The metasurface refractive index sensor has a high degree of tunability and flexibility, providing excellent performance for high precision refractive index sensing applications. The metasurface absorber with metallic structure has been hindered in further sensor applications due to the inherent Ohmic loss of the metallic material. In this study, a dual nanorod metasurface structure based on semiconductor Si was designed, introducing a symmetry‐breaking structure to excite dual ultra‐narrow q‐BIC resonance peaks with Fano line shapes. Both peaks are located in the near‐infrared region, and multipole analysis shows that this strong field enhancement effect is induced by a magnetic dipole. Experimental results demonstrate the potential of this sensor to provide dual‐channel detection while achieving high sensitivity and high Q‐factor. We believe that this device exhibits outstanding performance and high practicality, providing a reference for the development and application of biological and environmental sensors.

Micro-expressions, which are spontaneous and difficult to suppress, reveal a person’s true emotions. They are characterized by short duration and low intensity, making the task of micro-expression recognition challenging in the field of emotion computing. In recent years, deep learning-based feature extraction and fusion techniques have been widely used for micro-expression recognition, particularly methods based on Vision Transformer that have gained popularity. However, the Vision Transformer-based architecture used in micro-expression recognition involves a significant amount of invalid computation. Additionally, in the traditional two-stream architecture, although separate streams are combined through late fusion, only the output features from the deepest level of the network are utilized for classification, thus limiting the network’s ability to capture subtle details due to the lack of fine-grained information. To address these issues, we propose a new two-level spatio-temporal feature fused with a two-stream architecture. This architecture includes a spatial encoder (modified ResNet) for learning texture features of the face, a temporal encoder (Swin Transformer) for learning facial muscle motor features, a feature fusion algorithm for integrating multi-level spatio-temporal features, a classification head, and a weighted average operator for temporal aggregation. The two-stream architecture has the advantage of extracting richer features compared to the single-stream architecture, leading to improved performance. The shifted window scheme of Swin Transformer restricts self-attention computation to non-overlapping local windows and allows cross-window connections, significantly improving the performance and reducing the computation compared to Vision Transformer. Moreover, the modified ResNet is computationally less intensive. Our proposed feature fusion algorithm leverages the similarity in output feature shapes at each stage of the two streams, enabling the effective fusion of multi-level spatio-temporal features. This algorithm results in an improvement of approximately 4% in both the F1 score and the UAR. Comprehensive evaluations conducted on three widely used spontaneous micro-expression datasets (SMIC-HS, CASME II, and SAMM) consistently demonstrate the superiority of our approach over comparative methods. Notably, our approach achieves a UAR exceeding 0.905 on CASME II, making it one of the few frameworks in the published micro-expression recognition literature to achieve such high performance.

This paper focuses on image segmentation, image correction and spatial-spectral dimensional denoising of images in hyperspectral image preprocessing to improve the classification accuracy of hyperspectral images. Firstly, the images were filtered and segmented by using spectral angle and principal component analysis, and the segmented results are intersected and then used to mask the hyperspectral images. Hyperspectral images with a excellent segmentation result was obtained. Secondly, the standard reflectance plates with reflectance of 2 and 98% were used as a priori spectral information for image correction of samples with known true spectral information. The mean square error between the corrected and calibrated spectra is less than 0.0001. Comparing with the black-and-white correction method, the classification model constructed based on this method has higher classification accuracy. Finally, the convolution kernel of the one-dimensional Savitzky-Golay (SG) filter was extended into a two-dimensional convolution kernel to perform joint spatial-spectral dimensional filtering (TSG) on the hyperspectral images. The SG filter (m = 7,n = 3) and TSG filter (m = 3,n = 4) were applied to the hyperspectral image of Pavia University and the quality of the hyperspectral image was evaluated. It was found that the TSG filter retained most of the original features while the noise information of the filtered hyperspectral image was less. The hyperspectral images of sample 1–1 and sample 1–2 were processed by the image segmentation and image correction methods proposed in this paper. Then the classification models based on SG filtering and TSG filtering hyperspectral images were constructed, respectively. The results showed that the TSG filter-based model had higher classification accuracy and the classification accuracy is more than 98%.

Fast on‐site detection of potentially toxic metals in soils is crucial for soil remediation and contamination monitoring. The moisture in soils is a challenge for quantitative analysis of potentially toxic metals. Here, we used laser‐induced breakdown spectroscopy (LIBS) and partial least squares regression (PLSR) for fast on‐site quantitative analysis of potentially toxic metals in wet soil samples. The model offered direct measurement of the moisture content in soil samples: the coefficient of determination ( R ² ) value and the root‐mean‐square error of prediction (RMSEP) of predicted moisture content were 0.98% and 1.3%, respectively. The correlation between laser ablation factor and moisture content of the samples was determined by analysing the influence of sample moisture content on laser ablation. Quantitative analysis of wet soil samples based on ablation factor was established with an R ² of 0.97. Potentially toxic elemental Cu and Cr in sample 4# were verified by the quantitative analysis model through weighing and direct measurement. The relative error of concentration was within 10%, and the accuracy was improved by over 80%. The preliminary results show that the quantitative analysis model of potentially toxic metals in wet soil samples based on LIBS technology can detect potentially toxic metals in wet soil samples quickly and accurately on‐site. This has important guiding significance for real‐time and on‐site monitoring of contaminated soils. Highlights: Fast on‐site quantitative analysis of potentially toxic metals in wet soil samples Quantitative analysis of wet soil samples based on ablation factor was established The relative error of concentration was within 10%, and the accuracy was improved by over 80% The model can detect potentially toxic metals in wet soil samples quickly and accurately on‐site

A swing ion-beam etching method to fabricate convex blazed gratings used in shortwave infrared hyperspectral imaging spectrometers is presented. This method solves the consistency problem of blaze angles by swing etching through the meridian direction of the gratings. The mathematical relationship of the curvature, aperture, and diffraction efficiency of convex gratings is studied to demonstrate the limitation of conventional translational lithography and the necessity of swing etching. A geometric model is built to analyze the influence of swinging speed and beam slit width on groove evolution. Convex gratings with a 45.5 gr/mm groove density, 67 mm aperture, 156.88 mm radius of curvature, and 2.2° blaze angle have been fabricated and measured where the peak and average diffraction efficiency in the shortwave infrared band reach 90% and 70%, respectively. Experimental results validate that high-efficiency convex gratings of small blaze angle and high groove consistency can be produced by swing etching, which satisfy the requirements for high spectral resolution and miniaturization of imaging spectrometers.

Hyperspectral unmixing (HU) aims to decompose mixed pixels into a set of endmembers and corresponding abundances. Deep learning-based HU methods are currently a hot research topic, but most existing unmixing methods still rely on per-pixel training or employ convolutional neural networks (CNNs), which overlook the non-local correlations of materials and spectral characteristics. Furthermore, current research mainly focuses on linear mixing models, which limits the feature extraction capability of deep encoders and further improvement in unmixing accuracy. In this paper, we propose a nonlinear unmixing network capable of extracting global spatial-spectral features. The network is designed based on an autoencoder architecture, where a dual-stream CNNs is employed in the encoder to separately extract spectral and local spatial information. The extracted features are then fused together to form a more complete representation of the input data. Subsequently, a linear projection-based multi-head self-attention mechanism is applied to capture global contextual information, allowing for comprehensive spatial information extraction while maintaining lightweight computation. To achieve better reconstruction performance, a model-free nonlinear mixing approach is adopted to enhance the model’s universality, with the mixing model learned entirely from the data. Additionally, an initialization method based on endmember bundles is utilized to reduce interference from outliers and noise. Comparative results on real datasets against several state-of-the-art unmixing methods demonstrate the superior of the proposed approach.