Vision-based early warnings against civil drones are crucial in the field of public security and are also challenging in visual object detection. Because conventional target detection methods built on handcrafted features are limited in terms of high-level semantic feature representations, methods based on deep convolutional neural networks (DCNNs) have facilitated the main trend in target detection over the past several years. Focusing on the development of civil drone-detection technology based on DCNNs, this paper introduces the advancements in DCNN-based object detection algorithms, including two-stage and one-stage algorithms. Subsequently, existing drone-detection methods developed for still images and videos are summarized separately. In particular, motion information extraction approaches to drone detection are investigated. Furthermore, the main bottlenecks in drone detection are discussed. Finally, potentially promising solutions and future development directions in the drone-detection field are presented.

Inspired by the contrast mechanism of the human visual system (HVS), this study proposed an improved high boost filter (IHBF)-based enhanced local contrast measurement method for solving the low detection rate of infrared (IR) small targets with a non-homogeneous background. First, based on the frequency characteristics of the small target, the IHBF operation was used to discard the low-frequency signal containing the background. An enhanced local contrast measure method was proposed to construct the contrast operator of the ratio-difference joint form. Thus, the target contrast can be enhanced further to obtain an optimal saliency map. Finally, the adaptive threshold technology was used to extract small targets. The simulation results demonstrate that compared with existing local contrast algorithms, the proposed method is better in terms of detection rate and false alarm rate and is an effective method for detecting IR small targets in non-homogeneous backgrounds.

Considering the problem in which the existing thermal infrared target tracking algorithms have difficulty dealing with similar object interference and target occlusion, the multi-task framework in the MMNet algorithm is introduced to obtain the specific discriminant features and fine-grained features of thermal infrared targets, which are fused to identify thermal infrared objects between and within classes. In addition, the peak side-lobe ratio is adopted to dynamically set the model update parameters and obtain the target change information more efficiently, in addition to evaluating the tracking results. For unreliable tracking results, a Kalman filter was unutilized to predict the target. The experimental results on the LSOTB-TIR dataset demonstrated that the performance of the improved algorithm was optimal. Compared with MMNet, the tracking accuracy and success rate were improved by 5.7% and 4.2%, respectively. It can effectively address the challenges of occlusion and deformation and can also be applied to the field of infrared target tracking.

In the target recognition of infrared thermal imaging images, a detection algorithm based on improved YOLOv5 for infrared low-resolution targets was proposed to address the poor detection of low-resolution small targets and low detection rate of complex-scale targets. The LLVIP infrared dataset was selected and the detection effect was compared by introducing different attention mechanisms. The attention mechanism with the best effect was selected to improve the loss function of the target detection network and improve the detection rate of small targets. A TiX650 thermal imager was utilized to acquire small target image samples for optimal sampling and broadening of the original dataset, and the YOLOv5 network was trained using the improved before and after, respectively. The performance improvement of the model was evaluated from the model-training and target detection results, and the experimental results demonstrate that compared with the original training model, the improved YOLOv5 training model has a significant improvement in the detection accuracy of low-resolution small targets in the same scene of infrared imaging and exhibits a low miss detection rate.

Both infrared and visible images are widely used in the field of target detection; however, unimodal images find it difficult to satisfy the requirements of low-visibility road target detection. Therefore, this study proposes a low-visibility road target detection algorithm based on infrared-visible fusion from the perspective of bimodal fusion. First, the input images were pre-processed using various IR-visible dual-mode image fusion algorithms, five parameters, including mean, standard deviation, information entropy, mean gradient, and spatial frequency of the fused images, were quantitatively analyzed, and the detection model for low-visibility road targets was obtained by optimizing the training detection network. Finally, the accuracies of the algorithm and model were evaluated in terms of the model-training results and target detection results. The experimental results demonstrate that the false- and missed-detection rates of the model trained by the algorithm in this study were significantly reduced compared with other algorithms, and the detection accuracy was improved from 75.51% to 88.86% compared with the existing algorithm using unimodal images; in addition, the image processing speed satisfied the requirement for real-time detection.

Target recognition based on 3D features has problems, such as easy misjudgment in similar point cloud domains and large amounts of total data computation, which results in a low target detection rate and high misjudgment rate. To improve the accuracy and speed of target recognition, a three-dimensional target recognition algorithm based on infrared features was proposed. The system simultaneously obtains the 2D infrared image and 3D point cloud data of the target area, obtains the projection range of the target using the salient features of the target's infrared characteristics, and calculates the pose relationship between the system and the target. The limited range of the target in the point cloud data is calculated according to the infrared feature mapping relationship, thereby significantly reducing the total number of point clouds that need to be matched and calculated. In the experiment, the same target vehicle was tested under the same background conditions, and the recognition data for three different test angles were recorded and analyzed. The obtained results indicated that the average target detection rate of the conventional point cloud recognition algorithm, average false positive rate, and convergence time were 93.4%, 19.5%, and 4.77 s, respectively. In addition, the average target detection rate of this algorithm, average false positive rate, and convergence time were 98.7%, 1.5%, and 1.23 s, respectively. It can be inferred that the detection and misjudgment rates of the target recognition algorithm based on infrared features are more advantageous, and the processing speed is faster.

Owing to the problems of high noise and poor contrast in infrared images, the accuracy of target detection is easily reduced. Here, an improved YOLOX model combined with YOLOX and a Swin Transformer is proposed. To improve the feature extraction ability, reduce the activation functions and standardization layers of the neck and head parts in YOLOX, and optimize the network structure, the Swin Transformer is used to replace the CSPDarknet backbone extraction network in YOLOX. This study tests the improved model on both the InfiRay and FILR datasets. The obtained experimental results indicate that the improved YOLOX network has significantly improved the average detection accuracy on both datasets and is more suitable for infrared image target detection.

In an actual scenario, as the detection distance decreases, the size of the infrared weak and small targets increases dynamically. Commonly used infrared weak and small target detection and tracking algorithms cannot continue to detect and track stably. To address these problems, we propose an adaptive infrared target size change detection and tracking method. The initial screening of weak and small targets is realized with the help of a low threshold signal-to-noise ratio and circumvents the missed detection and false detection of large targets via adaptive size segmentation. Subsequently, we built an alternative target library. Finally, the Kalman algorithm model was adopted to predict the motion trajectory, complete the small-scale wave-gate detection, and realize target tracking. Compared with the DBT conventional detection and tracking algorithm, our method considers the detection and tracking of weak and small targets and large-sized targets simultaneously. In the selected scene, where the target size dynamically increases, the detection and tracking rate of the algorithm in this study is improved by approximately 10%.

With the widespread application of satellite payloads in the field of target detection and recognition, such as the atmosphere and ground objects, space infrared camera technology has been rapidly developed, which also necessitates increasingly higher requirements for the technical level of space infrared optical systems. This study analyzed and summarized the research status and development trend of space infrared optical technology by investigating the technical characteristics and changes in typical spaceborne infrared photoelectric loads at home and abroad over the past ten years.

In the field of aviation remote sensing, the two-band optical system is the most representative optical system. The dual-band system can detect both the background signal and the target signal to obtain more accurate information compared to a single-band system. Compared with the off-axial reflection system, the optical system is miniaturized while satisfying a longer focal length. Simultaneously, choosing the reflection system of a free surface as the blueprint of telefocal length system design has several advantages, including a large field of view angle, easy optical road folding, and a high system imaging quality, and can achieve high-resolution imaging and light weight design of the system. The system was added to the free-form surface for better image quality. The effective focal length of the system is 2000 mm, the relative aperture is 1/2, the field of view is 6°×1°, and the working bands are 3-5 .m and 8-2 .m. The selected model is the LA6110 non-refrigeration type detector. The design results show that the free-form surface can greatly improve the imaging quality of the system, and the modulation transfer function in the whole field of view can achieve a modulation transfer function greater than 0.3 at 14 lp/mm.

Traditional image stabilization algorithms cannot be directly applicable for video sequences with small background transformation and jitter components. In this study, a panoramic image stabilization algorithm based on the Harris image mosaic is proposed. First, the Prewitt operator was used to extract the edge information of the image, and then, Harris feature points were detected. Subsequently, both NCC and RANSAC algorithms are combined to achieve the exact matching of feature points between images, and the weighted average fusion method is used for image fusion. Finally, the fused panoramic image is clipped to complete image compensation and output a stable video sequence. The experimental results show that the improved Harris algorithm improves the algorithm efficiency and number of correct feature points. Moreover, the image stabilization algorithm in this study demonstrated good real-time performance, which can effectively eliminate video jitter and output stable video sequences.

To reduce the training samples of hyperspectral images and obtain better classification results, a double-branch deep network model based on DenseNet and a spatial-spectral transformer was proposed in this study. The model includes two branches for extracting the spatial and spectral features of the images in parallel. First, the spatial and spectral information of the sub-images was initially extracted using 3D convolution in each branch. Then, the deep features were extracted through a DenseNet comprising batch normalization, mish function, and 3D convolution. Next, the two branches utilized the spectral transformer module and spatial transformer module to further enhance the feature extraction ability of the network. Finally, the output characteristic maps of the two branches were fused and the final classification results were obtained. The model was tested on Indian pine, University of Pavia, Salinas Valley, and Kennedy Space Center datasets, and its performance was compared with six types of current methods. The results demonstrate that the overall classification accuracies of our model are 95.75%, 96.75%, 95.63%, and 98.01%, respectively when the proportion of the training set of Indian pines is 3% and the proportion of the training set of the rest is 0.5%. The overall performance was better than that of other methods.

Infrared thermal imaging technology is often used to detect internal defects in carbon-fiber-reinforced polymers; however, commonly used optical heat sources are inefficient and require the specimen to be heated over a short distance. Lasers offer the advantages of energy concentration and low-energy attenuation. As a heating source, it helps achieve long-distance nondestructive testing (NDT). This paper introduces the technology of line laser scanning infrared thermal imaging and analyzes the thermal transmission process inside the material during the heating process. Second, the uniformity and contrast of the infrared image are poor, which is not conducive to defect-feature extraction. Here, image segmentation methods were based on the intuitionistic fuzzy C-means (IFCM) clustering algorithm to extract defect edges. Compared with the K-means clustering method, this method can improve the recognition and detection of fuzzy edges of defects, retain more detailed information of images, and help extract the features of the defect edges accurately.

Here, a novel method for measuring an aspheric surface using an infrared laser interferometer is proposed. The model is divided into three modules. First, the wave front aberration between the aspheric surface and the standard spherical surface was measured by an infrared laser interferometer. Then, the theoretical value of the wave image difference between the aspheric surface and the standard spherical surface was calculated according to the aspheric equation. Finally, the surface shape deviation of the aspheric surface was calculated. To verify the validity and reliability of this method, the surface shape error of the same paraboloidal mirror was measured using the ZYGO visible light interferometer and the compensation mirror method. The obtained results indicate that these two measurement methods were identical. Hence, the novel method is convenient, fast, and highly versatile and can be used to test aspheric surfaces during processing.