基于径向基神经网络回归预测的船舶轴频电场实时检测方法

Real-time Detection of Shaft-rate Electric Field of Ships Based on RBF Neural Network Regressive Prediction

针对交变电磁信号在海洋环境中衰减较快的特点,在分析船舶轴频电场以及海洋环境电场噪声与测量电路自噪声的基础上,提出了一种能够在低信噪比下检测船舶轴频电场的方法.对信号进行小波分解并在轴频段进行重构,以噪声数据作为训练集训练RBF神经网络建立预测模型,以预测的误差作为检验统计量,达到提高信噪比的目的.再对该统计量做滑动功率谱检测,与得到的浮动阈值做比较来判断目标是否存在.将该方法用于实测船舶轴频电场的检测,取得了较好的效果,与其它目标检测算法比较,该方法在计算时间和检测能力方面具有明显优势.

Electric field signal attenuats quickly in marine environment and is hard to be detected. Through analysing shaft-rate （SR）electric field and noise of environmental electric field and measurement circuit,an effect way for detecting the SR electric field signal under a low signal- noise ratio（SNR）. Based on the noise, a RBF neural network predictive model was set up, the prediction error was made to be testing statistic and it was detected using slide detection method, the detecting result was compared with the floating threshold so we could judge whether there was a target. The simulation results showed that this method could detect the target effectively. Compared with other detecting method, it costed less time and was better at detection.