蜂巢式類神經網路於震測水平連接與圖型識別

Abstract

我們應用蜂巢式類神經網路(cellular neural network)於震測水平連接(seismic horizon linking)與圖型識別。蜂巢式類神經網路具有區域連接的特性，適合處理一些有區域規則性連接的特定運算，所以適合於震測水平連接。我們經由建構數種不同的peak分佈條件，以其形成的能量方程式和蜂巢式類神經網路之標準能量方程式做比較，完成網路訓練的過程，然後我們利用這完成訓練的網路來處理震測水平連接的問題。檢取震測水平反射層，將有助於震測資料的處理及震測資料的解釋。另一個應用是震測圖型識別。我們將蜂巢式類神經網路設計成聯想記憶體，然後運用此聯想記憶體辨識震測圖型。震測圖型識別將有助於震測資料之分析及解釋。

We apply cellular neural networks for seismic horizon linking and pattern recognition. Cellular neural networks have the characteristic of local connection. It is suited for some operations which have the characteristic of local regular connection. So it is suited for seismic horizon linking. We establish the energy function by setting several different constraints of peak distribution. And we compare this energy function and the standard energy function of a cellular neural network, and then finish the process of network training. Then we use this network which is trained to deal with seismic horizon linking. Detecting seismic horizons will help us to deal with seismic data and interpret seismic data. Another application is seismic pattern recognition. We design cellular neural networks to behave as associative memories, and then use the associative memories to recognize seismic patterns. Seismic pattern recognition will help us to analyze and interpret seismic data.