不同插值方法在断层检测模型中的应用效果对比分析

Comparative Analysis of the Application Effects of Different Interpolation Methods in Fault Detection Models

为了了解同一断层检测模型在不同插值方法下的检测效果和同一插值方法在不同断层检测模型的检测效果。对比视觉几何组VGG(Visual Geometry Group)、多层感知机MLP(Multi-layper Perceptron)、支持向量机SVM(Support Vector Machine)三种断层检测模型在linear、nearest、cubic、area四种插值方法下的检测效果,其实现过程是利用CNN卷积层提取特征,全连接层分别用MLP、SVM作为分类器,建立MLP、SVM模型,根据地震剖面数据和断层数据构建样本,对VGG、MLP、SVM模型进行训练,训练好后通过模型数据测试。发现相同插值方法下SVM模型检测结果较好,同时发现同一模型中area插值法下检测效果最好。该测试表明:在选择的模型数据中使用area插值方法可以提升断层检测的性能。

In order to understand the detection effect of the same fault detection model under different interpolation methods and the detection effect of the same interpolation method under different fault detection models,the detection effect of three fault detection models,namely VGG(visual geometry group),MLP(multi-layper perceptron),and SVM(support vector machine)under the four interpolation methods of linear,nearest,cubic,and areaare compared.The realization process is to use the CNN convolutional layer to extract features,and use MLP and SVM as classifiers in full connection layer to establish MLP and SVM modelsrespectively.Then,samples are constructed based on seismic profile data and fault data,and VGG,MLP and SVM models are trained.After that,model data are tested.It is found that the SVM model detection result is better under the same interpolation method,and the area interpolation method in the same model has the best detection effect.The test shows that using area interpolation method in selected model data can improve the performance of fault detection.