基于泊松碟采样的小生境遗传算法剩余静校正

针对线性剩余静校正容易陷入局部解，导致复杂山地地震资料成像效果不佳的难题，本文提出一种基于泊松碟采样的小生境遗传算法剩余静校正方法。在常规遗传算法剩余静校正基础上，应用泊松碟采样对求解的静校正量实数编码，增加初始解空间的均匀度；采用多线程构建多种群小生境，种群内采用“精英+赌轮”选择、余下个体与精英两两交叉、小扰动变异策略进行遗传进化；种群间进行“联姻”操作，优选几个种群中适应度最大个体并与其他种群联姻，充分利用计算机多核并行运算功能以快速搜索到最优剩余静校正量。理论模型及实际资料应用结果均表明，该方法能增强同相轴的连续性，显著提高山前带低信噪比资料成像品质。

Residual static corrections with the niche genetic algorithm based on Poisson disk sampling

For residual static corrections,linear methods are easy to fall into local solutions and lead to poor seismic imaging in complex mountain areas.Therefore we propose in this paper a new approach with a niche genetic algorithm based on the Poisson disk sampling.First based on cnventional residual static corrections,real number codes are calculated with the Poisson disk sampling to increase the uniformity of the initial solution.Then multi-populations of niches are constructed with multi-threads to estimate residual statics based on "marriage" operation.Multi-core parallel calculations of machines can quickly search optimal residual statics within the multi-populations and their "marriage".Both theoretical model tests and real data applications confirm that the proposed residual statics approach can increase the event continuity and improve low signal-to-noise ratio (SNR) data imaging.