基于神经网络分类的异类传感器目标关联算法

摘要

In the data fusion system composed of radar and infrared sensor installed in high speed of dynamic platform, the system error estimation and target correlation are dependent and are difficult very much. To solve the problem, a new target correlation algorithm based on pattern classification is proposed in the article according to the property of system errors variation. The approach realizes pattern classification by BP neural network. It needn’t estimate the system error and compensate it, and has a tolerance to system error. The experiment shows that the average correct probability for target-correlation in the data fusion between the above two kind of sensors is more than 86%.%　　为解决雷达与高速动态平台上的红外传感器构成的融合系统中，系统误差估计、目标关联紧密耦合难于解决的问题，根据系统误差数值变化特性，该文提出了一种基于模式分类的系统误差补偿与目标关联联合处理的方法。该方法采用BP神经网络进行分类，省略了系统误差补偿环节，简化了融合处理流程，对一定范围内变化的系统误差有较好的容限。实验结果表明，该方法用于文中涉及的两类传感器组成的数据融合系统中，能实现目标正确关联的平均概率大于86%。