Dynamics-Based Stereo Visual Inspection Using Multidimensional Modal Analysis

This paper proposes the concept of multidirectional-modal-parameter-based visual inspection with high-frame-rate (HFR) stereo video analysis as a novel active sensing methodology for determining the dynamic properties of a vibrating object. HFR stereo video is used for observing the 3-D vibration distribution of an object under unknown excitations in the audio-frequency range, and the projections of the vibration displacement vectors along multiple directions can be verified using output-only modal analysis that can estimate their modal parameters such as resonant frequencies and mode shapes. Through implementing a fast output-only modal parameter estimation algorithm on a 10 000-fps stereo vision platform, we developed a real-time multidirectional-modal-parameter-based visual inspection system; it can measure the 3-D vibration displacement vectors of 30 points on a beam-shaped object from 512 x 96 pixel stereo images at 10 000 fps and can determine its resonant frequencies and mode shapes along 72 different directions around its beam axis as its input-invariant modal parameters. To demonstrate the performance of our system in modal-parameter-based visual inspection, the asymmetric dynamic properties, caused by cracks, of several steel beams vibrating at dozens of hertz and having artificial cracks were inspected in real time by determining the modal parameters along 72 directions around their beam axes.