摘要
关于密集多回波环境下机动多目标跟踪,因其JPDA问题计算量出现的组合爆炸的现象,从而导致实时跟踪效果极差的问题。引入能解决组合优化问题的连续型Hopfield神经网络方法,来减少其计算量,并结合改进的模拟退火算法来优化网络收敛性和全局最优问题。其中,在已知网络能量函数下,对应的网络参数对网络性能产生了直接影响。因此,在实验的基础上,对神经网络参数进行分析研究,并充分证实了该方法解决JPDA问题的有效性,并给出了相应的积极结论。
About multi-maneuvering target tracking in dense multi-return environments, JPDA problems because of the combination of explosive phenomenon, result in real-time tracking the issue very poor results. Therefore, the introduction ofcombinatorial optimization problems are solved by continuous Hopfield neural network method to reduce its computation and it use the improved simulated annealing algorithm to optimize network convergence and global optimization problem. But among this, based on the neural network energy function, the corresponding neural network parameters have the direct influence to the neural network performance. So in experimental basis, it analyzes to the neural network parameters, further confirms this new method solving the JPDA question validity, and providing the corresponding conclusion.
出处
《计算机工程与设计》
CSCD
北大核心
2009年第2期494-498,502,共6页
Computer Engineering and Design
基金
2006年教育部新世纪优秀人才计划基金项目(NCET-06-0487)
国家自然科学基金项目(60472060、60572034)
江苏省自然科学基金项目(BK2006081)
关键词
多目标跟踪
数据关联
卡尔曼滤波
神经网络
参数
multi-targettracking
data association
Kalmanfiltering
neural network
parameters