Generalised discriminative optimisation algorithms for augmented reality: theory and practice

Gao, Qinghong and Zhao, Yan (2021) Generalised discriminative optimisation algorithms for augmented reality: theory and practice.

Augmented Reality (AR) technology achieves the seamless registration between virtual scenes and the real world. Three-dimensional registration is a core method for integrating virtual information and the real world. Most registration experiments are conducted on simple data models (such Standford Bunny model and the Bimba model) currently. Actually, augmented reality is significantly related to registration and object tracking in real scenes. In order to verify the better performance of the proposed algorithms, we created the point clouds of real scenes to evaluate the performance of the proposed algorithms on object tracking and the registration of scenes and models. We use KinectV2 to attain the depth images and RGB images of the moving models (bunny model, chicken model and parasaurolophus model ) standing in the hand of a people. Then point clouds are reconstructed via camera calibration. And these reconstructed point clouds involve to rotation, outliers, occlusions and other perturbations, which is enough to evaluate the performance of any registration algorithms.