Generation of 3D surface model of complex objects based on non-metric camera

A reconstruction method for complex objects was developed using non-metric images. Firstly, a rotatable platform was designed to capture images and a planar grid board on its top was used for camera calibration. With the help of the platform, images were captured and then processed with a multi-baseline image matching algorithm to produce corresponding point features. The initial external elements of the images can be obtained with traditional photogrammetry approach. These parameters and coordinates of corresponding points were input to block bundle adjustment as initial values, and then point clouds and accurate image parameters were obtained. In the end, a visibility constrained Delaunay triangulation algorithm was used to produce surface models. The methods include artful hardware design, efficient image matching, and rational 3D surface reconstruction. It is low-cost and automatic, and experiment results demonstrate its feasibility and effectiveness.