Owing to the diversity of underwater environments and the scattering and selective absorption of light in water, acquired underwater images usually suffer from severe quality degradation problems, such as color deviation, low clarity, and low brightness. To solve these problems, an underwater image enhancement algorithm that combines a transformer and generative adversarial network is proposed. Based on the generative adversarial network, a generative adversarial network with transformer (TGAN) network enhancement model is constructed by combining the coding and decoding structure, global feature modeling transformer module based on the spatial self-attention mechanism, and channel-level multi-scale feature fusion transformer module. The model focuses on color and spatial channels with more serious underwater image attenuation. This effectively enhances the image details and solves the color-deviation problem. Additionally, a multinomial loss function, combining RGB and LAB color spaces, is designed to constrain the adversarial training of the network enhancement model. The experimental results demonstrate that when compared to typical underwater image enhancement algorithms, such as contrast-limited adaptive histogram equalization (CLAHE), underwater dark channel prior (UDCP), underwater based on convolutional neural network (UWCNN), and fast underwater image enhancement for improved visual perception (FUnIE-GAN), the proposed algorithm can significantly improve the clarity, detail texture, and color performance of underwater images. Specifically, the average values of the objective evaluation metrics, including the peak signal-to-noise ratio, structural similarity index, and underwater image quality measure, improve by 5.8%, 1.8%, and 3.6%, respectively. The proposed algorithm effectively improves the visual perception of underwater images.

In the process of underwater imaging, the light source is one of the key factors affecting image quality owing to the scattering and absorption of light. This results in many problems such as color distortion, low contrast, and visibility of underwater images. Underwater images with degraded quality are not conducive to analysis and utilization. To address these problems, we propose an underwater image-enhancement algorithm based on color correction and dark–bright dual-channel prior. First, a color-compensation algorithm based on the standard deviation ratio is proposed to effectively solve the color-distortion problem. Sharpening is used to enhance the details and edges of the image to obtain a contrast-enhanced image. Conversely, we propose a dark and bright dual-channel to remove image blur based on channel-difference weighting to obtain a visibility-restored image. Finally, a multiscale fusion method is used to fuse the contrast-enhanced and visibility-restored images based on the weights. The proposed algorithm is compared with other underwater image enhancement algorithms for qualitative and quantitative evaluations. Experimental results show that the proposed algorithm can effectively eliminate color deviations and restore image clarity. The enhanced images generated are better than those generated by other algorithms in terms of Undewater Color Image Quality Evaluation(UCIQE), Underwater Image Quality Measurement (UIQM) andInformation Entropy(IE) parameter indices. Underwater images are of high quality owing to the scattering and absorption of underwater light.

An improved object detection method (YOLO with EffectiveSE, Focal-EIOU, DCNv2, CARAFE, and DyHead) is proposed based on YOLOv5 to address issues in underwater waste infrared target detection, such as blurred boundary details, low image quality, and the presence of various irregular or damaged coverings. The InceptionNeXt network is selected as the backbone network to enhance the model's expressive power and feature extraction capability. Additionally, the EffectiveSE attention mechanism is introduced in the feature fusion layer to adaptively learn the importance of feature channels and selectively weight them. Deformable convolutions are used to replace the C3 module in the original model, enabling it to better perceive the shapes and details of the targets. Moreover, the CARAFE operator is employed to replace the upsampling module, thereby enhancing the representation ability of the fine-grained features and avoiding information loss. In terms of the loss function, the Focal-EIOU loss function is adopted to improve the accuracy of the model in target localization and bounding box regression. Finally, DyHead is introduced to replace the head of YOLOv5, thereby enhancing the model accuracy via dynamic receptive field mechanisms and multiscale feature fusion. The improved EFDCD-YOLO model is applied to underwater waste infrared target detection and compared to the YOLOv5 model. The model achieves a 21.4% improvement in precision (P), 9.7% improvement in recall (R), and 13.6% improvement in mean average precision (mAP). The experimental results demonstrate that EFDCD-YOLO effectively enhances the detection performance in underwater waste infrared target detection scenarios and effectively meets the requirements of underwater infrared target detection.

To address the issues of existing underwater image enhancement methods, which lack focus on critical target objects in images and exhibit poor enhancement effects on edge detail information, in this study, an underwater image enhancement approach is proposed based on an attention mechanism and feature reconstruction. First, a superpixel image enhancement model is constructed by integrating the residual module with the Convolutional Block Attention Module (CBAM), which not only improves the overall quality of underwater images but also enhances the clarity and visibility of target objects in images. Second, an edge difference module is designed to enable the model to focus on high-frequency information in the images, thereby strengthening the edge details of the target objects. Finally, a multi-granularity feature reconstruction module is built to reconstruct the hidden layer features of the superpixel image enhancement model, restore the input image, and further optimize the model parameters. Experimental results demonstrate that when compared with contrastive methods, the proposed model realizes improvements in three evaluation metrics: Structural Similarity (SSIM), Peak Signal to Noise Ratio (PSNR), and Underwater Image Quality Measures (UIQM), indicating better enhancement performance. Notably, it exhibits a remarkable effect in enhancing critical target objects in underwater images.

To address the problems of misdetection, omission detection, and low detection efficiency when detecting underwater targets due to the complex underwater environment, a lightweight underwater target detection algorithm with an improved YOLOv8 model is proposed. First, to ameliorate the problem of insufficient feature fusion in the neck network, the neck network of YOLOv8 is fused with a BiFPN bidirectional feature pyramid structure to improve the detection of the small target layer. Second, to address the problem of the large number of parameters of the convolution module in the network and high computational complexity, an Adaptive-Attention Down-Sampling(AADS) module is designed to replace the convolution module in the backbone network to reduce the number of model parameters and amount of computation. Finally, Large Separable Kernel Attention (LSKA) is introduced to strengthen the feature extraction capability such that the model can focus on important information more accurately and improve target detection accuracy. The experimental results show that in the underwater target detection dataset, the improved algorithm improves the average detection accuracy by 1.4%, reduces the number of model parameters by 43.3%, and reduces the computational complexity of the model by 15.9% when compared with YOLOv8. This realizes a good balance between detection accuracy and detection speed.

Hyperspectral images (HSIs) are polluted by a large amount of mixed noise during the acquisition process, which affects the performance of subsequent applications of remote sensing images. Therefore, restoring clean HSI from the mixed noise is an important preprocessing step. In this study, a hyperspectral mixed noise image restoration model based on nonconvex low-rank tensor decomposition and group-sparse total variational regularization is proposed. On the one hand, by using logarithmic tensor nuclear norm to approximate the low-rank characteristics of the HSI, the inherent tensor structure of hyperspectral data can be utilized, and the shrinkage of larger singular values can be reduced to preserve more detailed features of the image. On the other hand, the group sparse total variational regularization can be used to enhance the spatial sparsity of the HSI and correlation between adjacent spectra. ADMM algorithm is used to solve the problem, and an experiment shows that the algorithm converges easily. In simulated and real hyperspectral image experiments, this method performs better in removing HSI mixed noise when compared to other methods.

Infrared image recognition of substation electrical equipment is an important prerequisite for defect and fault diagnosis to ensure the safe and stable operation of power systems. To realize high-precision and high-efficiency recognition of substation equipment, in this study, an infrared image recognition method of substation equipment is proposed based on an improved YOLOv7 network. The infrared image acquired by the substation is used as the input for the YOLOv7 network. In the recognition of infrared images, a CoordConv convolution layer is used to increase the image coordinate information, enhance the information details of the network layer, and enrich the image feature content. The attention mechanism is introduced to eliminate other information interference, enhance the feature expression ability of the model, and improve the accuracy of network training. To further improve the recognition accuracy, unlike the traditional loss function, the WIoU loss function is used to accelerate the network convergence and improve the model accuracy. By analyzing the actual infrared images acquired by the substation, the experimental results show that the recognition accuracy of the infrared image recognition model of the substation equipment based on the improved YOLOv7 network can reach 97.1%. Compared with the YOLOv7 network and other typical networks, the proposed model has higher accuracy and robustness and can be effectively applied to intelligent monitoring and maintenance of substation equipment, providing basic conditions for subsequent fault diagnosis.

Object detection is a fundamental task in computer vision. Drones equipped with infrared cameras facilitate nighttime reconnaissance and surveillance. To realize small target detection, slight texture information, weak contrast in infrared aerial photography scenes, limited accuracy of traditional algorithms, and heavy dependence on computing power and power consumption in infrared object detection, a pedestrian detection method for infrared aerial photography scenes that integrates salient images is proposed. First, we use U2-Net to generate saliency maps from the original thermal infrared images for image enhancement. Second, we analyze the impact of two fusion methods, pixel-level weighted fusion, and replacement of image channels as image-enhancement schemes. Finally, to improve the adaptability of the algorithm to the target scene, the prior boxes are reclustered. The experimental results show that pixel-level weighted fusion presents excellent results. This method improves the average accuracy of typical YOLOv3, YOLOv3-tiny, and YOLOv4-tiny algorithms by 6.5%, 7.6%, and 6.2%, respectively, demonstrating the effectiveness of the designed fused visual saliency method.

To address challenges such as blurry edges, low contrast, and unclear details in infrared images used in artillery shooting, night vehicle reconnaissance, aerospace, and soldiers' patrolling, this study proposes an enhanced Retinex image enhancement algorithm. The method integrates wavelet transform, improved bilateral filtering, an enhanced threshold function denoising algorithm, and fuzzy set functions. First, the infrared image undergoes wavelet decomposition to extract low and high-frequency coefficients. Subsequently, high-frequency components are enhanced using an improved threshold function, adapting r for denoising purposes. An improved bilateral filtering Retinex algorithm is employed to smooth the infrared image while preserving essential details. The high and low-frequency components are recombined through wavelet reconstruction to reconstruct the enhanced infrared image. A fuzzy set function is applied to further enhance the contrast of the infrared image. Experimental results validate the effectiveness of the proposed algorithm. It effectively reduces noise, enriches image details, suppresses background interference, and enhances contrast compared to conventional methods such as adaptive histogram equalization and multi-scale Retinex image enhancement. This approach not only enhances the quality of infrared images for critical applications but also demonstrates significant improvements over existing methods in terms of clarity and detail retention.

A multiscale and convolutional attention-based infrared and visible image fusion algorithm is proposed to address the issues of insufficient single-scale feature extraction and loss of details, such as infrared targets and visible textures, when fusing infrared and visible images. First, an encoder network, combining a multiscale feature extraction module and deformable convolutional attention module, is designed to extract important feature information of infrared and visible images from multiple receptive fields. Subsequently, a fusion strategy based on spatial and channel dual-attention mechanisms is adopted to further fuse the typical features of infrared and visible images. Finally, a decoder network composed of three convolutional layers is used to reconstruct the fused image. Additionally, hybrid loss function constraint network training based on mean squared error, multiscale structure similarity, and color is designed to further improve the similarity between the fused and source images. The results of the experiment are compared with seven image-fusion algorithms using a public dataset. In terms of subjective and objective evaluations, the proposed algorithm exhibits better edge preservation, source image information retention, and higher fusion image quality than other algorithms.

The existing infrared small-target detection method based on convolutional neural networks (CNN) exhibits the problem of a limited receptive field in the encoder stage, and the decoder lacks an effective feature interaction when fusing multiscale features. To address the aforementioned issues, in this study, a new method is proposed based on an encoder–decoder structure. Specifically, a vision transformer is used as an encoder to extract multiscale features from small infrared target images. The vision transformer is an emerging deep-learning architecture that uses a self-attention mechanism to capture the global relationship between all pixels in the input image, thereby effectively processing long-range dependencies and contextual information in the image. Furthermore, a dual-decoder module, comprising an interactive decoder and auxiliary decoder, is proposed to improve the ability of the decoder to reconstruct small infrared targets. The dual-decoder module can make full use of the complementary information between different features, promote interaction between deep and shallow features, and better reconstruct small infrared targets by combining the results of the two decoders. Experimental results on widely used public datasets show that the proposed method outperforms other methods in terms of two evaluation indicators: F1 and mIoU.

Considering the widespread application of infrared technology in military, security, medical, and other fields, improving the transmission quality and reliability of infrared system signals has become an urgent problem that should be solved. In this study, the application of temperature compensation technology in infrared systems for packet loss is investigated. A method is proposed for detecting packet loss via asynchronous triggering judgment, and systematic analyses, measurement, and verification are performed. This study provides a new solution and technical method for improving the signal transmission quality, signal integrity, and reliability of infrared systems.

A mechanical cryocooler is the main equipment used to provide the low-temperature environment required for an infrared detector. The temperature of the cold chamber is lower than that of the IRFPA as measured by the diode. The temperature difference is related to the temperature measurement position of the diode and heat transfer resistance. In this study, the influence of the diode position and difference in the heat transfer resistance on the actual temperature of the cold chamber is theoretically analyzed and experimentally verified. The influence of the actual temperature of the cryocooler cold chamber on its performance is further examined via theoretical derivation and experimental testing. During the testing of the performance of the cryocooler with the Dewar test, the following points should be considered. (1) An inherent temperature difference exists between the position of the diode and the cold chamber of the cryocooler. The greater the temperature difference, the lower the actual temperature of the cold chamber, and the worse the performance of the cryocooler and test Dewar. (2) The temperature difference between the diode and cold chamber of the cryocooler is affected by the thermal resistance of the Dewar module. The greater the thermal resistance, the greater the temperature difference and the lower the actual temperature of the cryocooler chamber. (3) The position of the diode affects the thermal resistance between the diode and cold chamber, thus affecting the temperature of the cold chamber of the cryocooler. The position of the diode should be fully considered in the design of the Dewar test to avoid the influence of simulation distortion on the evaluation of the cryocooler performance.

A temperature compensation and calibration method of the thermal infrared radiometer array sensor were introduced to improve the measurement accuracy. A maximum temperature calibration model were established based on the matrix IR temperature sensor, and a distance uniform method were introduced in this paper. By using the method, measurement error can be calibrated without the limitation of the angle and distance of the IR array sensor. Test results showed that this method can reduce the influences of the angle or distance of the measurement, and improve the accuracy.

To optimize and realize the rapid identification of three-dimensional features of road aggregates in complex environments, in this study, an infrared image three-dimensional reconstruction method is proposed based on temperature-assisted enhancement of aggregate profile. First, a 3D model library is reconstructed by calculating the point-cloud data of the aggregates with different key sieve holes. On this basis, the particle size characteristics of the 3D model, OBB enveloping box characteristics, and recognition of aggregate pinflake particles are evaluated and analyzed. The results show that the correlation coefficient between the 3D model particle size and aggregate particle size reaches 0.9916, and the OBB scale feature can reflect the 3D size information of the aggregate with a small fitting error. Compared with the traditional two-dimensional identification method, the accuracy of identifying needle flake particles based on the OBB scale feature is improved by 12%, providing new ideas for the practical application of road aggregate feature recognition.