Under hypersonic conditions, the existence of aero-optical effects seriously affects the accuracy of infrared imaging guidance, which has become one of the key technical problems in the development of a new generation of hypersonic precision strike weapons. It is necessary to study the optical aperture with supersonic film cooling suitable for hypersonic flight environment so as to solve the thermal protection and imaging problems of optical aperture under hypersonic flight conditions and break through the speed and temperature limitations of existing infrared imaging guidance weapons. Finally, it provides support for the realization of weapons' high-precision strike against air, ground, and sea under hypersonic conditions.

Aiming at the problems existing in long-distance infrared target detection, such as less feature information, complex environment, more noise interference, and high missed detection rate and false alarm rate of traditional target detection algorithms, an infrared small target detection algorithm based on dual-channel feature enhancement attention network is proposed in this paper. The overall network structure mainly includes three parts: dual channel feature extraction module, feature enhancement module and integrated top-bottom attention module. Compared with single channel feature extraction, dual channel feature extraction can obtain more feature information. Feature enhancement module can enrich target features further. Moreover, the integrated top bottom attention module can adaptively enhance target features and weaken background noise. And then the algorithm improves the detection effect of dim and small targets in the infrared images. Finally, it is verified that the algorithm proposed in this paper has a better detection effect, and has a lower rate of missed detection and false detection.

In order to solve the problem of complex structure and high cost of traditional multi-mode target source, three-mode target simulation system for generating microwave signal, infrared signal and laser signal using beat frequency modulation technology is proposed in this paper. The microwave signal can be generated by optical heterodyne method, and the modulation information can be modulated by the arbitrary waveform generator to achieve the microwave signal with different frequency, wave, repetitive frequency and pulse width. The laser signal and infrared signal conversion module can be used to convert the laser signal to the infrared signal, and the temperature tuning of the infrared signal can be achieved by adjusting the output power of laser signal. The seed source laser is used to generate the laser signal, and the laser signal with modulation information can be generated through the arbitrary waveform generator and the modulator. The generated microwave signal can reach up to 40 GHz, at the same time, it can generate infrared signals with the temperature of 230℃ and laser signals with 10Gbps modulation information. The target source simulation system can realize three emission modes of microwave signal, infrared signal and laser signal. Furthermore, the structure is mainly composed of optical fiber devices, which has the advantages of low cost, simple structure and strong anti-interference ability.

Aiming at the lateral-directional control problem of solar-powered unmanned aerial vehicle (UAV) with normal configuration, a moving-mass control scheme with battery as a mass slider and slide rail arranged along the spanwise is proposed. Firstly, the motion model of moving-mass solar-powered UAV is established. Then based on the classical PID control theory, the rolling angle controller and the yaw angle controller are designed respectively with the slider offset as the input. The simulation results show that when the mass of the slider is certain, the moving speed and maximum stroke of the slider are the main factors affecting the control effect of the moving-mass control scheme, and the sliding stroke mainly affects the manipulability and overshoot of the control system, and when the slider speed is too slow, the delay of the control system will be obvious, and even the system will be diverged.

With the rapid development of new photodetectors, optical materials, signal processing and MEMS techno-logy, the photoelectric precision guidance technology has gradually become one of the most important technologies in the field of precision guidance today, and also lays the foundation for the development of a new generation of light photoelectric guided missile. In this paper, the typical models of light photoelectric guided missile in foreign military are summarized, and the technology development status is introduced. The key technologies in missile overall design technology, detection guidance and control technology, fuse-warhead coordination technology, and motor system technology are analyzed in detail. Finally, the paper envisages the development trend of light photoelectric guided missile.

In this paper, the tracking differentiator is used to extract LOS angular rate of strapdown seeker. Firstly, the mathematical mechanism of extracting LOS angular rate of strapdown seeker by tracking differentiator is described. Then, a method of fast obtaining the frequency-domain characteristics of tracking differentiator by using the rotating equivalent force is proposed. Based on the frequency-domain characteristics, the parameter design and performance analysis of two typical tracking differentiators are carried out. Finally, based on the above design results, modeling simulation and calculation analysis are carried out to verify the LOS angular rate decoupling effect of tracking differentiator.

For the problem of multi-constrained trajectory planning of gliding vehicle in the unpowered glide phase, a set of fast solving algorithms is presented based on convex optimization method. The algorithm establishes an extended dimension state model by introducing control derivative, obtains discrete optimization problems based on piecewise Gaussian pseudo-spectral method, and accelerates the convergence solving process by relying on three designed target penalty functions. The simulation results show that the algorithm can quickly converge to a set of feasible solutions of the original problem at any initial value that does not satisfy the constraint, and the computational efficiency is higher than that of the traditional Gaussian pseudo-spectral method.

In the flight of spinning missile, the change of engine thrust has a significant impact on the response characteristics of the missile, which will reduce the stability of the missile and lead to the non-convergent conical motion of the missile. To solve this problem, a cross-coupling dynamic model of spinning missile considering thrust variation is established, the mechanism of dynamic stability reduction caused by thrust variation is analyzed, and the design stability boundary conditions of thrust change rate are given. Numerical simulation verifies the correctness of the proposed method. Within the designed boundary of stability, the rapid change of thrust from large to small will lead to the reduction of the dynamic stability of the missile, and once it exceeds the design boundary, the missile will diverge in the form of conical motion.

Deep neural network training needs a large number of sample data, but for infrared ship targets, the sample size of infrared ship targets with different types and perspectives is small and difficult to collect, which makes it very difficult for deep learning training. In order to reduce the demand for real infrared ship target data in deep learning, a large number of simulated infrared ship images and a small number of real infrared ship images are used as samples for training. In order to solve the problem of cross domain adaptability between simulated infrared ship image and real infrared ship image, the feature adaptive method from coarse to fine is used to realize the cross domain target detection function. Experimental results show that the proposed algorithm has high detection accuracy for real infrared ship targets.

The star sensor distortion calibration method mainly adopts the fitting method, which is limited by the number of calibration points and errors in the engineering application. The traditional least square fitting method or the toolbox fitting method will produce over-fitting phenomenon in the fitting process, resulting in a decrease in the accuracy of the star sensor's distortion calibration. This paper proposes a high-precision distortion calibration method based on the Dropout method. This method first networkizes the high-order surface distortion model of the star sensor, and then constructs the distortion model of the star sensor with part of the convolutional layer hidden. Recently, supervised learning is performed to complete the calibration of star sensor distortion model. The test results show that the use of the star sensor calibration method based on the Dropout method can effectively improve the training accuracy of the star sensor. Compared with the fitting results of the high-precision toolbox, the distortion calibration residual is increased from 2.02″ to 1.12″.

Orbital angular momentum beam has been widely used in communications, military and other fields in recent years because of its anti-electromagnetic interference, high security and multimode multiplexing. Firstly, this paper reviews the theoretical background of orbital angular momentum beam in electromagnetism, discusses the characteristics of orbital angular momentum of Laguerre-Gauss beam, and reviews the method of orbital angular momentum screening. A solution to the problem that the scattered light field changes from single mode to superposition state is proposed. Secondly, combined with the CGH technology, a single Laguerre-Gaussian beam and a multi-mode superposition state Laguerre-Gaussian beam are generated and applied to the whole screening system. Finally, the basic idea of spatial coordinate transformation is introduced, and the mode of orbital angular momentum is decomposed based on the coordinate transformation method. Through the orbital angular momentum beam screening experiment, the distortion of orbital angular momentum under turbulent scattering is given. The experimental results show that with the increase of the turbulence intensity, the dispersion between the orbital angular momentum states increases, and the strong turbulence causes the distortion of the orbital angular momentum multiplexed beam.

In order to study the infrared radiation characteristics of smoke interference from diesel combustion, first, we introduce the collection experiment of the smoke produced by combustion, and obtains the morphological characte ristics of the smoke particles. In the establishment of the infrared radiation transmission model, the physical parameters of a single particle combined with the physical parameters of the particle system are used to calculate the extinction, absorption and scattering characteristics of the particle system, and then we use the method called Planck average to compute the spectral characteristics of the system. Then we use the Euler method to divide the entire flow field into a structured grid and then solve the smoke flow field in the discrete process. Finally, by combining the infrared radiation transmission model with volume rendering technology, we render the smoke from diesel combustion in the infrared band. In the end, we simulate the diffusion of smoke at different times, and the shielding of the blackbody by the smoke in diffe- rent wavebands and different wind fields.

In order to enhance the visual effect of dual light fusion and improve the recognition of fused images by human eyes, this paper proposes infrared-visible light image fusion methods based on hue-saturation-value (HSV) color system. According to the relationship between the information of each channel of HSV system and color perception physiology, the gray-scale infrared image and pseudo-color coded infrared image are mapped with the visible light image in hue, saturation and value channel. The HSV information of the fused image is obtained by establishing the fusion mathematical model of each channel through brightness adaptation. Finally, the RGB computer color display image of the fused image is obtained by color model transformation. The results show that methods proposed in this paper can retain the key information in infrared and visible light images when realizing the fusion of infrared and visible light, and complete the information complementation.

The gradient characteristics of dark targets is analyzed first in this paper. The gradient of infrared target is isotropic, which can be used as the basis for the design of target detection algorithm. Besides, the imaging of medium and long wave infrared images is analyzed. It is believed that long-wave infrared imaging is less affected by target reflection. On this basis, preliminary target detection of the pyramid decomposition image is carried out using multidirectional gradient algorithm. Then, the antisotropic diffusion method is used to further distinguish between the target and the background. Finally, the detection results of medium and long infrared images are fused base on the characteristics of dual-band infrared imaging. Effective detection of dual-band infrared dark targets is realized.

Dynamic vision sensors, with microsecond-level time resolution and low-delay characteristic, have a great application value in the challenging scenarios with high speed and high dynamic range. In order to compare the differences of the different detection algorithms in event detection, the integral model and Leaky Surface model are used to handle the output events. In addition, two kinds of event-based object detection algorithms are listed, that is, event-based feature detection algorithm and event-based convolutional neural networks (CNN) detection algorithm. By means of object detection for MINST-DVS and POKER-DVS event dataset, the detection accuracy of the two algorithms is compared, and the advantage of event-based deep learning detection algorithm in multi-target and high-speed scenarios is verified .

With the continuous development of deep learning in the field of object detection, the detection accuracy is constantly improved. However, the neural network algorithm model has high requirements on the computing resources of the hardware platform, so it is difficult to be applied in the embedded platform. In order to ensure that the neural network algorithm can meet the high accuracy and improve the efficiency of its operation, this paper carries out the research on the object detection algorithm based on neural network model compression technique. Firstly, K-means ++ clustering algorithm is used to cluster the prior box in the data set, in order to make algorithm have a good initial value in the initial stage of training. At the same time, aiming at the problem of computing speed, this paper optimizes the YOLOv3 which based on Darknet53 frame, and prunes the network model of YOLOv3.The experimental results show that the neural network model compression technique can improve the speed of the algorithm by two times with less loss of accuracy.

In order to improve the fusion effect of different sensors, a guided filtering infrared and visible image fusion algorithm based on weight and saliency information optimization is proposed in this paper. First, the image is divided into a base layer and a detail layer through Gaussian filter. Second, the target contour is extracted as saliency information to reduce the influence of regional noise. Third, the local gradient is calculated through a sliding window to optimize the weight map construction to reduce the influence of noise points and increase the confidence of the weight map. Fourth, the guided filtering is used to process the coarse weight map to suppress artifacts and remove noise. Finally, the appropriate detail layer and base layer fusion coefficient ratio is selected to strength the detail information and complete the fusion. The experimental results show that the algorithm in this paper strengthens the performance of the fusion image in texture details, and has a certain degree of improvement in the main fusion evaluation indicators such as information entropy, average gradient, and blind image quality, compared with the selected five commonly used fusion algorithms.

When using the distributed active/passive imaging detection system to detect the same key target, it is necessary to fuse the multi-source information of the target, integrate the different positioning accuracy of active/passive imaging detector, and calculate the precise spatial position of the key target. Aiming at the problems of different imaging systems, different positioning solution models and different positioning accuracy of distributed active/passive imaging detection system, a unified target space collaborative positioning mathematical model is proposed in this paper, and the linear weighted genetic algorithm is used to solve this mathematical model to obtain the target spatial collaborative positioning results. The simulation results show that the mathematical model of target spatial collaborative positioning constructed in this paper and the target spatial positioning method based on linear weighted genetic algorithm provide a unified model and solution for the target spatial collaborative positioning problem of distributed active/passive imaging detection system. Meanwhile, it also solves the problem of target information fusion when detectors of different imaging systems have different positioning errors. During the detection mission of key target by distributed UAV cluster, distributed cruise missile cluster and distributed fighter cluster, the model and algorithm proposed in this paper can effectively improve the accurate detection, guidance, attack and interception ability of the detection system.