The tree classification method is proposed for the detection of bright spots in the pinchout pattern. Envelope, instantaneous frequency, and polarity are used as the features. Envelope is selected as the first-level feature, and instantaneous frequency is selected as the second-level feature. An algorithm for the automatic selection of the training traces is proposed. A nonsupervised clustering analysis is performed by using envelope and instantaneous frequency as the features. After determining the tree classifier of envelope and instantaneous frequency from training traces, a classification experiment is then performed on the seismogram of the pinchout pattern. Using these intermediate classification results, polarity is computed and bright spots are detected if the hypothesis is satisfied. The classification result shows that the proposed method can be used to improve automatic seismic interpretation.
|出版狀態||Published - 1 1月 1984|
|事件||1984 Society of Exploration Geophysicists Annual Meeting, SEG 1984 - Atlanta, United States|
持續時間: 2 12月 1984 → 6 12月 1984
|Conference||1984 Society of Exploration Geophysicists Annual Meeting, SEG 1984|
|期間||2/12/84 → 6/12/84|