Sliding window based multi-scale morphology for detecting microseismic signal

Micro-seismic signal is contaminated with strong noise composed of random noise and low-frequency (LF) noise. Signal detection makes it essential way for micro-seismic data processing. Typical methods like band filtering can hardly separate signal and noise and damage signal in some degree. Mathematic morphological decomposition (MMD) is an effective tool to decompose data into several scales. Most random noise distributes in the 1st scale and LF noise distributes in last scale. Unlike traditional mathematic morphological reconstruction (MMR), each scale having fixed coefficient, we proposed sliding window based multi-scale morphological reconstruction (SWMMR). A window with certain length is used to segment scales along time axis with overlap. For segmented part of each scale in the same time window, we introduce least-square algorithm to calculate the coefficients. The reconstruction result suppresses background noise and detects weak signal. Successful performance of proposed method is demonstrated by synthetic and field micro-seismic data examples.