Next Generation 3D-DIC Technique with Sensor-Based Extrinsic Parameter Calibration and Natural Pattern Tracking

Abstract

Three-dimensional Digital Image Correlation (3D-DIC) is a computer vision technique capable of extracting full-field displacements of structures by processing images acquired from synchronized stereo cameras and performing triangulation. To properly reconstruct the 3D points in space, the stereo cameras must be calibrated to compute the lens distortion (i.e., intrinsic parameters) and the cameras’ relative position and orientation (i.e., extrinsic parameters). Traditionally, calibration is performed by taking pictures of a calibration target (e.g., a checkerboard). However, the calibration target must occupy most of the cameras’ field of view (FOV) and have a size comparable to that of the object being tested, which becomes impractical for large-scale FOVs. Additionally, 3D-DIC requires the application of a stochastic speckle pattern on the surface to identify and track unique points across multiple images. This can be a problem when structures on which the application of a speckle pattern is impractical have to be analyzed (e.g., wind turbine blades). This research proposes a multi-sensor system using three inertial measurement units and a laser distance sensor to compute the extrinsic parameters of a stereo vision system and accelerate the calibration procedure. In addition, a feature-based natural pattern tracking algorithm is proposed to exploit features that are naturally present on the structure instead of applying a speckle pattern to perform 3D-DIC. Laboratory tests show that cameras calibrated using the multi-sensor system reconstruct displacements with an accuracy above 95% compared to displacements measured using a traditional calibration method. At the same time, the proposed feature-tracking algorithm can identify and reconstruct the 3D positions and displacements of preexisting features without requiring an applied speckle pattern. The research shows that the proposed methods can simplify and streamline the use of 3DDIC and open the way to the use of this technique to perform condition monitoring of large-scale civil and mechanical engineering structures.