非常快速模擬退火與基因演算法於圖型偵測及震測圖形識別之應用

Abstract

我們採用4種最佳化演算法於圖形參數偵測及震測圖形識別，此4種演算法為模擬退火、快速模擬退火、和非常快速模擬退火、及基因演算。快速模擬退火演算法與非常快速模擬退火演算法是全域最佳化演算法，改良自傳統的模擬退火演算法。除了有機會避免局部局部最佳化，兩者在搜尋過程上均較傳統模擬退火演算法有較快的收斂。基因演算法也是一種全域最佳化演算法，基因演算法的優點也在於避免局部最小值。我們利用這4種演算法並建立圖形參數偵測系統，並且提出循序式的圖形參數偵測方式偵測影像中的直線、雙曲線、圓和橢圓。偵測參數採用階段性，可降低計算量及能快速收斂。系統能找到一組圖形參數向量使得圖形到影像上的點為 最小的距離。 在實驗的部分，圖形參數偵測系統配合4種演算法應用到偵測影像的圖形、單炸點震測訊號的圖形、和common depth point (CDP) gather的圖形。系統可偵測單炸點震測訊號中的直接波 (直線) 與反射波 (雙曲線)。系統可偵測CDP gather圖形中的雙曲線，由偵測到的參數，求出 root-mean-squared 速度，再作normal moveout correction (NMO) 修正，及stacking，得出stacked震測圖，此過程提供了自動化的速度分析方法，有助於震測的解釋與進一步的分析處理。

We adopt four global optimization methods to the parameterized pattern detection. They are simulated annealing (SA), fast simulated annealing (FSA), very fast simulated annealing (VFSA), and genetic algorithm. FSA and VFSA not only avoid local optimum with probability, but also have faster convergence than SA. Genetic algorithm (GA) is also a global optimization algorithm to avoid local minimum. We use these four global optimization methods to design the pattern parameter detection system. Also we propose the sequential methods to detect three types of patterns that include the lines, hyperbolas, circles and ellipses in the image. We use steps in the parameter detection for reducing the computation and getting fast convergence. This system has the capability of searching pattern parameter vectors with global minimal distance between the patterns and the image data. In the experiments, this system with four methods is applied to the image data, one-shot seismic data, the common depth point (CDP) gather data. The system can detect the parameters of direct wave (line) and reflected wave pattern (hyperbola) in the simulated and real one-shot seismograms. The system can detect the hyperbolic patterns in CDP gather data. The detected hyperbolic parameters are used to calculate the root-mean-squared velocity Vrms of the layers. Then we use Vrms to process the normal-moveout (NMO) correction. After stacking, we can get the stacked seismic signals. This system can provide an automatic velocity analysis and can improve the seismic interpretation and further seismic data processing.