3D Surface Reconstruction From 2D Binary Images

We introduce methods for reconstruction of three dimensional surfaces. They are based on moving the light source relative to an object whose shape is to be determined. Using a camera which is placed above the object and a moving light source, the shadows cast by the object at each angle are recorded and analyzed. In one method, the light source is rotated in a horizontal plane; this method is a simple way to get segmentation between different surfaces. The second method uses a light source which is rotated in a vertical plane. For each section of the object a shadow diagram (Shadowgram) is formed and analyzed to get the third dimension. The Shadowgram has some features which make the reconstruction very simple. By looking at some curves of the Shadowgram, invisible surfaces can be partially reconstructed. The above methods can be combined to achieve simple solutions for problems in the robotics area e.g., the bin-picking problem. A set of experimental results demonstrates the robustness and usefulness of the methods.