海底双相随机介质声反射的SVM神经网络分类识别研究

考虑海底沉积介质为双相介质,为了更好地模拟实际海底底质的不均匀性,将随机介质理论引入双相介质理论。首先,通过基于随机-双相介质理论的高阶有限差分数值技术模拟计算海底底质分别为泥质砂、泥、泥质砾时的地震反射波信号。然后利用小波变换分别求取不同底质的一次反射波的包络作为其特征向量,最后利用基于粒子群智能算法优化的支持向量机神经网络对这些反射波信号进行分类识别。为了进一步考察所用方法的抗噪能力,对正演得到的海底底质反射波信号分别加入10%、30%、50%的高斯白噪音之后再进行分类,支持向量机仍然取得了较好的分类预测效果。基于上述正演模拟及分类识别方法的论证,提出了一套行之有效的微机软件模拟海底沉积物分类识别的一般化流程,这将有利于开展海底沉积物反射特征的进一步研究。

The use of SVM to classify the reflection from submarine random two-phase medium

In this paper,to better simulate the actual heterogeneity of the seabed sediment,the random medium theory is introduced into the two-phase medium theory. Firstly,through the high-order staggered-mesh finite different simulation of random two-phase media,simulated the propagation of the seismic wave of three different the sediments,which are shaly sand,mudstones,muddy conglomerate. Then,the wavelet transformation technology is used to obtain the envelopes of reflection,called as the feature vector,which will be used as the input term of neural network. Finally,support vector machine neural network based on particle swarm optimization was applied to classify these data. To further investigate the anti-noise ability of the proposed method,the 10%,30% and 50% of Gaussian white noise was added into the original data and the optimized support vector machines still achieved good classification prediction. Based on the repeatable,convenient of the computer simulation and the relevant high accuracy and the robustness of SVM,a total solution of a classification,which will be easier,deeper,further to sturdy the feature of reflection of sediments is proposed in the article.