摘要
在运用恒虚警(CFAR)检测算法中,一个非常重要的工作是根据给定的恒虚警率确定其标称化因子。当标称化因子关于虚警率的关系式很难甚至于是不可能得到时,传统上采用仿真方法,但仿真方法的计算量非常大。为此文章利用BP神经网络具有强大的逼近任意非线性关系式的能力,提出了一种基于BP神经网络的CFAR检测器标称化因子确定方法。通过实例研究表明,通过对BP神经网络的输入进行自然对数的变换后,其对虚警概率和标称化因子的关系进行逼近时需要的训练次数将大为减少,研究还表明基于BP神经网络的标称化因子确定方法具有相当高的精度。
As applying constant false alarm rate（CFAR） detection algorithms, an important task is to determine its scale factor according to given false alarm probability. When analytic expression of scale factor vs false alarm probability is difficult or impossible to be obtained, simulation is adopted traditionally. But the computation of simulation is very large. A method for determining scale factor of CFAR detector based on BP neural networks is proposed in the paper using powerful ability to approximate any non-linear expression. Studies of examples indicate training times of BP neural networks approximating relation between false alarm probability and scale factor can be largely shorten, after nonlinear transformation of natural logarithm is applied to input of BP neural networks. Studies also indicate method for determining scale factor based on BP neural networks can provide high accuracy.
出处
《舰船电子工程》
2009年第10期125-127,共3页
Ship Electronic Engineering
