基于深度学习的复杂观测条件下工频磁异常探测

Power Frequency Magnetic Disturbance Signal Prediction Based on CNN-LSTM Hybrid Model

复杂观测条件下使用工频磁场探测人员、车辆、飞行目标等多类型目标造成的磁场扰动时,受到复杂环境下电磁噪声、供电设备及外来物体扰动等影响,工频磁场扰动信号具有噪声多、干扰强等特征,为有效削弱噪声及干扰对工频磁场扰动信号的影响,实现工频磁场扰动探测,该文利用实验数据对复杂观测条件下的磁场扰动信号进行特征分析,提出了一种基于深度学习的工频磁场异常探测方法,通过提取正常状态与有扰动状态的信号序列,将该信号输入神经网络训练,得到准确检测工频磁场异常信号的网络模型。实验结果表明,该方法的识别准确率在80%以上。

When the magnetic field disturbance caused by the use of power frequency magnetic field detectors,vehicles,flight targets and other types of targets under complex observation conditions is affected by the electromagnetic noise,power supply equipment and external object disturbance in complex environment. The power frequency magnetic field disturbance signal has the characteristics of high noise and strong interference,in order to effectively reduce the influence of noise and interference on power frequency magnetic field disturbance signal. To detect disturbance of power frequency magnetic field,this paper uses experimental data to analyze the characteristics of magnetic field disturbance signals under complex observation conditions. A deep learning method for detecting power frequency magnetic field anomalies is presented. By extracting the signal sequence of normal and disturbed states,the signal is input to the training of the neural network. A network model for detecting abnormal signal of power frequency magnetic field is obtained. The experimental results show that the recognition accuracy of this method is above 80%.