A Novel Single Camera Robotic Approach for Three-Dimensional Digital Image Correlation with Targetless Extrinsic Calibration and Expanded View Angles

Abstract

Three-Dimensional Digital Image Correlation (3D-DIC) is a technique used widely in both academic and industrial settings to measure object morphology, velocity, and full-field strain. However, conventional methods require the use of two synchronized cameras and a lengthy calibration process of taking thirty or more images of specialized target, after which the cameras must be kept fixed. This paper proposes a new approach for obtaining 3D-DIC measurements using only a single camera moved between different positions by a robotic arm, with the extrinsic calibration parameters obtained by manipulation of the internal position and orientation information known to the robot, obviating the need for a calibration target in order to calculate the extrinsic matrix for various positions (after the intrinsic matrix has been established). The specifications of a prototype are presented, and the method is quantitatively and qualitatively compared to the traditional 3D-DIC approach to determine relative accuracy and reliability in shape measurements. By correcting for errors in the reported orientation of the robotic arm, it was possible to achieve the same accuracy as the traditional 3D-DIC approach for stereo angles up to 30^o, beyond which the loss of image overlap results in linear increasing errors that are up to an order of magnitude greater. The method is also further extended to allow for images from more than two positions to be combined in a single analysis, which enables an expanded view of up to 360° using only a single camera and without the additional calibrations required by current multi-camera expanded view systems. The integration of this new 3D-DIC approach into future robotic systems will facilitate new capabilities for performing vital tasks, such as material analysis, infrastructure assessment, and human-robot interactions.