First-arrival picking is a necessary step in seismic data processing. Existing algorithms are either inaccurate or inefficient when the number of geophone groups is large and signal-to-noise ratio is low. In this paper, we propose the automatic first-arrival picking through convolution kernel construction and particle swarm optimization (FPCO) algorithm. First, abrupt energy fluctuations are detected using convolution kernels through convolution. Second, the locations of abrupt fluctuation are calculated in each convolution result to produce the index vectors. Third, the index vectors are integrated to one vector called the centerline by the Gaussian kernel. Finally, adaptive threshold is applied to obtain first arrivals around the centerline. The particle swarm optimization is applied to train the parameters. Experimental results on field datasets show that FPCO is more accurate and stable than popular algorithms.

初至拾取是地震数据处理的必要步骤。当检波器组数量多且信噪比低时，现有算法要么不准确，要么效率低下。在本文中，我们通过卷积核构造和粒子群优化（FPCO）算法提出了自动初到拣选。首先，通过卷积使用卷积核检测突然的能量波动。其次，在每个卷积结果中计算突变的位置以产生索引向量。第三，索引向量被高斯核整合为一个称为中心线的向量。最后，应用自适应阈值来获得围绕中心线的初到。应用粒子群优化来训练参数。