Research Of Seismic Signal Denoising Based On Sparse Decomposition Algorithm

With the deepening of the geological exploration, geological structure and geologicalenvironment of exploration area is increasingly complicated, which will damage theregularity and completeness of the seismic data, making the processing and interpretation ofsubsequent seismic data the big trouble, ultimately affecting the judgment of stratumstructure and the exploration of oil and gas reservoirs. So how to improve the resolution andsignal-to-noise ratio of seismic data have become the primary problems many scholarsstudy.This article mainly research seismic signal denoising based on sparse decompositionalgorithm.This paper first introduces the classification and formation cause of noise in theseismic data, as well as the main denoising method, and the principle, the advantages anddisadvantages of various common denoising method in this paper, the introduction of a newdenoising method and train of thought, namely the denoising algorithm based on sparsedecomposition.Then the paper introduces the basic theory of sparse decomposition,emphatically introduces principle of matching pursuit algorithm and denoising of matchingpursuit algorithm and advantages, and select the Ricker wavelet whose structure is closer toseismic signal as the time-frequency atom to build a over-complete atom dictionary, whichcan achieve very good denoising effect.Aimed at the great calculation amount problem of matching pursuit algorithm, thispaper puts forward two kinds of optimization algorithm, one is: using the geneticoptimization algorithm to find the best atoms, greatly reducing the running time of thealgorithm, at the same time using adjacent residual ratio threshold as the terminatingconditions to improve the reconstruction accuracy of the atom, it can achieve betterdenoising effect, which is demonstrated in simulation experiment. The second is: geneticalgorithm(GA)has shortcomings of the early maturity and weak local searching ability, thearticle puts forward a kind of sparse decomposition denoising algorithm based on simulatedannealing genetic optimization algorithm. It generate initial population and select operationbased on the hamming distance between individuals in a population and adopt adaptivecrossover operator and mutation operator to improve the genetic algorithm, simulatedannealing mind is embedded genetic algorithm at the same time, use simulated annealingoperation to best individuals produced genetic operation in population of newgeneration,the best individual in the group of the previous generation is original solution,the best individual in the group of new generation is new solution, use the Metropoliscriterion to process data, it can improve the global search ability of the algorithm, whichincreases the ability and efficiency of the improved algorithm to search the best atoms.Theimproved algorithm reduced the amount of calculation, and can achieve very gooddenoising effect.In simulation experiments, using two kinds of optimization algorithm and the MP algorithm to denoise doppler signal, block signal, synthetic seismic signals, which addednoise and real seismic signal, through the denoising effect, comparing and analyzing theexperimental data, it is proved that the effectiveness of the two kinds of optimizationalgorithm.