Calibration and 3D Reconstruction of Omni-directional Multi-view Image Measurement System

Abstract

This article presents an omnidirectional multi-view image measurement system to provide multi-camera 3D reconstruction and multi-view image information. The system consists of four multi-view cameras, one of which is the fixed center camera that measures the geometric center point and provides a top-view image. The other three cameras provide side side-view images. They are 120 degrees apart in a sphere and are adjustable on arc rods. Any camera can be moved along arc measuring rods to a specified position to capture images within the sphere. Altering the angles among the cameras, the range of images captured by the cameras is changed accordingly. If the multi-view images of four cameras are captured at the same time, any two cameras are used to form a stereo camera pair. Using the principle of parallax matching between left and right images, stereo camera pairs can be formed for three-dimensional reconstruction. Finally, through the verification of calibration and 3D reconstruction examples, it is shown that the system can be successfully applied to omnidirectional multi-view image measurement and 3D reconstruction.