Ghost vibration imaging via feature extraction from the spectral centroid

Vibration measurement is essential for applications ranging from structural monitoring to biomedical diagnostics. The integration of vibration sensing with imaging techniques offers a powerful approach for the comprehensive analysis of dynamic targets. This can be achieved by combining coherent detection with correlation imaging. A key challenge in such hybrid systems stems from the coupling among signal components, which complicates the simultaneous retrieval of distinct physical features. To address this, we propose a novel decoupling method that enables the blind extraction of both image and vibration characteristics of different parts, without requiring any prior information. We experimentally validate the method by simultaneously capturing and identifying the vibrations and images of multiple targets. Notably, the approach remains effective even when the vibrational states of the targets vary dynamically over time, demonstrating its robustness for target identification and classification. This work provides a simple and feasible pathway toward the seamless integration of vibration measurement and optical imaging.