Automatic First Break Detection Using Support Vector Machine Classification

An automatic first arrival detection method using support vector machine classification is presented in this paper. Evaluation of a 2D seismic data set confirms that this multi-attribute classification method can identify first break half cycles with high accuracy. The calculated first break times are of high quality. Classification results can be used to assign a quality indicator to each of the detected first arrival times. Moreover, by thresholding the quality indicator one can quickly eliminate any mis-picks which have closer distance to created SVM.