Vision-Based Rigid-Body and Deformation Motion Extraction of High-Speed and Flexible Mechanisms with Multiple Connected Components

Abstract

A novel approach is presented for the experimental motion extraction of mechanisms with multiple connected components. By exploiting vision-based measurement techniques, the presented method is able to extract both the rigid-body motion and the deformation motion of the mechanism components. To achieve this, a motion tracking method is proposed which combines a motion-by-detection approach (i.e., through consecutive marker detection) with a motion-by-tracking approach (i.e., through optical flow). It will be shown that the methodology is sub-pixel accurate, is able to detect occlusions and can track motions with a large displacement between the frames of the camera recording. These are three current challenges when working with vision-based motion tracking on high-speed, multi-component mechanisms. The extracted motion is then decomposed into its rigid-body and deformation motion components by using a Procrustes method. Lastly, the dominant proper orthogonal deformation modes of the flexible components are calculated by using a singular value decomposition. The presented approach is experimentally validated on two cases. The first validation case shows the capabilities of the approach on a translational and partly occluded motion of the weaving frame of a weaving loom mechanism. The second validation case shows the capabilities of the approach on a combined rotational-translational motion of a slider-crank mechanism.

Keywords: Vision-based motion measurements, Optical flow, Procrustes, Multi-component mechanisms, Deformation modes