基于神经网络的探地雷达探雷研究

根据对探地雷达回波信号的分析，提出了一种新的适合于现场处理的探雷方法，采用延时校正、维纳自适应滤波及减背景等手段对原始数据进行预处理，提取Welch功率谱密度估计作为目标的特征，将特征送入神经网络进行分类训练。使用地雷目标与共相近的物体的数据进行对比试验和神经网络测试，结果表明，经过训练的神经网络可有效地将地雷与其它干扰物分开。

Study of mine detection with GPR based on neural networks

s Based on the analysis of ground penetrating radar returns a novel mine detection method suitable for on site processing is proposed.First the original data is preprocessed by adopting time adjusting,Wiener adaptive filter and background subtraction approaches, and then object feature is extracted using Welch power spectral density estimation.At last the extracted features is used for training the networks.The network is tested on data containing mines and some other objects which are like mines in shape and size, and the results indicate that the network can distinguish above objects effectively.