Infrared Physics & Technology最新文献_第7页

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 : 2026-01-05 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":"2026-01-05","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

Application of the GWO-RBFNN algorithm for data fitting in a sapphire fiber temperature measurement system GWO-RBFNN算法在蓝宝石光纤测温系统数据拟合中的应用

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

Infrared Physics & Technology

Pub Date : 2026-01-05 DOI: 10.1016/j.infrared.2026.106373

Wenjie Zhao , Sen Yang , Haoyu Wu , Jingmin Dai

In the process of high-temperature kiln ironmaking, molten iron temperature is an important parameter affecting yield and safety, and real-time accurate monitoring is crucial. However, the complexity of kiln conditions and serious environmental interference lead to the low accuracy of traditional temperature measurement, which is difficult to meet the demand for accurate control. To solve this problem, this paper proposes a data fitting method using the Gray Wolf Optimization (GWO) algorithm to optimize the Radial Basis Neural Network (RBFNN) based on the sapphire fiber optic temperature measurement system in order to improve the temperature measurement accuracy. The gray wolf algorithm optimizes the radial basis neural network center, width and weights by simulating the hunting behavior of gray wolves, and adaptively adjusts the parameters with the goal of minimizing the prediction error, which avoids the overfitting of the neural network and improves the model accuracy. The experimental results show that compared with the traditional algorithm, the proposed GWO-RBFNN method reduces the RMSE of iron water temperature prediction by 64%, MAE by 73%, and R² is improved to 0.9992, which further improves the prediction accuracy and training stability.

在高温窑炉炼铁过程中，铁水温度是影响产量和安全的重要参数，实时准确的监测至关重要。然而，由于窑炉条件复杂，环境干扰严重，传统的测温方法精度较低，难以满足精确控制的需求。针对这一问题，本文提出了一种利用灰狼优化（GWO）算法对基于蓝宝石光纤测温系统的径向基神经网络（RBFNN）进行数据拟合的方法，以提高测温精度。灰狼算法通过模拟灰狼的狩猎行为，对径向基神经网络的中心、宽度和权值进行优化，并以预测误差最小为目标自适应调整参数，避免了神经网络的过拟合，提高了模型精度。实验结果表明，与传统算法相比，本文提出的GWO-RBFNN方法将铁水温预测的RMSE降低64%，MAE降低73%，R2提高到0.9992，进一步提高了预测精度和训练稳定性。

{"title":"Application of the GWO-RBFNN algorithm for data fitting in a sapphire fiber temperature measurement system","authors":"Wenjie Zhao , Sen Yang , Haoyu Wu , Jingmin Dai","doi":"10.1016/j.infrared.2026.106373","DOIUrl":"10.1016/j.infrared.2026.106373","url":null,"abstract":"<div><div>In the process of high-temperature kiln ironmaking, molten iron temperature is an important parameter affecting yield and safety, and real-time accurate monitoring is crucial. However, the complexity of kiln conditions and serious environmental interference lead to the low accuracy of traditional temperature measurement, which is difficult to meet the demand for accurate control. To solve this problem, this paper proposes a data fitting method using the Gray Wolf Optimization (GWO) algorithm to optimize the Radial Basis Neural Network (RBFNN) based on the sapphire fiber optic temperature measurement system in order to improve the temperature measurement accuracy. The gray wolf algorithm optimizes the radial basis neural network center, width and weights by simulating the hunting behavior of gray wolves, and adaptively adjusts the parameters with the goal of minimizing the prediction error, which avoids the overfitting of the neural network and improves the model accuracy. The experimental results show that compared with the traditional algorithm, the proposed GWO-RBFNN method reduces the RMSE of iron water temperature prediction by 64%, MAE by 73%, and R<sup>2</sup> is improved to 0.9992, which further improves the prediction accuracy and training stability.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106373"},"PeriodicalIF":3.4,"publicationDate":"2026-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146023729","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

IPCNet: An interactive parallel convolutional network for wheat variety recognition in hyperspectral depth images IPCNet：用于高光谱深度图像小麦品种识别的交互式并行卷积网络

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

Infrared Physics & Technology

Pub Date : 2026-01-05 DOI: 10.1016/j.infrared.2026.106372

Dongfu Xu , Lei Du , Wei Zhang , Pu Liu , Xiaonan Wang , Yizhou Chen , Lixin Qiao , Tie Shi , Jingjing Liu

The identification of wheat varieties based on hyperspectral images is of significant importance in wheat seed breeding. However, the full utilization of the complex information in hyperspectral images presents challenges, and existing feature extraction methods have not fully analyzed them. To address this issue, this paper proposes a wheat seed classification method that integrates hyperspectral imaging technology and an interactive parallel convolutional network (IPCNet). This method builds multiple parallel feature extractors using cross-stacked minimal convolution units(MCU) and multilayer perceptron(MLP). These components jointly process multi-band data to perform multi-level global–local spatial–spectral feature learning in hyperspectral images. Hyperspectral image data for 6 wheat varieties, totaling 720 seeds, had been acquired in the experiment with sufficient spectral band information preserved. Individual wheat seeds are segmented using morphological thresholding, and recognition is achieved through IPCNet with over 95%. 5-fold cross-validation validates the performance of IPCNet, with the model achieving over 97% accuracy on both the training and validation sets. The results demonstrate that IPCNet overcomes the limitations of existing methods for processing high-dimensional hyperspectral image data of wheat seeds and excels in recognizing high-dimensional, complex wheat hyperspectral images. It provides significant technical support for seed variety identification using hyperspectral high-dimensional images in agricultural production.

基于高光谱图像的小麦品种识别在小麦育种中具有重要意义。然而，充分利用高光谱图像中的复杂信息是一个挑战，现有的特征提取方法并没有充分分析高光谱图像。为了解决这一问题，本文提出了一种结合高光谱成像技术和交互式并行卷积网络（IPCNet）的小麦种子分类方法。该方法利用交叉堆叠最小卷积单元（MCU）和多层感知器（MLP）构建多个并行特征提取器。这些组件共同处理多波段数据，在高光谱图像中执行多级全局-局部空间光谱特征学习。实验获得了6个小麦品种共720粒种子的高光谱图像数据，保留了足够的光谱波段信息。采用形态阈值分割方法对小麦种子进行分割，IPCNet的识别率在95%以上。5倍交叉验证验证了IPCNet的性能，该模型在训练集和验证集上的准确率均超过97%。结果表明，IPCNet克服了现有小麦种子高维高光谱图像数据处理方法的局限性，在识别高维复杂小麦高光谱图像方面表现优异。为农业生产中利用高光谱高维图像进行种子品种鉴定提供了重要的技术支持。

{"title":"IPCNet: An interactive parallel convolutional network for wheat variety recognition in hyperspectral depth images","authors":"Dongfu Xu , Lei Du , Wei Zhang , Pu Liu , Xiaonan Wang , Yizhou Chen , Lixin Qiao , Tie Shi , Jingjing Liu","doi":"10.1016/j.infrared.2026.106372","DOIUrl":"10.1016/j.infrared.2026.106372","url":null,"abstract":"<div><div>The identification of wheat varieties based on hyperspectral images is of significant importance in wheat seed breeding. However, the full utilization of the complex information in hyperspectral images presents challenges, and existing feature extraction methods have not fully analyzed them. To address this issue, this paper proposes a wheat seed classification method that integrates hyperspectral imaging technology and an interactive parallel convolutional network (IPCNet). This method builds multiple parallel feature extractors using cross-stacked minimal convolution units(MCU) and multilayer perceptron(MLP). These components jointly process multi-band data to perform multi-level global–local spatial–spectral feature learning in hyperspectral images. Hyperspectral image data for 6 wheat varieties, totaling 720 seeds, had been acquired in the experiment with sufficient spectral band information preserved. Individual wheat seeds are segmented using morphological thresholding, and recognition is achieved through IPCNet with over 95%. 5-fold cross-validation validates the performance of IPCNet, with the model achieving over 97% accuracy on both the training and validation sets. The results demonstrate that IPCNet overcomes the limitations of existing methods for processing high-dimensional hyperspectral image data of wheat seeds and excels in recognizing high-dimensional, complex wheat hyperspectral images. It provides significant technical support for seed variety identification using hyperspectral high-dimensional images in agricultural production.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106372"},"PeriodicalIF":3.4,"publicationDate":"2026-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923445","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

Edge-Guided dynamic diffused fusion for infrared and visible images 红外和可见光图像边缘制导动态扩散融合

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

Infrared Physics & Technology

Pub Date : 2026-01-05 DOI: 10.1016/j.infrared.2026.106370

Longlong Liu , Qimin Yang , Kan Ren, Qian Chen

Image fusion aims to integrate complementary features from multi-modal images into a single, more informative composite. To generate a fused image that simultaneously preserves salient targets from infrared images and rich textural details from visible images, we introduce a novel fusion method: Edge-Guided Dynamic Diffused Fusion (EGD²-Fusion), designed to enhance image quality and optimize downstream tasks like object detection. Addressing the common issue of edge degradation in existing fusion algorithms, EGD²-Fusion incorporates an adaptive weight allocation module that dynamically adjusts multi-scale, multi-timestep features during the reverse diffusion process. Furthermore, we introduce an edge-supervised loss function for auxiliary training, ensuring the fused image effectively retains sharp edge information from the source images while preserving target color and texture. Extensive experiments on public datasets demonstrate that our proposed method surpasses various state-of-the-art methods across multiple quantitative metrics, promoting advanced computer vision tasks.

图像融合的目的是将多模态图像的互补特征整合到一个单一的、信息更丰富的合成图像中。为了生成融合图像，同时保留红外图像中的显著目标和可见光图像中的丰富纹理细节，我们引入了一种新的融合方法：边缘引导动态扩散融合（EGD2-Fusion），旨在提高图像质量并优化下游任务，如目标检测。为了解决现有融合算法中常见的边缘退化问题，EGD2-Fusion集成了一个自适应权重分配模块，该模块可以在反向扩散过程中动态调整多尺度、多时间步特征。此外，我们引入边缘监督损失函数进行辅助训练，确保融合图像在保留目标颜色和纹理的同时有效地保留源图像的锐利边缘信息。在公共数据集上进行的大量实验表明，我们提出的方法在多个定量指标上超越了各种最先进的方法，促进了先进的计算机视觉任务。

{"title":"Edge-Guided dynamic diffused fusion for infrared and visible images","authors":"Longlong Liu , Qimin Yang , Kan Ren, Qian Chen","doi":"10.1016/j.infrared.2026.106370","DOIUrl":"10.1016/j.infrared.2026.106370","url":null,"abstract":"<div><div>Image fusion aims to integrate complementary features from multi-modal images into a single, more informative composite. To generate a fused image that simultaneously preserves salient targets from infrared images and rich textural details from visible images, we introduce a novel fusion method: Edge-Guided Dynamic Diffused Fusion (EGD<sup>2</sup>-Fusion), designed to enhance image quality and optimize downstream tasks like object detection. Addressing the common issue of edge degradation in existing fusion algorithms, EGD<sup>2</sup>-Fusion incorporates an adaptive weight allocation module that dynamically adjusts multi-scale, multi-timestep features during the reverse diffusion process. Furthermore, we introduce an edge-supervised loss function for auxiliary training, ensuring the fused image effectively retains sharp edge information from the source images while preserving target color and texture. Extensive experiments on public datasets demonstrate that our proposed method surpasses various state-of-the-art methods across multiple quantitative metrics, promoting advanced computer vision tasks.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106370"},"PeriodicalIF":3.4,"publicationDate":"2026-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923451","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

YOLO-TFFM: Dual-domain filter fusion network for small target detection in cluttered infrared scenes YOLO-TFFM：用于杂乱红外场景小目标检测的双域滤波融合网络

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

Infrared Physics & Technology

Pub Date : 2026-01-05 DOI: 10.1016/j.infrared.2026.106369

Ke Xu , Yiqing Li , Wen Zhou

In scenarios characterized by cluttered background and target sparsity, existing infrared small target detection algorithms mainly focus on spatial features that cannot handle challenges such as significant noise interference and weak feature expression. Taking inspiration from the time-frequency domain properties that distinguish objects from backgrounds in infrared images,this paper proposes a dual-input detection method named YOLO-TFFM, which is built on the YOLOv11n model. This approach enhances the visibility of small targets and reduces background interference through time-frequency domain convolutional filtering, thereby improving the original object detection framework. Furthermore, an innovative IR-Filter Fusion (IFF) module is designed to integrate filtered image information, effectively enhancing detection accuracy. In the object detection module, an Edge Information Enhancement Module (EIEM) is incorporated to improve the C3k2 and introduce a small target detection layer, thereby augmenting the network’s capacity to extract small target features. To further enhance localization accuracy,a CSP Enhanced Dual-Layer Aggregator (CEDA) module is introduced to reinforce the model’s feature aggregation capability. Regarding loss function design, a joint optimization of Shape-NWD and CIoU is proposed for localization regression, effectively mitigating the sensitivity of traditional IoU to deviations. Extensive experimental results on two public datasets demonstrate that the proposed YOLO-TFFM outperforms mainstream models in overall performance while also exhibiting strong robustness and superior detection capabilities.

在背景杂乱、目标稀疏的场景下，现有红外小目标检测算法主要关注空间特征，无法应对噪声干扰大、特征表达弱等挑战。本文从红外图像中区分目标和背景的时频域特性中获得灵感，提出了一种基于YOLOv11n模型的双输入检测方法YOLO-TFFM。该方法通过时频域卷积滤波增强了小目标的可见性，减少了背景干扰，从而改进了原有的目标检测框架。此外，还设计了一种创新的红外滤波器融合（IFF）模块，用于集成滤波后的图像信息，有效提高检测精度。在目标检测模块中，加入边缘信息增强模块（Edge Information Enhancement module， EIEM）对C3k2进行改进，引入小目标检测层，增强了网络对小目标特征的提取能力。为了进一步提高定位精度，引入CSP增强双层聚合器（Enhanced Dual-Layer Aggregator， CEDA）模块来增强模型的特征聚合能力。在损失函数设计方面，提出了Shape-NWD与CIoU联合优化的局部回归方法，有效降低了传统IoU对偏差的敏感性。在两个公开数据集上的大量实验结果表明，所提出的YOLO-TFFM在整体性能上优于主流模型，同时也表现出较强的鲁棒性和优越的检测能力。

{"title":"YOLO-TFFM: Dual-domain filter fusion network for small target detection in cluttered infrared scenes","authors":"Ke Xu , Yiqing Li , Wen Zhou","doi":"10.1016/j.infrared.2026.106369","DOIUrl":"10.1016/j.infrared.2026.106369","url":null,"abstract":"<div><div>In scenarios characterized by cluttered background and target sparsity, existing infrared small target detection algorithms mainly focus on spatial features that cannot handle challenges such as significant noise interference and weak feature expression. Taking inspiration from the time-frequency domain properties that distinguish objects from backgrounds in infrared images,this paper proposes a dual-input detection method named YOLO-TFFM, which is built on the YOLOv11n model. This approach enhances the visibility of small targets and reduces background interference through time-frequency domain convolutional filtering, thereby improving the original object detection framework. Furthermore, an innovative IR-Filter Fusion (IFF) module is designed to integrate filtered image information, effectively enhancing detection accuracy. In the object detection module, an Edge Information Enhancement Module (EIEM) is incorporated to improve the C3k2 and introduce a small target detection layer, thereby augmenting the network’s capacity to extract small target features. To further enhance localization accuracy,a CSP Enhanced Dual-Layer Aggregator (CEDA) module is introduced to reinforce the model’s feature aggregation capability. Regarding loss function design, a joint optimization of Shape-NWD and CIoU is proposed for localization regression, effectively mitigating the sensitivity of traditional IoU to deviations. Extensive experimental results on two public datasets demonstrate that the proposed YOLO-TFFM outperforms mainstream models in overall performance while also exhibiting strong robustness and superior detection capabilities.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106369"},"PeriodicalIF":3.4,"publicationDate":"2026-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923452","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

Defect detection for micro-LED by multi-illumination and deep learning 基于多点照明和深度学习的微型led缺陷检测

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

Infrared Physics & Technology

Pub Date : 2026-01-03 DOI: 10.1016/j.infrared.2025.106357

Qicheng Lin , Zhiguo Xie , Kiyoshi Takamasu , Meiyun Chen

Micro-LED wafer defect detection faces challenges such as weak optical contrast, diverse defect types, and stringent real-time requirements. To tackle these issues, this paper proposes the Multi-illumination Movable Detection System (MIMDS) and DMI-Net. MIMDS uses multi-angle illumination and multi-spectral fusion for higher-quality defect images. DMI-Net integrates three core innovations: Dual Receptive-field Attention Convolution (DRAConv) module that enhances feature extraction while reducing computational cost through grouped and segmented convolution; Mixed Local Channel Attention RepNBottleneck (MRB) that improves small defect recognition via multi-scale feature refinement; and Inner Focaler IoU (IF-IoU) loss function that adapts to sample difficulty for faster convergence and better accuracy. Experiments on VOC2012 and Micro-LED wafer datasets show that DMI-Net achieves mAP scores of 0.754 and 0.893 at 21 FPS, meeting real-time requirements and demonstrating strong reliability.

微型led晶圆缺陷检测面临光学对比度弱、缺陷类型多样、实时性要求严格等挑战。为了解决这些问题，本文提出了多照度移动检测系统（MIMDS）和DMI-Net。MIMDS采用多角度照明和多光谱融合技术来获得高质量的缺陷图像。DMI-Net集成了三个核心创新：双接受场注意卷积（DRAConv）模块，通过分组和分段卷积增强特征提取，同时降低计算成本；通过多尺度特征细化提高小缺陷识别的混合局部通道注意瓶颈（MRB）；以及适应采样难度的Inner Focaler IoU （IF-IoU）损失函数，以实现更快的收敛和更好的精度。在VOC2012和Micro-LED晶圆数据集上的实验表明，DMI-Net在21 FPS下的mAP得分分别为0.754和0.893，满足实时性要求，可靠性强。

引用次数: 0

LGCNet: Local and global collaborative network for hyperspectral image classification LGCNet：用于高光谱图像分类的本地和全球协作网络

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

Infrared Physics & Technology

Pub Date : 2026-01-02 DOI: 10.1016/j.infrared.2025.106365

Muzi Wang , Jinliang An , Longlong Dai , Zheng Liang , Wenyi Zhao , Weidong Zhang

Hyperspectral image classification is a fundamental task in hyperspectral analysis and forms the basis for numerous applications. Deep learning methods have succeeded remarkably, particularly convolutional neural networks (CNNs), graph convolutional networks (GCNs), and Transformers. However, effectively integrating local detail features with global structure information while maintaining flexible spectral-spatial relationships remains a significant challenge. To cope with these issues, we propose a local and global collaborative network (LGCNet) that efficiently integrates local detail features and global semantic information via the synergistic interaction of local and global branches. In the local branch module, a multi-scale depthwise separable convolution is employed to extract multi-scale spatial–spectral features. In the global branch module, we combine the GCN and graph Transformer with complementary strengths, where the GCN exploits intrinsic graph topology. In contrast, the graph Transformer enables dynamic relationship learning via self-attention. Extensive experiments demonstrate that LGCNet achieves overall classification accuracies of 97.36% on Indian Pines, 98.68% on Pavia University, and 98.47% on Salinas datasets, outperforming state-of-the-art methods.

高光谱图像分类是高光谱分析中的一项基本任务，也是众多应用的基础。深度学习方法取得了显著的成功，特别是卷积神经网络（cnn）、图卷积网络（GCNs）和变形金刚。然而，如何有效地将局部细节特征与全局结构信息相结合，同时保持灵活的光谱-空间关系仍然是一个重大挑战。为了解决这些问题，我们提出了一个局部和全局协同网络（LGCNet），该网络通过局部和全局分支的协同交互，有效地集成了局部细节特征和全局语义信息。在局部分支模块中，采用多尺度深度可分卷积提取多尺度空间光谱特征。在全局分支模块中，我们将GCN和具有互补优势的图转换器结合起来，其中GCN利用了固有的图拓扑结构。相反，图Transformer通过自我关注实现动态关系学习。大量实验表明，LGCNet在Indian Pines数据集上的总体分类准确率为97.36%，在Pavia University数据集上为98.68%，在Salinas数据集上为98.47%，优于目前最先进的方法。

{"title":"LGCNet: Local and global collaborative network for hyperspectral image classification","authors":"Muzi Wang , Jinliang An , Longlong Dai , Zheng Liang , Wenyi Zhao , Weidong Zhang","doi":"10.1016/j.infrared.2025.106365","DOIUrl":"10.1016/j.infrared.2025.106365","url":null,"abstract":"<div><div>Hyperspectral image classification is a fundamental task in hyperspectral analysis and forms the basis for numerous applications. Deep learning methods have succeeded remarkably, particularly convolutional neural networks (CNNs), graph convolutional networks (GCNs), and Transformers. However, effectively integrating local detail features with global structure information while maintaining flexible spectral-spatial relationships remains a significant challenge. To cope with these issues, we propose a local and global collaborative network (LGCNet) that efficiently integrates local detail features and global semantic information via the synergistic interaction of local and global branches. In the local branch module, a multi-scale depthwise separable convolution is employed to extract multi-scale spatial–spectral features. In the global branch module, we combine the GCN and graph Transformer with complementary strengths, where the GCN exploits intrinsic graph topology. In contrast, the graph Transformer enables dynamic relationship learning via self-attention. Extensive experiments demonstrate that LGCNet achieves overall classification accuracies of 97.36% on Indian Pines, 98.68% on Pavia University, and 98.47% on Salinas datasets, outperforming state-of-the-art methods.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106365"},"PeriodicalIF":3.4,"publicationDate":"2026-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145895867","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

Low-level wind shear lidar observation under typical weather process 典型天气过程下低空风切变激光雷达观测

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

Infrared Physics & Technology

Pub Date : 2026-01-02 DOI: 10.1016/j.infrared.2025.106359

Xinyu Zhang , Hongwei Zhang , Weijun Zhang , Xitao Wang , Xiaoying Liu , Jie Yang , Shengguang Qin , Jiaping Yin , Qichao Wang , Songhua Wu

Low-level wind shear may pose significant risks to aviation safety during critical flight phases such as take-off and landing. The pulsed coherent Doppler lidar has been widely applied in wind observation and wind shear identification. In this paper, pulsed coherent Doppler lidars deployed at Guangzhou Baiyun International Airport in 2023 effectively addressed this threat through optimized Doppler Beam Swinging (DBS) and Plan Position Indicator (PPI) scanning modes. We developed wind retrieval algorithms to quantify equipment performance, wind fluctuations, and low-level wind shear characteristics during three high-impact weather phenomena: low-level jets, thunderstorms, and typhoons. Statistical analysis of 12-hour high-resolution lidar data sets per weather process—aligned with meteorological records—revealed key dynamical parameters. This paper demonstrates the effectiveness of coherent Doppler lidar in wind observation and low-level wind shear identification under typical weather conditions, providing a reference for the selection of laser remote sensing equipment and optimization of observation modes. In addition, the results including intensity and scale of low-level wind shear based on lidar systems, are presented in this paper, which provide key refined dynamical parameters for the security of aircraft landing and take-off.

在起飞和降落等关键飞行阶段，低空风切变可能对航空安全构成重大威胁。脉冲相干多普勒激光雷达在风观测和风切变识别中得到了广泛的应用。本文将于2023年部署在广州白云国际机场的脉冲相干多普勒激光雷达，通过优化多普勒波束摆动（DBS）和平面位置指示（PPI）扫描模式，有效解决了这一威胁。我们开发了风检索算法来量化三种高影响天气现象（低空急流、雷暴和台风）中的设备性能、风波动和低空风切变特征。对每个天气过程的12小时高分辨率激光雷达数据集进行统计分析（与气象记录一致），揭示了关键的动力学参数。本文论证了相干多普勒激光雷达在典型天气条件下的风观测和低层风切变识别中的有效性，为激光遥感设备的选择和观测模式的优化提供参考。此外，本文还介绍了基于激光雷达系统的低空风切变强度和尺度分析结果，为飞机安全起降提供了关键的精细化动力学参数。

{"title":"Low-level wind shear lidar observation under typical weather process","authors":"Xinyu Zhang , Hongwei Zhang , Weijun Zhang , Xitao Wang , Xiaoying Liu , Jie Yang , Shengguang Qin , Jiaping Yin , Qichao Wang , Songhua Wu","doi":"10.1016/j.infrared.2025.106359","DOIUrl":"10.1016/j.infrared.2025.106359","url":null,"abstract":"<div><div>Low-level wind shear may pose significant risks to aviation safety during critical flight phases such as take-off and landing. The pulsed coherent Doppler lidar has been widely applied in wind observation and wind shear identification. In this paper, pulsed coherent Doppler lidars deployed at Guangzhou Baiyun International Airport in 2023 effectively addressed this threat through optimized Doppler Beam Swinging (DBS) and Plan Position Indicator (PPI) scanning modes. We developed wind retrieval algorithms to quantify equipment performance, wind fluctuations, and low-level wind shear characteristics during three high-impact weather phenomena: low-level jets, thunderstorms, and typhoons. Statistical analysis of 12-hour high-resolution lidar data sets per weather process—aligned with meteorological records—revealed key dynamical parameters. This paper demonstrates the effectiveness of coherent Doppler lidar in wind observation and low-level wind shear identification under typical weather conditions, providing a reference for the selection of laser remote sensing equipment and optimization of observation modes. In addition, the results including intensity and scale of low-level wind shear based on lidar systems, are presented in this paper, which provide key refined dynamical parameters for the security of aircraft landing and take-off.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"154 ","pages":"Article 106359"},"PeriodicalIF":3.4,"publicationDate":"2026-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923450","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

Uncooled 2D InSe/3D PbSe van der Waals barrier infrared detector 非冷却2D InSe/3D PbSe范德华势垒红外探测器

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

Infrared Physics & Technology

Pub Date : 2026-01-01 DOI: 10.1016/j.infrared.2025.106344

Bingbing Li , Jianliang Li , Jintao Fu , Sainan Li , Yanan Bao , Jiao Xu , Meiqi An , Jing Li , Yiming Yang , Jijun Qiu

The two-dimensional van der Waals (2D vdW) photodetector is facing many challenges, such as the low quantum efficiency, the narrow detection range and the oxygen-free environment. In this study, a 2D Graphene-InSe/3D PbSe van der Waals barrier (vdW-B) vertical architecture was proposed and fabricated for high-performance infrared detection at room temperature. In the atmosphere environment, the 2D Gr-InSe/ 3D PbSe vdW-B detector exhibits a low dark current of 10⁻¹²A with high current on/off ratio (∼10⁴), with high responsivity (∼5 × 10³ mA/W), and shows a room-temperature detectivity of 2.6 × 10¹⁰ Jones at 2700 nm wavelength. The 2D-3D vdW-B detector architecture explored a new pathway for high-performance uncooled infrared photodetectors.

二维范德华光电探测器面临着量子效率低、探测范围窄、无氧环境等诸多挑战。在本研究中，提出并制作了用于室温下高性能红外探测的二维石墨烯- inse /三维PbSe范德华势垒（vdW-B）垂直结构。在大气环境下，2D Gr-InSe/ 3D PbSe vdW-B探测器具有10−12A的低暗电流，高电流开/关比（~ 104），高响应率（~ 5 × 103 mA/W）， 2700 nm波长的室温探测率为2.6 × 1010 Jones。2D-3D vdW-B探测器架构为高性能非制冷红外探测器探索了一条新途径。

引用次数: 0

Design of a 2 mm aperture metasurface single lens for integrated thermopile sensors 集成热电堆传感器用2mm孔径超表面单透镜的设计

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

Infrared Physics & Technology

Pub Date : 2026-01-01 DOI: 10.1016/j.infrared.2025.106362

Yang Yujie , Wang Fucen , Xie Qiming , Rong Yu , Ma Haiyang , Zhao Xinyu , Dong Haikuo , Cheng Haijuan , Wu Jingwei , Liu Guangsen

This paper presents a single-layer germanium metalens operating across the 8–11 μm wavelength band, designed to overcome the miniaturization challenges posed by conventional optics in infrared thermopile arrays. Leveraging the relatively large pixel size of these arrays, which relaxes the requirement for diffraction-limited imaging performance, we strategically constrained the system’s time-bandwidth product during the ray-tracing optimization process. This approach reduced the required group delay by approximately 20 % compared to conventional broadband achromatic designs, thereby significantly relaxing the demanding high-dispersion-compensation requirement for large-aperture (2 mm) metalenses. The lens was designed using a hybrid methodology that combines ray tracing, finite-difference time-domain (FDTD) full-wave simulation, and scalar diffraction theory, achieving a compact profile (12-μm nanopillars on a 0.5-mm substrate) and a 20° field-of-view without the need for complex multilayer structures or front aperture stops. Simulation results demonstrate that the modulation transfer function remains above 0.3 at 16.7 lp/mm in the central field, substantially surpassing the 3.85 lp/mm cutoff frequency of a commercial 130-μm-pitch thermopile array. This work provides a viable design pathway toward highly integrated, cost-effective thermal imaging systems for thermopile sensors via semiconductor processes.

本文提出了一种工作在8-11 μm波段的单层锗超构透镜，旨在克服传统光学在红外热电堆阵列中的小型化挑战。利用这些阵列相对较大的像素尺寸，放宽了对衍射受限成像性能的要求，我们在光线跟踪优化过程中战略性地限制了系统的时间带宽积。与传统的宽带消色差设计相比，这种方法将所需的群延迟降低了约20%，从而大大降低了大孔径（2mm）超透镜的高色散补偿要求。该透镜采用射线追踪、时域有限差分（FDTD）全波模拟和标量衍射理论相结合的混合方法设计，实现了紧凑的轮廓（在0.5 mm的衬底上有12 μm的纳米柱）和20°的视场，而无需复杂的多层结构或前孔径停止。仿真结果表明，在中心场16.7 lp/mm时，调制传递函数保持在0.3以上，大大超过了商用130 μm-pitch热电堆阵列3.85 lp/mm的截止频率。这项工作为通过半导体工艺实现热电堆传感器的高度集成、成本效益高的热成像系统提供了一条可行的设计途径。

{"title":"Design of a 2 mm aperture metasurface single lens for integrated thermopile sensors","authors":"Yang Yujie , Wang Fucen , Xie Qiming , Rong Yu , Ma Haiyang , Zhao Xinyu , Dong Haikuo , Cheng Haijuan , Wu Jingwei , Liu Guangsen","doi":"10.1016/j.infrared.2025.106362","DOIUrl":"10.1016/j.infrared.2025.106362","url":null,"abstract":"<div><div>This paper presents a single-layer germanium metalens operating across the 8–11 μm wavelength band, designed to overcome the miniaturization challenges posed by conventional optics in infrared thermopile arrays. Leveraging the relatively large pixel size of these arrays, which relaxes the requirement for diffraction-limited imaging performance, we strategically constrained the system’s time-bandwidth product during the ray-tracing optimization process. This approach reduced the required group delay by approximately 20 % compared to conventional broadband achromatic designs, thereby significantly relaxing the demanding high-dispersion-compensation requirement for large-aperture (2 mm) metalenses. The lens was designed using a hybrid methodology that combines ray tracing, finite-difference time-domain (FDTD) full-wave simulation, and scalar diffraction theory, achieving a compact profile (12-μm nanopillars on a 0.5-mm substrate) and a 20° field-of-view without the need for complex multilayer structures or front aperture stops. Simulation results demonstrate that the modulation transfer function remains above 0.3 at 16.7 lp/mm in the central field, substantially surpassing the 3.85 lp/mm cutoff frequency of a commercial 130-μm-pitch thermopile array. This work provides a viable design pathway toward highly integrated, cost-effective thermal imaging systems for thermopile sensors via semiconductor processes.</div></div>","PeriodicalId":13549,"journal":{"name":"Infrared Physics & Technology","volume":"153 ","pages":"Article 106362"},"PeriodicalIF":3.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145880179","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