SEISMIC VELOCITY PICKING BY HOPFIELD NEURAL NETWORK

Abstract

The Hopfield neural network (HNN) is adopted for velocity picking in the time-velocity semblance image of seismic data. A Lyapunov function is generated from the velocity picking problem. We use the gradient descent method to decrease the Lyapunov function and derive the equation of motion. The Lyapunov function can reach the minimum. According to the equation of motion, each neuron is updated until no change. The linking of the converged network neurons represents the best polyline in velocity picking. We have experiments on simulated seismic data. The picking results are good. It can improve the seismic data processing and interpretation.