In addition to RGB images, thermal IR images can be used to extract salient information, which is crucial for salient object detection. With the development and popularization of IR sensing equipment, thermal IR images have become readily available, and RGB-T salient object detection has become a popular research topic. However, there is currently a lack of comprehensive surveys on the existing methods. First, we briefly introduce machine learning-based RGB-T salient object detection methods and then focus on two types of deep learning methods based on CNNs and vision transformers. Subsequently, relevant datasets and evaluation metrics are introduced, and both qualitative and quantitative comparative analyses are conducted on representative methods using these datasets. Finally, challenges and future development directions for RGB-T salient object detection are summarized and discussed.

With technological developments, complex optoelectronic application environments have increased the requirements for optical materials and their properties. Due to their excellent comprehensive performance and relatively mature preparation technology, optoelectronic applications of sapphire have attracted considerable attention. Sapphire has high spectral transmission performance in multiple spectral bands such as ultraviolet, visible light, and mid-wave IR, and maintains relatively stable optical performance under complex usage conditions. It is frequently used in hypersonic missile domes, airborne IR optical windows, optoelectronic mast windows, and high-end civilian optical components. Doped sapphire can be used as an optical component in lasers and thermoluminescence detectors, and sapphire fibers can be used in high-temperature sensors. The basic characteristics of sapphire crystal materials and their research progress in optoelectronic applications are summarized, and future development trends are discussed.

To meet the needs of low-light detection with large field-of-view astronomical telescopes in high-altitude areas in our country, this study utilized a high-performance scientific CMOS (sCMOS) detector, GSENSE1516BSI, from Chang Guang Satellite Technology Co., Ltd. The imaging system, which is controlled by a Xilinx Kintex-7 FPGA, was designed to be low-noise, highly sensitive, and to utilize thermoelectric cooling (TEC) technology for active cooling. The hardware components encompass the peripheral circuitry of the sCMOS detector, FPGA readout circuit, and TEC active cooling module. The software includes a timing control module, a data alignment and reception training module, and a DDR3 high-speed cache module. In addition, a large-array pixel data DDR3 read-write validation module was developed to validate the reliability of storing large-array pixel data in the DDR3 submodule during the design phase. This early validation helped identify potential data transmission issues, thereby optimizing the system testing approach. Ultimately, comprehensive testing confirmed the stable performance of the designed sCMOS imaging system with a readout noise of 4.33e- and a cooling temperature of -30℃; the dark current at -30℃ is 0.15 e-/pixel/s. It can stably read 4 k×4 k large array pixel data, satisfying the astronomical observation requirements of large-field telescopes.

Modeling and simulation have wide applications in the performance design, test evaluation, and combat simulation of IR detection systems. To meet the need for the performance modeling of IR detection systems in complex combat scenes, the effects of the detector array structure, sampling efficiency, image blur, and other factors on detection were analyzed based on the traditional geometric model, and the effects of clutter and human interference on detection in complex scenarios was analyzed. All these factors were introduced into the performance modeling of the IR detection system, and an IR detection system model that can be applied to complex scenes was established. Simulation examples show that the model can realistically simulate the performance characteristics of IR detection systems in complex combat scenes and has important application value in IR detection system design, performance evaluation, and combat test simulation.

To address issues such as slow response time, complex calibration, and limited suitability for special environments in high-speed temperature measurements, a refrigeration-based high-speed IR temperature measurement system is designed with an STM32F429IGT6 microcontroller as the core controller and an optical system combined with a refrigeration-type IR detector as the main components. The optical system employs a metal parabolic reflector to construct the optical components. In conjunction with a refrigeration-type IR detector, photoelectric conversion is accomplished with high-speed data acquisition performed by the microcontroller. The system has a sampling rate of 1 MHz, utilizes an SD card for data storage, and displays the temperature-change waveforms on the screen. Temperature calibration is conducted using the blackbody furnace method, and a least squares fitting is applied to establish the voltage-temperature relationship formula, resulting in a measurement error of ±1℃. The effective temperature measurement range is 10-200℃. Following calibration, the system underwent Hopkinson pressure bar impact testing for practical application. The test results indicated that the refrigeration-based high-speed IR temperature measurement system possesses a microsecond-level detection capability for temperature changes, with a fast response time, simple operation, and calibration, thus demonstrating its promising applications in the field of dynamic high-speed temperature measurements.

To address the problem of low accuracy and slow recognition speed of infrared (IR) image target recognition of substation equipment in complex backgrounds, this study proposes an IR image recognition algorithm for substation equipment based on the improved YOLOv7-Tiny. First, the lightweight bottleneck structure GhostNetV2 bottleneck is introduced to replace a part of the CBS module and build a lightweight and efficient aggregation network known as a lightweight-efficient layer aggregation network. Simultaneously, a coordinate attention mechanism is embedded in the feature extraction stage to reduce the number of network parameters while strengthening the network's extraction of key features of the target and improving detection accuracy. The network coordinate loss function is replaced by SIoU~~Loss to improve the anchor frame positioning accuracy and network convergence speed. The results show that the accuracy of the improved network is 96.28%, the detection rate is 26.42 frames/s, and the model size is reduced to 7.82 M. Compared with the original YOLOv7-Tiny algorithm, the detection rate is increased by 21.69%, the identification accuracy is improved, and the model size is reduced by 36.89%. These results meet the requirements of accurate real-time identification of substation equipment and lay a foundation for subsequent substation equipment fault diagnosis.

Hyperspectral images contain rich spectral information, and multispectral images have exquisite geometric features. More comprehensive remote sensing images can be obtained by merging high-resolution multispectral and low-resolution hyperspectral images. However, most existing fusion networks are based on convolutional neural networks. For remote sensing images with complex structures, convolution operations dependent on the kernel size tend to lead to a lack of global context information in the feature fusion stage. To ensure the quality of image fusion, this study proposes a convolutional neural network (CNN) combined with a multi-scale transformer network to realize multispectral and hyperspectral image fusion, combining the feature extraction capability of the CNN and the global modeling advantage of the transformer. The network divides the fusion task into two stages: feature extraction and fusion. In the feature extraction stage, different modules are designed for feature extraction based on the CNN. In the fusion stage, a multi-scale transformer module is used to establish a long-distance correlation between local and global information, and the features are mapped into high-resolution hyperspectral images through multilayer convolution layers. Experimental results on the CAVE and Harvard datasets show that the proposed algorithm can improve the quality of fused images better than other classical algorithms.

Efficient and rapid identification of substation equipment is a crucial part in substation safety status detection. To better fit the complex backgrounds and different target devices of the substation, the PSA module is introduced on the basis of YOLOv7 to realize the information interaction between local and global channels and improve the recognition accuracy of the model for devices of different scales. Combined with PConv and GSConv, a lightweight network is established to accelerate the inspection progress while ensuring model accuracy. Using Dyhead to embed three perceptions into a target detection head, improving target detection capabilities. Compared with the original YOLOv7 algorithm, the accuracy is improved by 3% and the model is reduced by 10%, which meets the requirements of efficient and rapid identification of substation equipment and provides a basis for subsequent substation equipment fault diagnosis.

To address the problems of complicated data and low detection accuracy for deep-learning target detection of IR images of power equipment in complex environments, this study proposes a convolutional block attention module (CBAM) based on YOLOv7 to improve the classification algorithm for IR images of power equipment. First, the existing dataset is labeled and divided into training, validation, and test sets in a certain proportion and then introduced into the backbone network of YOLOv7 to enable the model to emphasize the region of interest and suppress useless information. Second, the divided dataset is put into the improved YOLOv7 for model training, and six improved YOLOv5s models are compared. The experimental results show that the improved YOLOv7 model outperforms YOLOv7, YOLOv5s, and six attention models based on YOLOv5s under the same experimental conditions. The improved YOLOv7 exhibits significantly improved performance and achieves fast and accurate IR image classification.

In complex environments such as power grids and bridges, inspection robots can effectively replace manual inspections and equipment maintenance. To achieve autonomous obstacle avoidance for inspection robots in complex environments, this study proposes an image-matching recognition technique based on the YOLOv5 deep-learning network model, which utilizes both IR and visible-light images. This enables the robot to identify and classify various obstacles, including living and nonliving obstacles. During the inspection operations, the robot is equipped with an IR camera and a regular camera to monitor its environment in real time. With the YOLOv5 network model trained on a large dataset, the robot can quickly and accurately identify and categorize obstacles along its path. The robot not only identifies the nature of obstacles but also performs appropriate proactive actions to address different situations. The average recognition accuracy is approximately 99.2%. Experimental results demonstrate the effectiveness of the comprehensive obstacle avoidance method based on multiple-image information. The robot can detect, classify, and navigate obstacles under various scenarios, thereby enhancing their autonomy and adaptability. This technology has wide applications in areas such as automated inspections, safety monitoring, and rescue missions, providing strong support for the continuous development of robotic technologies.

To address the memory and computational resource constraints of IR small-target detection under an embedded hardware platform, high-frame-rate detection demand, and higher target-level detection performance requirements, a detection network called CAMNet is proposed. The network combines the advantages of self-attentive global modeling with the lightweight and fast processing characteristics of convolution and adopts a four-stage stacked encoder and decoder architecture, which effectively reduces the algorithmic resource requirements and improves the detection frame rate. A center-of-mass loss function is proposed in terms of the loss function, which effectively improves the target-level detection performance of the algorithm. Experimental results on the public SIRST dataset show that CAMNet achieves a detection frame rate of 107 FPS on common embedded platforms. Compared with other state-of-the-art networks, such as ISTDU-Net and UIU-Net, CAMNet improves the probability of detection by at least 0.76% and reduces the false alarm rate by at least 87.30%. These findings indicate that the proposed detection network offers both fast detection speed and superior target-level detection performance.

The degrees of freedom in the motion state of aerial targets are higher, and the target motion state is more difficult to obtain. Existing methods focus on estimating the relative motion trajectory in two-dimensional space (azimuth and pitch), ignoring the interference of the reconnaissance platform's own attitude changes on the target's motion trajectory estimation, making direct application to airborne IR platform applications difficult. To address this problem, this study proposes a three-dimensional (3D) motion estimation method for airborne targets under an airborne IR platform to measure the target's motion status in all directions in the coordinate system of the northwest sky. To improve the accuracy of the target position estimation, this method introduces the target distance and the attitude of the detection platform to enhance the anti-interference performance of the IR target motion state estimation. Our method first uses a target-tracking module based on the TLD and a Kalman filter utilizing a detection-based tracking strategy. The Kalman filter is employed to alleviate the effects of target centroid jitter on the target position estimation accuracy. Second, a long-short strategic distance prediction module is proposed to supplement the target distance information not obtained by the laser rangefinder. Finally, the motion status of the target in each direction in the northwest-sky coordinate system is obtained using the aerial target motion estimation module based on prior information. Under the condition that the 3D motion information of the aerial target is known, the 2D spatial information of the target in the current reconnaissance system can be solved in reverse using this method. Experimental results show that the error in the target distance prediction result of this method is less than 50 m, and the velocity error of the northwest-sky coordinate system is less than 25 m/s. When the attitude angle of the detection system is changed, the target-tracking stability of this method is better than that of the Kalman filter.

In the laser soldering process, real-time measurement of the solder joint temperature and adjustment of the output power of the semiconductor laser are crucial for ensuring welding quality. To avoid faults such as solder scorching, virtual soldering, and false soldering caused by excessive temperature measurement errors or slow measurement speeds, a high-precision IR radiation temperature measurement device is designed. First, the principles of IR radiation temperature measurement are introduced, and the design method of the IR radiation signal conversion circuit is explained. Second, the primary signal processing method used in this study, which is a Butterworth-type infinite impulse response filter, is introduced. Finally, the performance of the device is validated through experimental analysis. The experiments demonstrate that the IR radiation temperature measurement device designed in this study is suitable for non-contact measurement of solder joint temperature in laser soldering, with a maximum error of 2℃ within the test range of 70-260℃ in a standard blackbody furnace. During the laser soldering process, the highest temperature error is less than 0.6%, making it widely applicable to the field of laser soldering.

The microchannel plate (MCP) ion barrier film produced by traditional technology affects the signal-to-noise ratio (SNR) of the image tube to a certain extent, and long-term working conditions reduce the reliability of the image tube. Due to the urgent need to improve and enhance the SNR and reliability of image tubes caused by defects in traditional ion barrier films, a new type of ion barrier film for MCP must be developed. In this study, a continuous high-quality U-shaped ion barrier film was prepared using atomic layer deposition on the inner wall of MCP channels with high aspect ratios and channel holes at the input end. The film quality and electrical performance were tested using an MCP comprehensive detection device and compared with the performance of MCP components prepared by traditional technology. The new ion barrier film is dense, with high gain values and low carbon content in the components. The new MCP components exhibit a high SNR and high lifespan reliability after tube preparation. New MCP components produced using this technology are of great significance for improving the SNR and reliability of image intensifiers.

To further study the nondestructive testing technology of line laser scanning IR thermography, the effects of adjustable parameters and tested material parameters on the thermal imaging results was explored to obtain the best defect detection results. COMSOL software was used to establish a finite element analysis model of line laser scanning carbon fiber composite materials, and the maximum surface temperature difference between the defect center and non-defect area was selected as the characteristic quantity to measure the detection effect. The influence of the linear laser length, defect size, and defect depth on the thermal imaging results of carbon fiber reinforced plastics (CFRP) was analyzed, and the relationship between the three and the detection effect was fitted. This study provides a reference for the establishment of a laser scanning thermal imaging detection system and accurate and reliable detection standards.