Large-field-of-view three-dimensional imaging from multiple virtual views

We present a virtual multiview imaging method that incorporates a pair of Risley prisms and a stationary camera coaxially mounted. By rotating double Risley prisms, the virtual array camera (VAC) is generated to capture multiframe images from different perspectives for three-dimensional (3D) reconstruction, which not only extends the field of view (FOV) but also generates rich point cloud data. The VAC imaging process, including the virtual camera imaging theoretical model, automatic virtual optical center calibration, point cloud data acquisition, and target pose estimation, is developed to enable virtual camera field construction. Moreover, an improved iterative closest point registration algorithm by combining the overlapping prepositioning method is proposed to improve the speed of multiframe point cloud stitching. The experiments on real data validate the feasibility and flexibility of our 3D imaging technique capable of realizing large-FOV 3D imaging and accurate pose estimation of the target, which demonstrates a promising application prospect in visual imaging fields.