基于改进MP的地震信号去噪

地震资料中的噪声分为相干噪声和随机噪声.针对随机噪声,采用匹配追踪（MP）算法进行去噪,能取得一定的去噪效果,但是MP算法过大的计算量严重影响去噪的效率.为解决这个问题,本文采用遗传算法（GA）寻找最优原子,大大减少了算法的运算量,提高了算法的运算速度.本文对Ricker子波进行改进,加入尺度参数、位移参数和相位参数,利用改进后的Ricker子波构建的过完备原字库进行去噪可以取得更好的效果.本文采用相邻残差比阈值作为迭代终止条件,与采用硬门限作为迭代终止条件相比增强了算法的鲁棒性.实验仿真时,首先对加有随机噪声的人工合成地震信号进行去噪,仿真结果表明：本文改进的去噪算法与基础MP算法相比,在信噪比、均方差和计算速度上具有明显优越性,然后将本文改进的方法应用于实际地震信号去噪,并取得很好的去噪效果.

Seismic Signal Denoising Based on the Improved MP

The noise of the seismic data is divided into coherent noise and random noise. Aiming at the random noise, using the matching pursuit （MP） algorithm to denoise can obtain certain denoising effect, but the computational burden of the MP algorithm is too large,which seriously affects the efficiency of denoising. In order to solve this problem, this paper uses the genetic algorithm （GA） to search for the optimal atom, greatly reducing the computational complexity of the algorithm and improving the operation speed of the algorithm. In this paper, the Ricker wavelet was improved, through joining the scaling parameters ,displacement and phase parameters, it can obtain better effect that using the improved Ricker wavelet to construct a over-completed dictionary to denoise. This article use the adjacent residual ratio threshold as the terminating condition,eompared with using hard threshold as the terminating conditions which enhances the robustness of the algorithm. Firstly,Compared with the MP algorithm ,the improved method in this paper is used to denoise synthetic seismic signal added random noise in the experimental simulation. Simulation results proved that the improved denoising method has obvious superiority in noise ratio, the mean square error and calculation speed, then the improved method can he applied to practical seismic signal denoising, and obtain good denoising effect.