Abstract:Manual velocity spectrum pickup results in low efficiency and poor accuracy because both 3D seismic survey and high-density 2D seismic survey have very great data volume. So, people advance continuously many new methods to solve the problem. I put forward a new method which involves neural network and fuzzy mathematics. The method works in the following steps: ·Make boundary search and clustering processing of the input data with the aid of fuzzy mathematics. ·Learn the given sarnpics to train network model by using I3P algorithm for neural network error. ·Input the preprocessed data for recognition. ·Automatically pick up stack velocity after the data recognition. The method has good noise-resistance and offers accurate velocity, so that velocity spectrum can be interpreted automatically.