Coupled Multiview Vision and Physics-Based Synthetic Perception for 4-D Displacement Field Reconstruction

Abstract

This paper presents a novel method that takes advantage of real-world multiview vision measurements to identify parameters needed in physics-based models for creating synthetic perception of a 4-D displacement field. To distinguish the 4-D reconstruction from traditional graphic reconstructions in 3-D space, “time” plays the fourth dimension to provide an essential basis for kinematic or dynamic analyses based on physical fields. The coupled real-vision and synthetic-perception method is illustrated in the context of a compliant multibody mechatronic system where the displacement field of the deformable and rigid bodies are reconstructed from a multicamera system. The coupled method has been experimentally validated on a test case emulating the compliant system negotiating a right-angle corner in tight 3-D space; perception results of physics-based models closely agree with graphical details reconstructed from vision images. The synthetic perception results dynamically unveil hidden details, which are otherwise obscured in vision images in a virtual environment, demonstrating the effectiveness of the coupled method in overcoming problems commonly encountered in either vision or physics-based modeling.