Geometry constrained correlation adjustment for stereo reconstruction in 3D optical deformation measurements.

Abstract

Recovering the geometric shape of deformable objects from images is essential to optical three-dimensional (3D) deformation measurements and is also actively pursued by researchers. Most of the existing techniques retrieve the shape data with triangulation based on pre-estimated stereo correspondences. In this paper, we instead propose to recover depth information directly from images of a binocular vision system for 3D deformation estimation. Given a calibrated geometry of the system, the reprojection error is parameterized by the depth and then described with local intensity dissimilarity between a stereo pair in considering spatial deformation. Afterward, a correlation adjustment model is formulated to estimate the depth parameter by minimizing the error. As a solving strategy, we show the Gauss-Newton linearization of the proposed model and its initialization. 3D displacement estimation based on depth information is also presented. Experiments, including rigid translation and bending deformation measurements, are conducted to verify the performance of the proposed method. Results show that the proposed method is accurate yet precise in 3D deformation estimations. Other underlying developments are underway.