基于多核融合的目标大面积遮挡处理方法

提出了一种基于多核融合的目标遮挡处理方法,用于提高大面积遮挡情况下视觉目标跟踪算法的鲁棒性和准确性.与现有基于单个对称核加权直方图的mean shift跟踪算法不同,该方法以目标区域内的多个非中心位置为核函数中心,构建多个非对称核加权直方图.由于这些直方图对目标的不同区域赋予了不同的权重,使得在遮挡发生时总存在一些直方图受影响较小.依据各个直方图分别进行mean shift迭代获得一组目标位置估计后,利用D-S证据理论融合判定最终的目标位置.实验结果表明,该方法在目标被大面积遮挡时仍能够获得准确的跟踪.     

风场中基于神经网络的炸弹弹道落点拟合分析

介绍了两种炸弹弹道落点拟合神经网络模型,对各模型的优缺点以及处理结果进行了分析。提出了采用广义回归神经网络来处理炸弹弹道落点拟合问题,弹道的落点参数、初始投放条件与风场可通过神经网络的阈值和权值来表现。仿真结果表明,应用广义回归神经网络进行弹道落点拟合,具有算法可行性好、拟合精度高、速度快等优点,而且运算简单;该方法在实战中有很高的参考价值和工程实用价值。     

Hybrid optimization for multiple-impulse reconfiguration trajectories of satellite formation flying

A hybrid optimization method is developed for fuel-optimal reconfigurations of a group of satellites flying in formation. The genetic algorithm performs a global search to find two-impulse trajectories, and primer vector analysis finds multiple-impulsive local optimal trajectories with the two-impulse trajectories as initial guesses. Hybrid optimization finds globally optimal trajectories for formation reconfigurations, including formation resizing, reassignment and reorientation maneuvers. Multiple-impulse trajectories reduce the fuel consumption from the two-impulse trajectories by up to 4.4% for those maneuvers. In real missions, satellites can follow two-impulse trajectories to gain the advantage of a smaller number of impulses, with the cost of slightly more propellant. The qualitative characteristics of the optimal trajectories are analyzed from the number of optimal trajectories found by hybrid optimization.     

3-D position estimation of isolated target and optimal trajectory choice in parasitic synthetic aperture radar

Bistatic SAR (BSAR) systems have recently been the subject of several studies, but little attention has been given to the potential of the location of targets. In the paper, the performance of the estimation of 3-D position of a target (TPE) in parasitic SAR is obtained analytically and illustrated by computer simulation using ambiguity function analysis, and the maximum likelihood estimate (MLE) approach. It was shown that by using a multi look parasitic BSAR all three coordinates of an isolated point target (IPT) could be evaluated. Analytical closed form equations that characterize the measurement accuracy were derived. In the final these equations will be used to demonstrate various cases of more practical, including optimal trajectories choice under any geometry configuration case, moreover, the conclusion is verified via Matlab.     

SEQUENTIAL ALGORITHM FOR MULTISENSOR PROBABILISTIC DATA ASSOCIATION

Jointprobabillsticdataassociation(JPDA)isanalgorithmusedinsinglesensormultipletargettrackingsystems.Itemploysthenon-uniqueassignmentof"allneighbor"strategytoadaptforthedensemultitargettrackingenvironments[1].Becauseofitswideapplications,itisnecessarytoextendJPDAintosomemultiplesensortrackingsystems.Suchamultisensorsystem,forexample,canbeformedbycollocatingradarandinfraredsearchandtrack(IRST)whichcantakeadvantagesofboththesensorsbodatafusion.Undertheconditionofthesamesensors,acommonmeasure…     

临近空间高超声速滑跃式轨迹目标跟踪技术

针对临近空间目标飞行速度快、机动特性强和加速度突变的特性,提出一种地心直角(ECEF)坐标系下基于目标特性分析的修正强跟踪滤波(MSTF)算法。首先,通过对ECEF坐标系下目标量测的无偏转化处理,以有效减小目标高超声速飞行所带来的旋转、平移和线性化误差影响;接着,在对目标特性充分分析的基础上,合理构建强跟踪滤波(STF)模型,通过对模型参数的自适应调节,以有效实现临近空间高超声速滑跃式轨迹目标的精确跟踪;最后,结合统计学原理对目标加速度的突变进行合理检测和补偿,以进一步修正强跟踪滤波模型的跟踪精度。仿真结果表明,与现有的临近空间目标跟踪算法相比,该算法具有较高的定位跟踪精度。     

基于Matlab Simulink的弹道仿真方法

弹道仿真软件有助于飞航导弹武器飞行试验结果的分析与评定。为此,设计了基于Matlab Simulink的弹道仿真计算方法,建立了六自由度空间弹道数学模型,利用导弹相关的气动数据和控制参数测试了算法,实现了软件,仿真结果证明了该模型的准确性和可信度。     

基于动态贝叶斯的反舰导弹弹型识别

使用反舰导弹对舰艇进行饱和攻击是当前对舰攻击的主要方式。首先,提出了动态贝叶斯网络(DBN)的理论及算法;其次,根据反舰导弹的高度、速度、过载、机动方式等弹道特性和RCS、雷达频率等导弹自身特性,建立了对反舰导弹进行弹型识别的DBN模型;最后,通过仿真算例验证了DBN模型的有效性,这为实际解决反舰导弹进行弹型识别的问题,提供了一种思路。     

无人飞行器地面移动目标跟踪系统研究现状与展望

无人飞行器地面移动目标跟踪系统是一项在军事和民用方面都具有广泛应用前景的非常重要的研究课题。首先分析了系统的特点，然后详细介绍了国内外无人飞行器地面移动目标跟踪系统的研究方法和现状，并指出其不足。最后，针对现有系统中控制方法的欠缺，提出了将仿人眼控制系统用于无人飞行器上的摄像机云台控制的新方法。     

基于卫星观测数据的弹道导弹初始状态估计

利用卫星探测数据对大气层外飞行的弹道导弹弹道进行估计。在极大似然估计中,由于卫星探测导弹运动的可观测性较差,使初始点估计变得十分病态,传统的Gauss Newton迭代算法不再适用,而采用较为复杂Levenberg Marquardt算法,分别对单颗卫星和两颗卫星的情况进行了Monte Carlo仿真,得到了各自的归一化残差测量误差(代价函数),结果表明该算法有效。