Visual object detection and tracking methods have always been one of the most popular research directions.In recent years,with the rapid development of UAV technology,the use of UAV aerial photography technology for target detection and tracking has become a research hotspot.As for UAV target detection,the complexity and challenges of UAV aerial photography target detection and tracking tasks are discussed,and the target detection algorithms based on deep learning are introduced in detail.In addition,as for UAV target tracking technology,two discriminative tracking algorithms based on correlation filtering and deep learning are discussed in depth.Finally,the prospect of target detection and tracking technology applied to UAV field is summarized.

Digital interface is the main platform of Human-Computer Interaction(HCI).Scientific and reasonable interface design can promote efficient and harmonious HCI.The evaluation of the effectiveness of digital interface HCI can effectively save time resources and development costs,and lay a solid foundation for the design of the interface.Using eye tracking technology to evaluate the effectiveness can objectively reflect the users visual psychology and cognitive thinking process.Based on the evaluation mechanism of this technology in human-computer interface interaction,this paper summarizes the current research status,and points out that its future application and development trend will focus on algorithm optimization,model calculation,multiple fusion and insensible interaction.

Oriented to the space-air-ground integration battlefield information collection scenario,a channel allocation and device matching model with the optimization objective of maximizing system efficiency and performance is established.There are correlation and coupling between channel allocation and device matching decision variables.In view of the characteristics,a two-stage resource allocation algorithm based on genetic algorithm and deep reinforcement learning algorithm (GD2A) is proposed.The effect of hyper-parameters on the performance and convergence rate of the GD2A algorithm is analyzed through simulation validation.Comparative experiments show that the GD2A algorithm ensures the reliability and security of military communications and maximizes the system utility in a 6G space-air-ground heterogeneous network.

With the great improvement of the anti-jamming capability of the seeker,the probability that missiles break through "soft killing" defenses is increasing.There is an urgent need to develop an active interceptor system that can intercept and destroy the threats that have broken through the "soft killing" defenses,so as to improve the survivability of the aircraft.In view of the current situation of indeterminate orientation of the incoming threats,fast flight speed and short interception windows,this paper studies the key algorithms involved in multi-target interception decision-making,including shooting list generation,interception firepower assignment,real-time calculation of interception windows,automatic delivery instruction generation and other interception decision-making algorithms.It is proved that the designed active interception decision-making algorithms are feasible through closed-loop simulation of multi-target interception.

When a UAV is about to land on the deck of a carrier based on computer vision technology,it is necessary to consider whether its flight trajectory is reasonable and whether the motion of corresponding visual feature points in the image space is reasonable,which means the feature points should be in the field of view of the airborne camera.In addition,due to the influence of external disturbances such as sea wind and waves,it is difficult to obtain suitable parameters to make the visual feature points move along the desired trajectory and make the image errors converge in a desired dynamic process.To solve these problems,a visual guidance method is designed,which makes the trajectory planning based on the strategy of tau coupling compatible with Image-Based Visual Servoing (IBVS).Meanwhile,Prescribed Performance Control (PPC) theory is introduced to explicitly describe the dynamic process of image error convergence,and to endow parameter tuning with an explicit physical meaning,thereby reducing the difficulty of parameter tuning.Finally,the feasibility and effectiveness of the proposed method are verified by the simulation experiments.

A sliding mode adaptive iterative learning control algorithm is proposed for trajectory tracking of a quadrotor UAV under external disturbances.The designed controller consists of two parts.The first part is the sliding mode sub-controller,which is taken as the feedback controller of the quadrotor UAV system.The second part is the iterative learning sub-controller,which is taken as the feedforward controller of the quadrotor UAV system.In the iterative learning process,the control gain of the iterative learning sub-controller can be adaptively changed according to the trajectory tracking error,and this can improve the trajectory tracking performance of the quadrotor UAV.Finally,the effectiveness of the algorithm for the trajectory tracking of a quadrotor UAV is verified by software simulation.

In aircraft path planning,the traditional Sparrow Search Algorithm (SSA) has such shortcomings as complex calculation process and the tendency of falling into local solutions.To solve the problems,an SSA that combines Whale search factor with Cauchy-Gaussian variation,called WSSA,is proposed.Firstly,the population is initialized with the help of the best point set method.Secondly,the adaptive weight iteration factor is formed by using the spiral search factor in the whale algorithm,which improves the global search ability of the algorithm and ensures local convergence.Then,the Cauchy-Gaussian variant factor is added to the follower position update to perturb the original iteration to improve the global optimization ability of the algorithm.Finally,based on the comparison of model solving and fitness before and after improvement,the effectiveness of the improvement is verified.

To address the problems of weak anti-interference capability and insufficient reliability of single-moment air target threat assessment,a multi-moment aerial target threat assessment model based on the improved D-S evidence theory is established.Firstly,the time range of multi-moment aerial target threat assessment is defined based on the air combat timeline.Then,on the basis of single-moment aerial target threat level probability assignment,the D-S evidence theory is used to fuse the evidence information of each moment.Meanwhile,to address the problems that the D-S evidence theory cannot handle high-conflict evidence and its existing improvement method is computationally intensive,the concept of offset degree is introduced to determine the weight of evidence sources at each moment and perform D-S fusion on the weighted evidence.Numerical examples show that the model algorithm has low complexity and high stability,and can effectively handle fluctuating data.The algorithm can attenuate the adverse effects of high-conflict evidence fusion on threat assessment,providing a more accurate discriminatory basis for the final decision-making.

Laser active detection is a technology to detect and identify enemy reconnaissance equipment by using the "cat′s eye effect" of optical system.Its core technology is to pinpoint the location of enemy optical detector,so as to implement interference or other measures to the detector.The basic model of laser transmission in laser active detection is set up,and the influence of several conditions,such as atmospheric attenuation,defocusing amount and incident angle,on laser active detection is discussed.Based on the photoelectric echo image of laser active detection,as well as the laser transmission characteristics in actual detection process,the three criteria of echo power,target size and target shape are creatively proposed and combined with the traditional target detection algorithm,and the accurate identification and positioning of enemy detection equipment is realized.The testing via examples verifies that this algorithm has better recognition ability than the traditional algorithm does,and has strong universality and certain value in actual combat applications.

In the ADS-B receiving module,the detection peak value of low-power signal matching filter output signal is low,and the packet receiving error rate is high.To solve the problems,the pilot synchronization algorithm is studied and improved.Through theoretical analysis,the parameters of the matched filter are reasonably designed to improve the receiving performance,and the leading pulse sequence detection method based on cross-correlation detection is verified.The time delay model of the matched filter cross-correlation detection is optimized so that the sampling time is the maximum.The simulation results show that,compared with the original module-matched filter,the designed filter can effectively increase the peak value at low power and reduce the packet error rate.

To overcome the shortcomings of traditional methods in solving three-dimensional UAV path planning,such as high planning costs,poor accuracy and prone to obtain a local optimum,a three-dimensional UAV path planning algorithm based on Q-learning arithmetic optimization algorithm is proposed.In order to improve the optimization accuracy of the arithmetic optimization algorithm,the Circle chaotic mapping is introduced to improve the diversity and distribution uniformity of the initial population.Q-learning is introduced to adaptively adjust the updating of the acceleration function.The global searching and local development of the algorithm are made balanced.The perturbations in the optimal solution’s neighborhood are designed to optimize the global searching capability.By establishing a three-dimensional UAV path planning model,the path planning is transformed into a multi-objective function optimization problem,and the improved algorithm is used to solve the three-dimensional UAV path planning problem.The trajectory objective function that comprehensively considers the trajectory cost,the terrain cost and the boundary cost is used to evaluate the fitness of the particles,and the path planning is optimized through iterations.The simulation results show that the trajectory obtained by the proposed algorithm has lower total costs and the stability to adapt to different complex terrain environments.

In order to improve the recognition accuracy of Unexploded Ordnance (UXO) targets on the ground by UAVs in complex environments,a UXO target detection method based on the improved YOLOv5 is proposed.On the basis of YOLOv5,the method improves the loss function of the original YOLOv5 network to improve the recognition accuracy of UXO targets.At the same time,the method adds an attention mechanism,improves mosaic data enhancement,and improves the prediction frame screening mechanism to improve the recognition efficiency of UXO targets,and realizes the detection of UXO targets in air-to-ground scenarios with better accuracy and speed.Experimentally,multiple UXO datasets in different complex backgrounds are selected,labeled and trained to obtain UXO target models.Then,the correctness of the algorithm and model is evaluated from the perspectives of model training results and target detection results.The experimental results show that:1) The model obtained by NGG-YOLOv5 has a significant improvement in detection accuracy and detection speed in comparison with that obtained by the original YOLOv5,with an increase in accuracy from 78% to 91% and an increase in mean Average Precision (mAP) from 50% to 56%;and 2) It can effectively detect UXO targets in four kinds of complex backgrounds,with a low missed alarm rate.

How to compress the High Dynamic Range (HDR) data collected by the infrared detector into the observable low-dynamic-range image with outstanding details,high contrast and low background noise while retaining the original image information as much as possible has always been a difficult aspect of infrared technology.In order to solve this problem,an infrared image enhancement mapping method based on the fusion of multi-scale multi-histogram mapping results is proposed.Firstly,the global linear transform and the multi-scale window CLAHE are used to map the HDR infrared image separately,and then adaptive weight calculation is conducted based on local saliency and dynamic-range features,and the multi-histogram mapping results are fused to solve the scene adaptability problem.At the same time,the fusion weight is optimized by using guided filtering in the weight calculation,and finally the adaptive detail enhancement is realized by using the Gaussian kernel function in the gray domain.The experimental results show that,compared with other methods,the proposed method can effectively enhance the image contrast,enrich the image details and reduce the background noise,and has strong scene adaptability.

Image haze removal is an important issue in the field of image processing.Deep learning can effectively improve image clarity,but due to the lack of corresponding real haze matching data pairs in the training process,synthetic haze is usually used as the dataset.The existing synthetic haze mostly depends on such parameters as depth information and atmospheric scattering coefficient.To solve the problems of color distortion and incomplete haze removal caused by training on such a dataset,a synthetic haze method based on Cycle Generative Adversarial Network (CycleGAN) is proposed.Through the network,mismatched data pair training is conducted to learn the features of haze images,then real haze features are added to clear pictures,with which matching data pairs are formed,and finally such datasets are used for dehazing training.The results show that these datasets can effectively solve the problems of color distortion and incomplete haze removal.

In the one-stage object detection based on deep learning,the bounding box regression loss based on the Intersection over Union (IoU) is not sensitive enough to the change of the position relationship of the bounding box.The existing loss cannot accurately distinguish different inclusion relationships between the predicted frame and the true value frame.In response to the above problem,a loss based on Regression Position Relationship Sensitivity IoU (RPIoU) is proposed.This loss design can strengthen the sensitivity to the relative positional relationship between the predicted frame and the true value frame.Firstly,a penalty term is added after the IoU to make the corners of the two frames infinitely close.It solves the problem of IoU degradation when the center points coincide.Secondly,the exponential function taking the ratio of the area of non-overlapping region to that of the true value frame as the parameter is introduced as the penalty term,which can deal with the problem that the loss cannot distinguish different inclusion relationships between the predicted frame and the true value frame,and can guide the position of frame regression more accurately.Considering that each part of the total loss of the one-stage object detection algorithm has different degrees of contribution to the training results,this paper takes the mean Average Precision (mAP) as the fitness function,and uses the genetic algorithm to optimize the total loss of training to obtain the optimal weights of classification,regression and confidence loss respectively.The designed loss is applied to the one-stage object detection algorithm YOLOv5,which is verified on the public visible light dataset VisDrone and the self-made infrared aircraft dataset respectively.On the public visible light dataset,the mAP reaches 0.447,which is 0.037 higher than that of the original YOLOv5.On the infrared aircraft dataset,the mAP reaches 0.966,which is 0.014 higher than that of the original YOLOv5.

To address the problems of inadequate precision and unstable deception angles of cross-eye jamming,a new method of angular deception jamming based on the three-tuple antenna is proposed.The method adopts the three-tuple structure to improve the focusing performance of the linear array,significantly reduces the deception disturbance of the symmetric peak of the linear array,thus obtaining a more stable and accurate deception direction.At the same time,this method uses adaptive genetic factor optimization to form stable deception effects by constraining the waveform and the direction of arrival,and increases the coverage of angle deception.The effectiveness of this method is verified through simulation results.

In order to solve the problem that mobile robots have strict requirements on operation environment,a light and small spherical land-air amphibious robot is designed,which can be applied to variable and narrow environments.It has two motion modes of rolling on the ground and flying in the air,and it can complete the switching of motion modes autonomously.Firstly,the structure of the robot is designed,including the rolling part,the flight part and the mode-switching part.Based on the Lagrangian method,the kinetic model in different motion modes is established,and the kinematic model based on the Euler equation is established.Finally,according to the motion models of the robot,the attitude controllers in flight mode and in ground-rolling mode are designed,and the comparative simulation experiments of the controllers are conducted.The tests were conducted on a real robot.The experimental results show that the robot can achieve autonomous mode switching and stable motion in different motion modes with controllable attitude,and has certain obstacle-crossing ability in pure ground mode.The structure,the control method and the control strategy switching method are useful for the land-air amphibious deformable robot.

In order to study the issue of polarization state transmission in parallel phase-shifting coaxial digital holographic system, a modified model of parallel phase-shifting coaxial digital holographic system considering polarization is proposed, and a polarization imaging method is proposed based on this model.Firstly, the principle of parallel phase-shifting coaxial digital holography under ideal conditions is introduced.Then, the modified model of parallel phase-shifting coaxial digital holography considering polarization is introduced.Finally, based on this model, a polarization imaging method is proposed.The experimental results verify the reliability of the model.The appearance information can be obtained by using this method, and the imaging resolution can reach 14 μm.This method expands the functions of digital holographic system.Through this system, the polarization state of the object can be imaged.

As a classical instrument for measuring the infrared radiation intensity of the aero-engine,the infrared spectral radiometer has the advantages of high spectral resolution,high measurement precision,fast response,wide spectral range and so on.However,the infrared spectral radiometer needs to select an appropriate lens according to the size of the measured target and the measuring distance.The infrared radiation intensity of the aero-engine can be obtained by multiplying the measured spectral radiance by the effective field-of-view area of the lens.Due to the processing and assembly errors of the manufacturer,the field-of-view angle of the lens of the infrared spectral radiometer will be deviated from its default value,leading to testing deviation.In order to measure the field-of-view angle of the lens of the infrared spectral radiometer,an effective method is designed and a measurement device is made,which is used for measuring the effective field-of-view angle of the standard lens of the VSR-3 infrared spectral radiometer.The measurement results show that the effective field-of-view angle of the lens presents a 4% deviation from the factory default value.Through deviation correction,the precision of infrared radiation intensity measurement of the aero-engine can be improved.