Abstract:Being lost,object redetection is a crucial step for retracking object for long time visually,but due to background interference and the introduction of numerous similar objects,and the performance of re-detection is poor,for this reason,a method of redetecting object is proposed based on dual modulation.First,a modulator with search frame feature is designed to enhance the correlation between search frame features and the target,thereby improving the capability of the re-detection method to handle complex background interference.Secondly,a method with proposal features is designed to enhance the response of proposal features to the target,thereby improving the capability of the re-detection method to handle inter-ference from similar objects.In order to verify the effectiveness of the method proposed in this paper,TransT and ToMP are selected as the base trackers in combination with the algorithms proposed in this pa-per form two long-term visual tracking algorithms,and the experiments are carried out on four datasets,i.e.UAV20L,LaSOT,VOT2018LT,and VOT2020LT.The experimental results show that the pro-posed method significantly improves the long-term tracking performance of the mentioned-above two basic trackers.

Abstract:Aimed at the problems to improve the confidence in ground clutter simulation for ground-based tracking radar,a simulation method in combination of statistical models with terrain matching is proposed.First,the dis-tribution patterns jammed by ground clutter at amplitude and power spectra in combination of actual deployment position of radar,and in full consideration of the impact of terrain obstruction on radar detection are analyzed,and a database of terrain profiles in various quantification directions,obstruction angles,and geographical features,is established.And then,ground clutter echo signal sequence generated is a match for the detection signal waveform through digital convolution.The simulation results show that this method requires low signal synthesis power,effectively controlling the scale of signal processing,and is still more conformed with the actual working conditions for radar under terrain obstruction.

Abstract:Because of the problems that accuracy is low in recognizing complex flight action,and in order to enhance the accuracy and reliability of flight parameter data analysis,this paper proposes a flight ac-tion recognition method based on time convolutional network mapping anchor boxes.The method is to construct a FD̠Net recognition model by improving the YOLOv3 network structure,transforming the recognition of complex flight action into a problem of regional division and classification in the time di-mension.A selection method of complex flight action with key feature parameters is proposed,and an input system with 25 feature parameters is constructed.A data augmentation method based on scale scal-ing is adopted by solving the problem of sample imbalance.Loss functions for prediction box regression,confidence regression,and classification regression are designed to complete model training.The experi-mental results show that compared with the existing methods,the proposed method significantly im-proves the accuracy of complex flight action recognition and significantly enhances computational efficien-cy,and the effectiveness and practicality of the method are verified.

Abstract:In order to reduce the risk of destructive physical tests such as short circuits,a corresponding short-circuit fault protection and power supply redundancy conversion mechanism is designed in the light of a certain of small aircraft power supply architecture.Based on AMESim,a complete power system and backend load model are established,and protection logic is injected to simulate actual faults and predict phenomena.The established power supply and electromechanical management model can simulate faults such as generator main feeder short circuit and busbar short circuit to a certain extent,and automatically execute protection logic.The results show that the interruption conversion time of the collector voltage in the short-circuit protection is less than 50 ms,and the back-end actuator and electric pump can work nor-mally under condition of non-emergency short-circuit faults.The designed protection logic is basically ap-propriate.The maximum impulse current of system short circuit is 2 898.5 A,and the maximum reverse current of motor is 38.4 A,providing reference for subsequent device selection and physical testing.

Abstract:Directed against complex attack and defense confrontation between carrier aircraft and surface-to-air missiles in modern warfare,this paper proposes an online decision-making method based on neural networks.In synthetic consideration of dual constraints of carrier aircraft maneuverability and target en-gagement,the aircraft's maneuvering strategy and the launch timing are optimized,improving the success rate of combat missions and meeting the needs of real-time decision-making.Firstly,dynamics models with anti-radiation missile,surface-to-air missile,and carrier aircraft are established,and a model of hav-ing attack and defense confrontation scenario,including all these three,is constructed.Through the simu-lation,the impact of different maneuvering strategies and launch timing on combat outcomes is analyzed,and the operation time is defined to measure success or failure of mission.And then,a genetic algorithm is employed to optimize the discrete-continuous hybrid parameter problem offline,obtaining the optimal car-rier aircraft maneuvering strategy and anti-radiation missile launch timing,and taking such as these to con-struct a neural network training dataset and building a neural network model for training and validation.Finally,the effectiveness of neural network-based online decision-making is verified through simulation ex-amples.The results show that this method can significantly expand the dominance zone of anti-radiation missiles,increase the mission success rate,and provide rapid predictions,meeting the needs of real-time decision-making.

Abstract:In existing resource allocation schemes for aeronautical Ad hoc networks,there remain some problems that high control overhead is high and time slot is low in utilization in trans-oceanic scenario ap-plications,an improved learning automata is proposed for slot allocation(ILASA)scheme based on the clustered aeronautical Ad hoc network.Firstly,a model of clustered aeronautical Ad hoc network is given.Secondly,a time slot frame structure is designed,the time slot allocation mode of the learning automata al-gorithm is improved,and the probability updating method in the reward and punishment mechanism is op-timized,solving the probability selection bias problem of the learning automata algorithm by increasing the time slot reservation mechanism.Lastly,the network model is constructed on the basis of the OMNeT++platform for simulation.The results show that the proposed scheme can reduce the resource overhead caused by the control information,effectively reduce the average end-to-end delay of the network,and im-prove the network throughput and packet delivery rate.

Abstract:In view of the problems that successful efficiency of de-interleaving is low under conditions of signal-to-noise ratio being low and relative delay being low for Automatic Dependent Surveillance-Broad-cast signals,this paper proposes a method of ADS-B signal de-interweaving based on VMD-SSA-ICA.First-ly,the interleaved signals are decomposed by the Variable Mode Decomposition(VMD).Secondly,the sin-gular spectrum analysis(SSA)is adopted to reconstruct the modes,eliminate mode aliasing,effectively an-alyzing the potential structure of the ADS-B signal.Thirdly,the Independent Component Analysis algo-rithm is used to perform de-interleaving.Lastly,the Dn-CNN neural network is used to denoise the output signal,achieving the integration of signal separation and denoising.The experimental results show that this method achieves signal decoding in a range of success rates from 60.92%to 99.94%respectively under condition of 8 to 15 dB signal-to-noise ratios.The experiments on the relative delay of different signals show that the algorithm maintains stable de-interleaving performance even with a relative delay of 0 to 10 ms.The method significantly improves the robustness and anti-interference ability of ADS-B signal de-interleaving algorithm.

Abstract:Aimed at the problems that detection efficiency is low,and accuracy of conflict detection is not high among a large number of drones in high-density local battlefield airspace in the future,a real-time and rapid detection of conflicts is obtained among a large number of drones in airspace by constructing a drone protection zone model and a battlefield airspace grid model to extract the airspace position matrix for each drone,and by using the Hadamard product calculation method and the properties of composite and prime factor decomposition for drone conflict detection to identify the drone numbers,positions,altitudes,and other information of conflicts.And on the basis of this,operational layers,avoidance layers and collision thresholds are introduced to establish a quantifiable airspace-capacity evaluation framework.The simula-tion results indicate that compared to the traditional methods,this Hadamard product calculation method enables the complexity of conflict detection to reduce from O(Cn-2n)to level O(n-1),and the conflict de-tection time for 800 drones is controlled within 40 ms,greatly improving the efficiency of conflict detec-tion.Through further processing of the conflict set under an acceptable collision probability criterion,reli-able airspace-capacity values can be worked out,offering theoretical and technical support for the safe and efficient operation of large-scale,high-density drones in local airspace.

Abstract:Aimed at the problems that computational power is limit,storage space is small,and real-time is high for requirements of drones,a method of detecting interpretable drone network intrusion based on Kolmogorov-Arnold Networks(KAN),called KIDS,is proposed.Under the inspiration of Kolmogorov-Arnold Representation Theorem,KAN is to utilize spline-parameterized univariate functions for replacing the traditional linear weights to dynamically learn activation patterns,enabling effective handling of feature extraction,and achieving excellent drone network intrusion detection performance with a more lightweight network structure.Furthermore,the visualization of parameterized spline functions provides insights into the model's decision-making process during traffic feature extraction,enhancing trust in the model's ap-plication.The extensive experiments being over on the real-world drone network traffic dataset drone-IDS-2020,the results demonstrate that the KIDS achieves superior detection performance by still lower model complexity,and exhibits obvious generalization capability in intrusion detection for surpassing type of drones.

Abstract:In order to optimize the structural parameters of the single-stage direct fed coil launcher and the energy distribution of the three-stage direct fed coil launcher,a field path coupling mathematical model and finite element analysis model of the three-stage direct fed coil launcher are established.The accuracy of the two models is calculated by adopting numerical calculation methods and finite element analysis methods.The results show that the motion data and circuit data calculated by the two models are basically consist-ent,proving that the two models have high accuracy.Based on the established parameterized model of the 3-level direct fed coil launcher,the differential evolution algorithm is used to optimize the structural pa-rameters of the single-stage direct fed coil launcher and the energy allocation of the 3-level direct fed coil launcher.The results show that the iterative process is good at convergence with convergence being 60 it-erations for single-stage optimization and 100 iterations for 3-level optimization.The optimization being o-ver,the single-stage outlet speed increases from 27.5 m/s to 30.8 m/s with an output efficiency being improvement of 6%.The peak single-stage current decreases from 19 kA to 8 kA with a decrease of 57.9%.The 3-stage outlet speed increases from 56.4 m/s to 60.4 m/s with an output efficiency being im-provement of 5%.The peak 3-stage current decreases from 25 kA to 10kA with a decrease of 60%.The result show that the differential evolution algorithm has a good ability to optimize for multi-dimensional optimization problems of direct fed coil launchers.

Abstract:In view of the problem that deficiencies exist in current air target threat assessment methods,an air target threat assessment model based on regret theory and multi-criteria is proposed.Firstly,an as-sessment index system is constructed by individual threat,individual value and threat urgency,and the en-tropy weight method is improved by introducing Pearson's correlation index to get the objective weights of the indexes.And then,a threat assessment model based on the three-way decision methods improved by the regret theory is established.Finally,a multi-player game model is constructed based on the different assessment criteria,and the Nash equilibrium is solved to get the target's threat categorization and ordering results.The simulation results show that the proposed method enables commanders to obtain the custom-ized classification and the threat ranking results according to their decision-making preference,and the method is valid.

Abstract:The highly-squinted airborne synthetic aperture radar(SAR)with curved trajectory has signifi-cant potential in applications characterized by disaster monitoring and resource exploration due to its flexi-ble flight and wide-area coverage.However,the significant spatially variant phase errors are introduced by large squint angle,leading to severe defocusing or even imaging failure at the scene edges when the tradi-tional imaging algorithms are employed.For this reason,a method of residual spatially variant phase error in compensation for airborne curved trajectory SAR based on the radius/angle algorithm(RAA)is pro-posed.First,an accurate model of residual spatially variant phase error is established,and its spatial distri-bution characteristics are deeply analyzed,and then,the analytical expression of the residual phase error in the spatial frequency domain is derived,and a simple and efficient spatially variant phase error compensa-tion filter is designed by establishing a mapping relationship between the slow-time domain and the spatial frequency domain.The filter operates in the spatial frequency domain by using a block-wise processing strategy,effectively eliminating the spatially variant phase errors caused by the curved trajectory and large squint angle.Compared with the traditional algorithms,this proposed method significantly improves ima-ging quality at the scene edges while maintaining low computational complexity.The simulation and exper-imental results show that the phenomenon of azimuth defocusing at edge point targets is significantly sup-pressed,and the image resolution,the peak sidelobe ratio,and the integrated sidelobe ratio are still more close to the theoretical values,enabling the wide-swath imaging under large squint conditions.

Abstract:In response to the problems that segmentation is low in accuracy and detection is slow at speed in detecting drone aerial images in key areas by using existing semantic segmentation algorithm,an im-proved PP-LiteSeg lightweight drone image semantic segmentation algorithm is proposed.The algorithm,first,is to design a composite attention fusion module in which the parameter free attention mechanism Si-mAM is introduced into the unified attention fusion module to enhance global contextual information and improve the information richness of output features.Afterwards,the model parameters are reduced through replacing the calculation method of convolution in the backbone network from ordinary convolu-tion to a combination of partial convolution and small-scale convolution kernels.At the same time,a new backbone network SDTCM_PNet is designed to further enhance the lightweighting of the model by chan-ging the feature concatenation method of short-term dense connection modules in the multi-layer receptive field of the backbone network.The experimental results conducted on the self-collected drone aerial image dataset show that the algorithm proposed in this paper is valid.Simultaneously,the algorithm is to be de-ployed and tested on embedded devices,and the algorithm also meets the needs of real-time.

Abstract:In response to the difficulties in analyzing and quantifying the collaborative interception of multi-layer antimissile systems,a'three domain four network'collaborative interception model is constructed based on the theory of hypernetworks.A modeling method for interception networks is proposed,with kill chains as nodes and collaborative relationships as edges.The Kuramoto collaborative dynamics model is used to analyze the supporting role of horizontal and vertical collaboration on the system.Based on the ex-perimental data of the simulation platform,the analysis shows that the structural indicators of the collabo-rative interception network are significantly correlated with the damage rate in the correlation analysis,confirming that the network structural characteristics can effectively characterize the combat effectiveness;Self synchronization analysis shows that a scheme that combines horizontal and vertical collaboration can fully leverage the resource density superposition effect of horizontal collaboration and the task relay advan-tage of vertical collaboration,achieving a global order parameter of 0.93.However,schemes that only use a single collaboration mode can not achieve optimal collaboration effects.The multi-layered antimissile system has a strong correlation with combat effectiveness through the dual dimensional coordination mech-anism of'spatial resource density superposition+time task relay'.The organic combination of horizontal and vertical coordination is the key to improving the overall coordination effectiveness of the system,pro-viding a theoretical basis for the collaborative optimization of the anti missile system.

Abstract:To enhance the level of air traffic services within approach control areas and ensure aircraft flight safety,this paper proposes a 4D trajectory prediction method based on a Spatial-Temporal Aware Trans-former model.The method is to construct a 4D trajectory time-series prediction model,realizing high-pre-cision multi-step trajectory forecasting through transforming aircraft trajectory prediction into a time-series forecasting problem,extracting spatial-temporal features from historical flight trajectories as model inputs in comprehensive consideration of departure airport,aircraft type,wake vortex category,landing runway,and aircraft attitude,and adopting Gated Recurrent Units(GRUs)to capture temporal dependencies with-in trajectory data,while a combined Temporal Convolutional Network-Gated Recurrent Unit(TCN-GRU)architecture extracts spatial features.The experimental results demonstrate that this model is prior to the traditional Transformer model.With the increase of prediction steps,the root-mean-square error of predic-tion accuracy is reduced to 20.1％,and the mean absolute error is reduced to 27.82％ by this model.These findings hold significant implications for improving the efficiency and safety of air traffic manage-ment.

Abstract:Interference forwarded by intermittent sampling based on digital RF memory is a new type of coherent interference that can achieve both suppression and deception effects,constituting a serious threat to radar target detection.Aimed at the problem that the interference is of forwarding by intermittent sam-pling in which radar transmission signal is linear frequency modulation signal,a suppression technique based on YOLOv8 algorithm is proposed.Through STFT used for time-frequency analysis,and the im-proved YOLOv8 algorithm used to identify interference features in the enhanced time-frequency distribu-tion grayscale map,a filter is designed in the time-frequency domain to suppress the interference signal.The simulation comparison experiment results show that this method is good at interference suppression effect,which is low in signal-to-noise ratio,weak in interference,and is good at coordination with multi-point source.Compared with the other methods,this method solves the problem of overlapping interfer-ence and target time-frequency distribution that cannot be recognized,and is robust.

Abstract:With the continuous enhancement of modern detection technologies,the joint detection tech-niques of visible light,infrared,and radar have been widely applied.How to achieve excellent stealth per-formance simultaneously in multiple frequency bands has become a key challenge.For the above-men-tioned reasons,this paper proposes and verifies a multifunctional flexible meta-surface in integrating spa-tially tunable infrared emissivity and radar cross-section(RCS)reduction.This meta-surface has both functions of infrared camouflage and radar stealth.Its sample structure is composed of aerogel felt/ITO/Air/PET/ITO from bottom to top successively.The functional layer is designed by etching ITO patch units on PET to form an encoded meta-surface.The surface achieves RCS reduction through microwave diffuse reflection and spatially tunable infrared emissivity by adjusting the ITO duty cycle.The simulation and experimental results show that the sample can effectively suppress the backscattering of electromag-netic waves in the 6～12 GHz band,achieving radar stealth,and the tunable range of infrared emissivity is approximately 0.25～0.45,enabling infrared camouflage.Additionally,the sample has good thermal in-sulation and flexibility,further expanding its application range.Moreover,the integrated design helps re-duce the structural thickness and processing complexity in engineering.In summary,this meta-surface de-sign holds promising application prospects in the fields of multi-spectrum compatible stealth and infrared camouflage.

Abstract:With the increasement of making the needs of flexible electronics in wearable,ion-gel materials have attracted considerable attention due to their flexibility and conductivity.This study systematically in-vestigates the regulatory mechanism of bending stress on the performance of ion-gel-gated metal-oxide-semiconductor field-effect transistors(MOSFETs)by simulating tensile and compressive strains through concave and convex molds in combination with electrical characterization and COMSOL Multiphysics simula-tions.The experimental results demonstrate that when the bending radius reaches 28.5 mm,the device's on/off ratio decreases from 2 658 to 2 215,the saturation carrier mobility drops by 53％(from 16.1 cm2/(V·s)to 8.6 cm2/(V·s)),and the hysteresis window expands by 9.3％.The simulation results further reveal that when the Mises stress exceeds 100 MPa,bending stress not only degrades the on/off ratio,carrier mobility,and hysteresis window but also induces a positive threshold voltage shift via interface trap states and energy band modulation.Notably,tensile strain exhibits a more pronounced impact on device per-formance compared to the compressive strain.These findings provide critical technical insights for the op-timization of ion-gel-based flexible devices in wearable electronics and smart sensors.

Abstract:In the light of the limited problems in model scalability and precision that scale is large in varia-tion,dense is irregular in distribution,and class is imbalance in vehicle object detection on drones,this study proposes an improved YOLO-QYF vehicle detection method based on the YOLOv7.This method is to introduce the QARepVGG module to replace the computationally intensive E-ELAN in the baseline,and minimize information loss during feature map transmission,a content aware reassembly of features is em-ployed,and a coordinate attention mechanism is integrated to enhance the localization ability for objects of interest.Additionally,the WIoU loss function is introduced to address the class imbalance issue and im-prove the model's generalization capability.The experimental results demonstrate that the proposed method achieves the mean average precisions of 96.4％,95.6％and 94.5％under the free flow,the syn-chronous flow and the blocking flow traffic scenarios respectively,increasing by 2.0％,1.6％ and 3.4％compared to the baseline,and the frames per second are 81.62,78.13 and 76.34 respectively.In the vis-drone2021 dataset,the mean average precisions of the proposed method achieves 60.6％,and is 1.4％higher than the baseline.

Abstract:Aimed at the problems that object is dense in distribution,small object is easy to omit,and scale is variable in drone aerial imagery,an improved algorithm TB-YOLOv8 is proposed.This method is to incorporate a triplet attention-based C2f_T module into the backbone to enhance shallow feature extrac-tion and key region awareness,reducing false positives and missed detections.A lightweight BiFPN is in-tegrated into the neck to improve multi-scale feature fusion efficiency.And simultaneously,the WIoUv3 loss function is introduced to enhance model robustness and localization accuracy.The experimental results on the VisDrone2019 dataset show that the TB-YOLOv8 achieves a 3.0％ improvement in mAP,a 6.31％reduction in parameter count,and a detection speed of 164.1 FPS compared to the original YOLOv8s mod-el.Moreover,the TB-YOLOv8 is prior to the Faster R-CNN,the RT-DETR,the YOLOv7,and the YOLOv11 in terms of detection performance,has high accuracy,efficiency,and real-time capability,and there is considerable potential in aerial target detection application.