KGTLIR:An Air Target Intention Recognition Model Based on Knowledge Graph and Deep Learning

As a core part of battlefield situational awareness,air target intention recognition plays an important role in modern air operations.Aiming at the problems of insufficient feature extraction and misclassification in intention recognition,this paper designs an air target intention recognition method(KGTLIR)based on Knowledge Graph and Deep Learning.Firstly,the intention recognition model based on Deep Learning is constructed to mine the temporal relationship of intention features using dilated causal convolution and the spatial relationship of intention features using a graph attention mechanism.Meanwhile,the accuracy,recall,and F1-score after iteration are introduced to dynamically adjust the sample weights to reduce the probability of misclassification.After that,an intention recognition model based on Knowledge Graph is constructed to predict the probability of the occurrence of different intentions of the target.Finally,the results of the two models are fused by evidence theory to obtain the target’s operational intention.Experiments show that the intention recognition accuracy of the KGTLIRmodel can reach 98.48%,which is not only better than most of the air target intention recognition methods,but also demonstrates better interpretability and trustworthiness.

Air target recognition method against ISRJ for radio frequency proximity sensors using chaotic stream encryption

The interrupted-sampling repeater jamming(ISRJ)can cause false targets to the radio-frequency proximity sensors(RFPSs),resulting in a serious decline in the target detection capability of the RFPS.This article proposes a recognition method for RFPSs to identify the false targets caused by ISRJ.The proposed method is realized by assigning a unique identity(ID)to each RFPS,and each ID is a periodically and chaotically encrypted in every pulse period.The processing technique of the received signal is divided into ranging and ID decryption.In the ranging part,a high-resolution range profile(HRRP)can be obtained by performing pulse compression with the binary chaotic sequences.To suppress the noise,the singular value decomposition(SVD)is applied in the preprocessing.Regarding ID decryption,targets and ISRJ can be recognized through the encryption and decryption processes,which are controlled by random keys.An adaptability analysis conducted in terms of the peak-to-side lobe ratio(PSLR)and bit error rate(BER)indicates that the proposed method performs well within a 70-k Hz Doppler shift.A simulation and experimental results show that the proposed method achieves extremely stable target and ISRJ recognition accuracies at different signal-to-noise ratios(SNRs)and jamming-to-signal ratios(JSRs).

Full Digital Smart Fuze on Air Target

The intelligent fuzing is a kind of perfect way to optimize detonation location.The fuze can autonomously configure the warhead detonation mode to optimize the desired effects against the target.For the air target,the fuze can sense the impact in the impact mode then detonate the warhead on impact,otherwise,it will operate in the proximity mode and detonate at a closest approach to the target.The anti-jamming ability is also added on fuze＇s full digital signal processing platform.The method for burst point＇s controlling is analyzed and the digital intelligent fuze system based on system on programmable chip（SOPC） is designed.

Air Target Fuzzy Pattern Recognition Threat-Judgment Model

Threat-judgment is a complicated fuzzy inference problem. Up to now no relevant unified theory and measur-ing standard have been developed. It is very difficult to establish a threat-judgment model with high reliability in the airdefense system for the naval warships. Air target threat level judgment is an important component in naval warship com-bat command decision-making systems. According to the threat level judgment of air targets during the air defense of sin-gle naval warship, a fuzzy pattern recognition model for judging the threat from air targets is established. Then an algo-rithm for identifying the parameters in the model is presented. The model has an adaptive feature and can dynamicallyupdate its parameters according to the state change of the attacking targets and the environment. The method presentedhere can be used for the air defense system threat judgment in the naval warships.

Hybrid hierarchical trajectory planning for a fixed-wing UCAV performing air-to-surface multi-target attack

This paper considers the problem of generating a flight trajectory for a single fixed-wing unmanned combat aerial vehicle （UCAV） performing an air-to-surface multi-target attack （A/SMTA） mission using satellite-guided bombs. First, this problem is formulated as a variant of the traveling salesman problem （TSP）, called the dynamic-constrained TSP with neighborhoods （DCT- SPN）. Then, a hierarchical hybrid approach, which partitions the planning algorithm into a roadmap planning layer and an optimal control layer, is proposed to solve the DCTSPN. In the roadmap planning layer, a novel algorithm based on an updatable proba- bilistic roadmap （PRM） is presented, which operates by randomly sampling a finite set of vehicle states from continuous state space in order to reduce the complicated trajectory planning problem to planning on a finite directed graph. In the optimal control layer, a collision-free state-to-state trajectory planner based on the Gauss pseudospectral method is developed, which can generate both dynamically feasible and optimal flight trajectories. The entire process of solving a DCTSPN consists of two phases. First, in the offline preprocessing phase, the algorithm constructs a PRM, and then converts the original problem into a standard asymmet- ric TSP （ATSP）. Second, in the online querying phase, the costs of directed edges in PRM are updated first, and a fast heuristic searching algorithm is then used to solve the ATSP. Numerical experiments indicate that the algorithm proposed in this paper can generate both feasible and near-optimal solutions quickly for online purposes.

Three-Dimensional Discrete Observability Analysis for Air-to-Air Missile Target Tracking

The interception information of infrared（ IR）-guided air-to-air missiles（ AAM） is mainly estimated only using the basic bearing measurements. In order to intercept highly maneuverable targets,it is essential to study the system observability to improve the target tracking system performance.The uniqueness of this paper is that the observability analysis is derived based on a discrete three-dimensional （3D） system model. During the maneuvering scenario,the system is approximated by a segment-by-segment system. The relationship between missile-target motion and observability is given by direct and dual approaches. Meanwhile sufficient observability conditions are derived. Moreover,a numerical simulation is conducted and an alternate method is provided to reinforce the proposed observability analysis results.

基于双线性插值的单目标检测算法

针对已有算法对于空中小目标检测效果差、精度低、漏检等问题,提出一种基于双线性插值的单目标检测算法。对主干网络进行设计,再根据特征图的特点,使用不同的方法进行特征融合。在特征融合的过程中使用双线性插值算法进行上采样。对算法的先验框进行设计,使算法的先验框尺寸与数据集目标框尺寸更加契合。实验结果表明,提出算法的平均检测精度对比SSD提升了10.29%,算法计算量对比Faster-RCNN减少了82.18%。

空战目标轨迹预测技术研究综述

信息化空战对抗中,快速获取和有效利用对手信息的一方,能够准确地预测对手的运动轨迹,更加迅速地完成OODA循环,进而取得空战优势。本文研究空战目标轨迹预测技术,以时序变化特征为研究视角,分别对基于物理、机器学习和深度学习的轨迹预测技术开展分析和总结,进一步归纳物理因素、空战因素和交互因素等作为轨迹预测模型的输入,单模态轨迹、多模态轨迹和行为意图三类作为模型的输出,并对未来智能化空战中的目标轨迹预测技术进行展望。

基于机器视觉的小白杏热风干燥控制系统设计

设计了一种基于机器视觉的小白杏热风干燥控制系统,该系统使用计算机视觉技术,对小白杏进行检测和分类,并自动控制热风干燥过程中的温度、风速。系统采用基于神经网络的目标检测算法,对小白杏进行检测和分割,利用图像处理技术提取小白杏的特征,并使用YOLOv7对小白杏进行分类。通过调整热风干燥的温度、风速和时间等参数,实现了对小白杏干燥过程中的品质保护,可保持小白杏品质和口感,提升杏干经济价值,该系统具有一定的应用前景。

基于深度强化学习的舰船导弹目标分配方法

针对对抗环境下的海上舰船防空反导导弹目标分配问题,本文提出了一种融合注意力机制的深度强化学习算法.首先,构建了舰船多类型导弹目标分配模型,并结合目标多波次拦截特点将问题建模为马尔可夫决策过程.接着,基于编码器–解码器框架搭建强化学习策略网络,融合多头注意力机制对目标进行编码,并在解码中结合整体目标和单个目标编码信息实现舰船可靠的导弹目标分配.最后,对导弹目标分配收益、分配时效以及策略网络训练过程进行了仿真实验.实验结果表明,本文方法能生成高收益的导弹目标分配方案,相较于对比算法的大规模决策计算速度提高10%~94%,同时其策略网络能够快速稳定收敛.

花椒热风-微波组合干燥失水特性研究

花椒热风干燥降速期水分含量低,水分扩散慢,导致热风干燥耗时长。为提高干燥效率,并通过热风与微波组合干燥,分别进行热风干燥、微波干燥和热风-微波组合干燥实验,探究不同干燥参数对花椒失水特性的影响,以确定合理的干燥转换临界点和最优组合干燥模型,并将傅里叶准则数(F_(0))引入Fick第二扩散定律方程,求解有效水分扩散系数(D_(eff))。研究结果表明:热风和微波单独干燥时,升高风温风速和增加微波功率均有利于缩短干燥时间;热风-微波组合干燥花椒时,热风段转微波段的最佳目标含水率即为热风干燥的临界点含水率(65%(w.b)),且高热风温度和高微波功率均可使微波干燥段获得高失水速率;热风-微波组合干燥花椒热风段和微波段对应的最优模型分别为Wang and Singh模型和Page模型,D_(eff)范围分别为1.908×10^(-9)~3.547×10^(-9)m^(2)/s和1.883×10^(-8)~3.321×10^(-8)m^(2)/s。热风-微波组合干燥方式能够显著提高干燥效率,促进花椒内部水分扩散,干燥模型可为优化干燥工艺和设计干燥设备提供理论依据。

一种空中目标关联关系挖掘分析方法

快速准确挖掘分析雷达历史数据中蕴含的目标关联关系,可以为空中目标活动规律研究、目标识别判性等提供辅助,充分发挥历史数据作用。根据目标时空特性,设计了一种新的空中目标关联关系挖掘分析方法。设计了一种时间匹配方法,生成时间关联数据集。设计了一种空间关联度量,生成空域关联数据集。运用DTW及生成树算法,匹配并搜索具有关联关系的空中目标,生成关联关系集合。经数字仿真验证,该方法可有效挖掘雷达数据历史信息,寻找目标之间的关联规律。

面向空中目标意图识别的时空Transformer模型设计

信息化条件下的战争环境瞬息万变,战场态势数据呈现海量、多样化的特点,导致利用专家经验识别空中目标作战意图的难度越来越高。结合目前先进的智能化方法,对Transformer模型进行研究并将其引入空中目标意图识别领域,设计出一种新的意图识别方法Temporal-Spatial Transformer(TST),可以有效地挖掘战场态势数据中时间域和空间域的深层特征信息,提高空中目标作战意图识别准确率。同时,对4种目前较为先进的神经网络意图识别方法进行效果对比,结果表明,TST模型在各类指标上都取得了突出的效果,优于所有对比的神经网络模型。TST模型不仅有优异的准确率,而且收敛速度极快,可以迅速地抓取战场态势数据中的关键信息进行意图识别。

空中目标动态电磁散射数据仿真系统设计与实现

电磁散射数据是目标识别研究的基础,但由于试验测量成本高、可重复性差等问题,空中目标电磁散射实测数据十分有限。基于去遮挡的N点模型和电磁计算数据插值研究了空中目标动态电磁散射数据仿真方法,设计了空中目标动态电磁数据仿真系统,该系统将电磁散射仿真与实际飞行场景相结合,支持场景自定义和空中目标三维模型库扩展。分析了飞机和巡航导弹2类目标仿真数据及成像结果,结果验证了仿真方法正确性和系统有效性,可为研究实际场景下的空中目标识别提供支撑。

基于深度残差收缩网络的雷达空中目标识别

雷达空中目标高分辨距离像(high resolution range profile,HRRP)中往往包含一定的杂波噪声,利用HRRP开展空中目标识别需要重点考虑噪声的影响。针对上述问题,提出一种基于深度残差收缩网络(deep residual shrinkage network,DRSN)的雷达空中目标HRRP识别方法。该网络将深度残差网络、软阈值函数和注意力机制结合起来,采用跨层恒等连接方式,不仅可以避免网络层数过深造成梯度消失或梯度爆炸,从而导致网络学习能力下降的问题,还可以有效过滤掉识别过程中噪声特征的影响,使模型专注于目标区域的深度特征识别,提升强噪声背景下模型的识别能力。实验结果表明,相对于其他常用的深度学习模型,所提方法在各个信噪比条件下,识别效果均有一定的优势,该模型对噪声具有较强的鲁棒性。

基于残差密集连接注意力的红外目标识别算法

针对红外目标识别时目标与干扰混淆、干扰遮挡目标导致算法性能下降的问题,提出了一种基于残差密集连接注意力的空中红外目标识别算法。首先,为跨层实现浅层与深层特征的融合、获取融合深度特征、提升特征复用性并增强语义信息,提出了改进的残差密集连接模块;其次,为增强融合深度特征的自适应表达能力,设计了并行混合注意力模块;最后,通过在大量红外数据集上的测试证明,相比于GoogLeNet算法,所提算法平均识别正确率提高了1.9个百分点,证明了算法的有效性。

基于PM_(2.5)浓度达标约束和区域联防联控的河南省地级市大气环境容量研究

为研究以河南省为代表的受大气污染传输影响显著省份的大气环境容量,本文基于CMAQ模型,采用嵌套迭代模拟的方法,计算了在周边省份区域联防联控的前提下河南省PM_(2.5)浓度达标(GB 3095-2012《环境空气质量标准》二级标准限值)时17个地级市SO_(2)、NO_(x)、一次PM_(2.5)和NH_(3)的大气环境容量.结果表明:①省外传输对河南省PM_(2.5)浓度的贡献率为50.29%,其中周边7个省份传输贡献率为36.19%,可见周边省份实施大气污染联防联控是河南省实现空气质量达标的必要条件.②在省级PM_(2.5)浓度达标时,河南省周边7个省份SO_(2)、NO_(x)、一次PM_(2.5)和NH_(3)的大气环境容量分别为279.07×10^(4)、465.61×10^(4)、172.67×10^(4)和182.96×10^(4)t,超载排放量分别为243.13×10^(4)、402.79×10^(4)、143.63×10^(4)和142.54×10^(4)t.③在区域联防联控的前提下,全省PM_(2.5)浓度达标时河南省17个地级市的SO_(2)、NO_(x)、一次PM_(2.5)和NH_(3)环境容量分别为23.48×10^(4)、52.40×10^(4)、22.74×10^(4)和32.17×10^(4)t,超载排放量分别为48.43×10^(4)、106.78×10^(4)、45.84×10^(4)和61.63×10^(4)t.④河南省东南部区域的SO_(2)和NO_(x)排放以及西北部区域的一次PM_(2.5)和NH_(3)排放会对全省PM_(2.5)浓度变化有显著贡献.研究显示,河南省PM_(2.5)浓度达标时,4种主要大气污染物的环境容量的最大值为52.40×10^(4)t,超载排放量的最大值为106.78×10^(4)t.为实现河南省PM_(2.5)浓度达标,河南省周边省份均需要实施大气污染联防联控,同时河南省需要加大对东南部区域的SO_(2)和NO_(x)排放以及西北部区域的一次PM_(2.5)和NH_(3)排放的管控.

基于RL-LSTM的空中目标意图识别方法

空中目标意图识别是战场态势认知的重要部分。为了进一步提高空中目标意图识别准确率及实时性,提出了基于改进长短时记忆(long short-term memory,LSTM)网络模型RL-LSTM的空中目标意图识别方法。首先获取目标实时的状态数据,以最后时刻目标状态作为模型输入,利用RL-LSTM模型来学习7种常见意图的运动及时间相关特征信息,最后,通过Softmax分类器实现目标意图识别。仿真实验表明,该模型提升了现有神经网络模型的识别准确率及识别效率。

基于多特征相似度的空中目标聚类研究

为准确识别空战场敌方目标的编队信息,本文提出一种针对空中目标的聚类模型,综合选取空中目标的位置、高度、航向、航速等多维度信息,计算目标之间的相似度,构建多特征相似度矩阵,将多特征相似度作为目标之间“距离”的度量方式,通过K-means算法聚类迭代,得到空中目标的编队信息。通过仿真实验,验证了本文所提方法在复杂空战场环境下,能准确识别敌目标的编队信息,相比传统聚类算法,识别准确率提高40%,有较好的工程应用价值。

基于K近邻算法的空中目标威胁度判断方法

针对传统的空中目标威胁度评估方法具有计算量大,实时性差,难以适用于数据缺失的情况,提出采用K近邻算法(KNN)对任意来袭目标实现威胁度评估的方法。该方法提取了空中目标的状态信息特征作为输入数据,使用离差最大化方法构建数据集,目标威胁度等级作为输出数据,利用K近邻算法构建了目标威胁度评估模型。仿真实验结果表明,该方法能够实现高准确度、实时化的目标威胁度评估,和TOPSIS方法与离差最大化方法进行对比,证明该方法对空中目标异常特征值具有更高的决策效率,更加适用于现代战场的高复杂性,进一步体现了该方法的优越性和可行性。