Construction of a 3D surface model from a series of 2D digital pictures of a solid object
Chowdhury, Tarik Ahmed (2003) Construction of a 3D surface model from a series of 2D digital pictures of a solid object. Master of Engineering thesis, Dublin City University.
Full text available as:
As the virtual world demands ever more realistic 3D models, attention is being increasingly focused on systems that can acquire graphical models from real objects. In this research, a system has been developed to create 3D surface model from a series of 2D digital pictures of a solid object. A rotating table and a calibration rail have been designed and constructed as part of the system, to capture images of a solid object at certain rotational angles and calibrate the camera respectively. A tripod is used to fix the camera in a certain position and background colour is selected depending on the colour of the object during image capturing. 3D Model software was developed by the author using Visual C++, as part of the proposed system, and was used to create 3D surface models of solid objects automatically from captured digital images. 3D surface models are created based on a “silhouette based” approach. In this process, quadtree is created from the silhouette of the object image and octree is created from the voted refined quadtree. Texture mapping was performed on the octree of the object using surface particles. Surface particles are created from the outer surface of the edge octree and their colours are captured by projecting them on the proper original image. For most objects without concavities, a fairly good approximation of the surface of the objects can be obtained using the proposed system. But the proposed system still has the same limitation as previous silhouette based volumetric construction algorithm, that is, a limited precision and an inability to detect concavities. The effect of shadows and spotlights were not considered in the proposed system.
Archive Staff Only: edit this record