Subpixel motion estimation for video-based target-free vibration monitoring under complex environmental conditions
Résumé
An emerging technique to measure structural vibrations is based on motion signals extracted from video images. The use of video cameras offers multiple advantages over traditional mechanical sensors: contactless measurement, large coverage, easy installation and maintenance. This paper proposes a new method for real-time motion signal extraction from video images with subpixel accuracy. It aims to address the challenges posed by complex operating conditions, namely illumination variations and background interference. Illumination robustness is achieved by efficiently combining image intensity interpolation with an affine brightness and contrast tuning transformation. Background robustness is obtained by automatically selecting active pixels in the processed images. Moreover, the on-line numerical computations are fast enough for real-time applications. Results of theoretical analysis ensure that the considered optimization criteria are well-posed and that, in the main step of subpixel motion estimation, the involved nonlinear optimization problem is efficiently solved in closed-form through a linear least squares problem. The performance of the proposed method is evaluated both on simulated images and on laboratory experiments with a target-free cantilever beam in comparison with existing methods. The reported results demonstrate the robustness and computational efficiency of the proposed method under complex environmental conditions, allowing real-time computation on a standard laptop for vibration monitoring with more than 100 virtual sensors at 600 frames per second.