RGB and infrared features. On this basis, a semantic-guided module is further employed to optimize cross-modal feature fusion using high-level semantic information, significantly improving the accuracy and robustness of target detection. At the same time, a multi-scale gated dynamic fusion module is incorporated to realize fine-grained fusion of multimodal features, further enhancing the model’s adaptability in complex scenarios. Experimental results demonstrate that the proposed R2TNet significantly outperforms existing state-of-the-art bimodal detection methods across multiple evaluation metrics, including Em, Sm, Fm, and MAE, and exhibits stronger robustness and generalization capability in complex backgrounds and small target detection tasks. Moreover, comparative results with unimodal infrared detection methods further validate the advantages of the proposed method in cross-modal fusion detection.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"155 ","pages":"Article 106426"},"PeriodicalIF":3.4,"publicationDate":"2026-01-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146172874","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}
Pub Date : 2026-01-29DOI: 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.
{"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":"2026-01-29","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号