Infrared Physics & Technology最新文献

It’s crucial for both producers and consumers to accurately trace the origin of millet, given the significant differences in price and taste that exist between millets from various origins. The traditional method of identifying the origin of millet is time-consuming, laborious, complex, and destructive. In this study, a new method for fast and non-destructive differentiation of millet origins is developed by combining terahertz time domain spectroscopy with ensemble learning. Firstly, three machine learning algorithms, namely support vector machine (SVM), random forest (RF), and kernel extreme learning machine (KELM), were used to build different discriminative models, and then the impact of six different preprocessing methods on the models’ classification performance was compared. It was observed that models employing Savitzky-Golay preprocessing exhibited pronounced superiority in accurately determining the millet’s geographical origins. Building upon these findings, the research introduces an innovative ensemble learning strategy, leveraging both topsis and stacking techniques, to harness the collective strengths of the three algorithms. The outcomes of this approach reveal its remarkable capacity to distinguish millets originating from five distinct locations without the necessity for any parameter fine-tuning. The accuracy, F1 score, and Kappa on the prediction set are all 100 %, which significantly outperforms the single model, traditional voting method, and stacking method. The culmination of this study suggests that the integration of terahertz time-domain spectroscopy and TOPSIS-Stacking ensemble learning emerges as a promising method for the swift and non-intrusive discrimination of millet geographical origins with remarkable precision.

The current mainstream object detection networks perform well in RGB visible images, but they require high computational resource and degrade in performance when applied to low-resolution infrared images. To address above issues, we propose a lightweight algorithm YOLO-SGF based on you-only-look-once version8 (YOLOv8). Firstly, the lightweight cross-scale feature map fusion network GCFVoV designed as neck to solve poor detection accuracy and maintain low complexity in lightweight networks. And a lightweight GCVF module in GCFVoV neck uses GSConv and Conv to process deep and shallow features respectively, which maximally preserves implicit connections between each channel and integrates multi-scale features. Secondly, we utilize ShuffleNetV2-block1 in combination with C2f for feature extraction, making the algorithm more lightweight and effectively. Finally, we propose the FIMPDIoU loss function, which focuses on overlooked objects in complex backgrounds and adjusts the prediction boxes using ratios specific to different sizes of objects. Compared with YOLOv8 in our infrared dataset, YOLO-SGF reduces the computational space complexity by 50 % and time complexity by 42 %, increases FPS₃₂ by 36.3 % and improves [email protected] ∼ 0.95 by 1.1 % in object detection. Our algorithm enhances the capability of object detection in infrared images especially in nighttime, low light, and occluded conditions. YOLO-SGF enables deployment on embedded edge devices with limited computing power, and provides a new idea for lightweight networks.

Efficiently sorting coal gangue and identifying coal types are vital operations in coal preparation, yet they are traditionally resource-consuming, labor-intensive, and potentially hazardous. This work puts forward an straightforward method employing mid-infrared spectroscopy with first derivative spectrum to address these issues. The proposed technique focuses on the delineation and enhancement of characteristic spectra to detect subtle differences among samples. The method utilizes just a few characteristic spectra of 3740–3700 cm⁻¹, 1790–1750 cm⁻¹, 1615–1583 cm⁻¹, 1580–1540 cm⁻¹, 1550–1440 cm⁻¹, 1270–1210 cm⁻¹ and 867–854 cm⁻¹ to achieve 100 % high-accuracy classification of coal gangue and identification of coal types with total 250 spectra, such as bituminite, anthracite, lignite, roof sandstone and gangue, without the need for secondary sample processing or the assistance of machine learning algorithms, simplifying the process considerably. Such a strategy not only significantly improves the efficiency of coal sorting but also endorses real-time on-site detection. It offers a theoretical foundation for advanced coal separation technology and its implementation in real-world mining operations.

Generating a single fused image that highlights important targets and preserves textural details is the aim of fusing visible and infrared images. The majority of deep learning-based fusion algorithms now in use can produce decent fusion outcomes; however, the modeling process still lacks consideration of the different amounts of information in different scenes or regions. Thus, we propose in this research SeACPFusion, a luminance-aware adaptive fusion network for infrared and visible images, which adaptively preserves the intensity information of the noticeable targets of the source images with the texture information of the background in an optimal ratio. Specifically, we design pixel-level luminance loss (PBL) to direct the fusion model’s training in real-time, and PBL retains the optimal intensity information according to the pixel luminance ratio of different source images. In addition, we designed the Channel Transformer (CTF) to consider the relationship between different attributes from the point of view of the feature channel and to focus on the key information by using the self-focusing mechanism to achieve the goal of adaptive fusion. Our extensive tests on the MSRS, RoadScene, and TNO datasets demonstrate that SeACPFusion surpasses nine representative deep learning methods on six objective metrics and achieves the best visual results in scenes such as overexposure or underexposure. In addition, the relatively efficient operation and fewer model parameters make our algorithm promising as a preprocessing module for downstream complicated vision tasks.

Non-contact sludge measurement methods for storage tanks can address the challenge of measuring the volume of sedimented sludge during long-term storage. While infrared thermography technology can address the issue of liquid level detection, its measurement accuracy for the undulating interface of sludge is insufficient. This study designed and constructed experimental setups for measuring sludge in storage tanks. In this study, infrared images taken by an infrared camera were used to record the temperature distribution of the outer wall of the storage tank. The threshold segmentation method is used to determine the accurate sludge boundary line in image processing. Finally, the Three-Dimensional Tank Residue Recovery Algorithm (3D-TRRA) was applied to fit the 3D distribution of the sludge and calculate accurate sludge volumes. The results indicate that the best segmentation is achieved with a threshold of 170. The measurement error for sludge volume is less than 5%. Accurate visual positioning and recognition of sludge are achieved.

The n⁻ region is crucial to the performance of n-on-p HgCdTe devices. However, the underlying mechanisms governing its formation process remain insufficiently elucidated in current literature. In this work, the influence of annealing temperature on the n⁻ region formation process was investigated systematically through experiments and one-dimensional (1D) simulation. The two key parameters, the transport rate of interstitials (Tr_I) and the diffusion coefficient of vacancies (D_V) were determined through the 1D model, and their accuracy was validated by experiments. The determination of Tr_I and D_V allows for more flexible and precise optimization of the n⁻ region in HgCdTe, thereby providing valuable guidance for cost-effective, high performance, and reliable preparation of HgCdTe detectors.

Infrared based visual perception is important for night vision of autonomous vehicles, unmanned aerial vehicles (UAVs), etc. Semantic segmentation based on deep learning is one of the key techniques for infrared vision-based perception systems. Currently, most of the advanced methods are based on Transformers, which can achieve favorable segmentation accuracy. However, the high complexity of Transformers prevents them from meeting the real-time requirement of inference speed in resource constrained applications. In view of this, we suggest several lightweight designs that significantly reduce existing computational complexity. In order to maintain the segmentation accuracy, we further introduce the recent vision big model — Segment Anything Model (SAM) to supply auxiliary supervisory signals while training models. Based on these designs, we propose a lightweight segmentation network termed SMALNet (Segment Anything Model Aided Lightweight Network). Compared to existing state-of-the-art method, SegFormer, it reduces 64% FLOPs while maintaining the accuracy to a large extent on two commonly-used benchmarks. The proposed SMALNet can be used in various infrared based vision perception systems with limited hardware resources.

We demonstrate a tri-corner cube Q-switched Ho:YAG spatial ring cavity laser, which was resonantly pumped by a 1908 nm fiber laser. The polarization state of the intracavity oscillating laser was adjusted by a half-wave plate, and a continuous s-polarized laser of 2.57 W at 2090.9 nm was obtained at a pump power of 18.5 W, corresponding to an optical-to-optical conversion efficiency of 13.9 % and a slope efficiency of 33.3 %. When the corner cube prism, which has the weakest anti-misalignment capability, was tilted vertically by 1° or horizontally by 0.92°, the laser could still output the laser. For Q-switched operation, the tri-corner cube Ho:YAG laser has a pulse energy of 9.86mJ and a pulse width of 178.8 ns at a repetition rate of 100 Hz. At the maximum output energy, the beam quality was M_x² = 1.3, M_y² = 1.2.