Journal of Graphic Science of Japan

Show abstractHide abstract

In recent years, three dimentional geometric shapes represented on computer systems have been widely used in areas both CAD/CAM (Computer Aided Design / Manufacturing) and Computer Graphics. However, input of such shapes, especially including surface shapes, requries much time and work. Many researches have been addressed for this topic, for examples, 3 D shape reconstruction from unorganized points from non-contact laserscanning machines, and 3 D shape recognition by stereos, and so on. In these methods, machines for measuring physical objects are generally expensive, and requires some special treatments for the measurement.
Our purpose is to recover 3 D shapes be use of a general 35 mm still camera, which are relatively low cost and easily obtained. We develop a new system pers for obtaining 3 D shapes from pictures. By using pers, 3 D shapes are reconstructed as follows : (1) Prepare several pictures for objects, and take into computers as digital images. (2) Using more than five points which positions in the world coordinates and corresponding image coordinates are known, pers calculate camera parameters (focal parameters of cameras, a camera rotation, position) . (3) Calculate 3 D points on the objects by interactively selecting two points between two points between two images to bet polygonal faces from 3 D points.

View full abstract

Download PDF (15950K)

Register with J-STAGE for free!