Phase motion estimation method via multi-scale and multi-directional adaptive Gabor filtering

Abstract Accurate measurement of vibration displacement is essential for industrial structural health monitoring and intelligent manufacturing. Phase motion estimation (PME) methods based on optical flow have gained attention for their high sensitivity and non-contact advantages. However, conventional PME approaches often suffer from degraded accuracy and stability in the presence of complex textures, multi-scale features, and multi-directional motions, primarily due to fixed filter parameters and limited directional adaptability. To overcome these challenges, this study proposes a novel PME method via multi-scale and multi-directional adaptive Gabor filtering. By constructing a Gabor filter bank with diverse scales and orientations, the method effectively captures local structural features in the spatial-frequency domain. A scale selection strategy using locally energy-weighted distributions dynamically adapts filter scales to match structural characteristics. Additionally, an amplitude-orientation consistency weighting mechanism enables adaptive alignment of filter parameters with motion direction, enhancing robustness. Experimental results demonstrate that the proposed method significantly outperforms conventional PME approaches, achieving a correlation coefficient above 96% with reference measurements across diverse excitation amplitudes, motion directions, and imaging resolutions. These findings highlight the method’s strong potential for accurate, robust, and non-contact vibration measurement in both laboratory and real-world applications.