针对传统的航迹关联算法在运动目标交叉、分岔时,常出现错漏相关航迹且计算量随着传感器和目标数量增加而飞速增长的缺陷,提出一种改进的Kohonen神经网络航迹关联算法.该算法由聚类关联、目标状态估计、神经元优化和状态融合估计等模块组成.通过给每个竞争层神经元加上一个合适的阈值,有效避免了常规的Kohonen神经网络因初始权值选择不合适而容易造成坏死神经元的问题.进一步设计了自组织竞争神经网络学习规则,将多传感器在同一时刻的测量数据进行自组织聚类,从而实现测量数据的有效关联.最后,利用连续时间下的关联数据,实现运动目标航迹关联.仿真研究验证了该算法的可行性和有效性.%There are several difficulties must be overcome to develop a track-to-track association algorithm in a distributed multi-sensor situation. For example, it might get errors and omissions related track when there are cross or bifurcation in the target moving patch. Besides, the computation might grow rapidly with the increasing of the number of sensors and target. In order to resolve these problems, a modified Kohonen neural network based correlation algorithm was presented. The algorithm was made up of four main modules, including clustering association, target state estimation, neurons optimization, and state fusion estimation. The suitable thresholds for each layer of the competition element in the neural network were set to avoid the phenomenon of necrotic neurons occurred in the conventional method due to inappropriate choice of initial weights. The learning rules were also designed for the clustering of multi-sensor measurement data in a self-organizing manner. Besides, data clustering association on continuous-time data were used to achieve moving target track correlation. The simulation results demonstrate the efficiency and effectiveness of the proposed algorithm.
展开▼
