Shan Gao

Multispectral radiation thermometry (MRT) estimates temperature from several spectral-radiance measurements while compensating for unknown emissivity, but the inversion stability is strongly determined by the wavelength configuration chosen before the inversion. We propose a temperature-parameter Cramér–Rao bound (CRB) guided wavelength-selection method that treats emissivity-model parameters as nuisance parameters and minimizes the CRB component associated with the target temperature. For a power-law emissivity model, an analytical Jacobian is derived for the Fisher information matrix, enabling deterministic evaluation of candidate wavelength sets. Numerical simulations in the 1.1–2.5 μm band show that the predicted temperature CRB agrees with the empirical Monte Carlo temperature-error standard deviation over 27 temperature–emissivity scenarios. Compared with uniform spacing, the proposed CRB-Opt design reduces the average CRB from 3.82% to 2.60%, the RMSE from 3.82% to 2.60%, and the 95th-percentile ARE from 7.59% to 5.23%. Experiments on a pre-oxidized GH4169 nickel-based superalloy further indicate that, under the same calibration and inversion procedure, the CRB-Opt wavelengths reduce the median ARE from 9.85% to 2.50% and suppress the high-error tail relative to uniform spacing. The method provides a temperature-prioritized, physically interpretable, and computationally feasible front-end channel-design criterion for MRT.

Accurate temperature assessment in complex combustion environments is crucial for the thermal diagnostics of aero-engines and gas turbines. To address the challenges posed by intense ambient radiation, a background-corrected multispectral light-field (MSLF) thermometry framework is developed. The system integrates single-snapshot 4D acquisition with an emissivity-adaptive model. A dual strategy is employed to mitigate radiative interference: a background-correction algorithm eliminates reflection artifacts, while discrete, narrowband wavelengths (specifically 700, 750, 800, and 850 nm) are selected to spectrally bypass gaseous radiation. Validation in a heating furnace demonstrates that reflection-induced errors can be reduced to within 6.5 K. Furthermore, jet-flame impingement tests confirm that absolute deviations were limited to within 20 K across varying emissivities (0.2–0.9), ensuring robust performance in high-temperature environments.

Near-infrared multispectral light-field (NIR-MSLF) thermography was developed for wide-range temperature measurement. By integrating an optical-filter-array with a near-infrared light-field camera, the system enables single-shot acquisition of two-dimensional (2D) temperature distributions of high-temperature objects. Blackbody calibration experiments demonstrated the lower temperature threshold successfully extended to 573 K compared with that of visible multispectral light-field thermography, and errors were maintained below ±10 K over 573–1873 K. Furthermore, to address the effect of intense ambient radiation on accurate temperature calculation, a three-dimensional view-factor mapping (3D-VFM) method was proposed to quantify ambient radiation. By projecting three-dimensional geometric elements onto the pixel plane, the method derives a 2D view-factor distribution, enabling pixel-level correction of ambient radiation. The reliability of the proposed NIR-MSLF thermography and 3D-VFM method was validated through turbine blade film cooling experiments. Under various coolant flow rates, in conditions where ambient radiation may exceed the radiation from turbine blade itself, the NIR-MSLF thermography measured blade surface temperatures with deviations consistently within ±15 K when compared to the buried thermocouples.

In hyperspectral image (HSI) classification, Transformer and CNN are widely used because they complement each other in extracting features. Nevertheless, existing Transformer-based methods still struggle to capture fine-grained local spatial details and positional relationships. Meanwhile, CNN-based attention mechanisms typically process all spectral bands simultaneously to obtain global features, often overlooking locally salient information. To address these challenges, we propose an innovative normalized grouping fused comprehensive relative position transformer network (NG-CRPTN). Specifically, NG-CRPTN comprises two core modules: the comprehensive relative position module and the normalized mutual information band division. The former employs tokens of different dimensions as its foundation, modeling local-to-global spatial structures and relative positional relationships through the local independent relative position transformer and global independent relative position transformer, while incorporating a local complementary enhancement fusion module to deeply integrate high and low-frequency information from spatial features across different dimensions. The latter achieves adaptive band grouping of HSI spectral bands to better focus on locally salient features in the spectral dimension. Extensive experiments conducted on four HSI datasets demonstrate that the proposed NG-CRPTN achieves superior classification accuracy with relatively few training samples compared to existing state-of-the-art methods.

Hyperspectral target detection (HTD) faces significant challenges due to the scarcity of labeled samples and the redundancy in spectral information. To address these challenges, this paper proposes a multi-scale lightweight transformer-based detector (MSLTD) specifically designed for hyperspectral target detection. First, a novel training sample generation strategy is developed that relies solely on raw hyperspectral data and a single prior target spectrum to generate a sufficient number of high-quality training samples. Second, a multi-scale feature lightweight network based on three-channel transformer is proposed. It first divides spectral pixels into consecutive groups to extract local spectral features across adjacent bands, and then introduces a multi-scale sparse subspace transformer to capture and fuse intrinsic spectral characteristics and long-range dependencies at multiple resolutions while effectively reducing spectral redundancy. In addition, an adaptive hardness-aware contrastive (AHC) loss function is introduced, which integrates triplet loss and focal loss to adaptively emphasize hard-to-classify samples and improve the discriminative ability between targets and backgrounds. The effectiveness of the proposed MSLTD is validated through comparative experiments on four real hyperspectral datasets against seven state-of-the-art target detection methods, demonstrating improved detection accuracy and competitive computational efficiency. Extensive ablation studies further confirm the contributions of key components in the framework, highlighting its robustness and applicability in various hyperspectral target detection scenarios. The code for this work is available at https://github.com/2474109721ct/MSLTD for reproducibility purposes.

The surface radiation characteristics of high-temperature alloy materials are influenced by various factors, which can decrease the accuracy of the temperature measurement. In this paper, a surface radiation thermometry method of high temperature alloy incorporating emissivity variation constraints is proposed to accurately describe the trend of emissivity variation with wavelength by constructing an emissivity variation constraint model based on ridge estimation. The golden sine algorithm and the dynamic weight factor are further used to improve the whale optimisation algorithm to achieve high-precision temperature solution of the objective function after the emissivity variation constraint. The trends before and after emissivity correction are analysed by simulation, and the experimental validation shows that the maximum temperature measurement error of the algorithm is no more than 0.3 % , and the average time is 0.057 s. The algorithm shows good accuracy and stability in the radiation thermometry of high-temperature metal samples, which verifies the feasibility and high efficiency of the method in practical applications.