| Summary: | 碩士 === 國立中央大學 === 機械工程學系 === 106 === The shape from silhouette (SFS) method is a simple and fast technique for reconstructing the three-dimensional (3D) model of an object from a series of two-dimensional images. It only employs the silhouettes of the object images to generate a set of cloud points describing the object surface. However, the problem is that it may result in some vertices that do not really exist on the object. Hence the resulting 3D model may have artifacts, indicating that the shape of the reconstructed model is different from that of the real object. The purpose of this study is to develop a method for improving quality of the SFS model. There are mainly three algorithms in the proposed method. First, a new triangulation algorithm is proposed for improving the quality of the triangular model obtained from the cloud points. The proposed algorithm mainly inserts new vertices for improving the occurrence of long-and-narrow triangles that might occur in the original method. Second, a Laplacian smoothing algorithm is employed for removing artifacts and improving the smoothness of the triangular meshes. Finally, to maintain the accuracy of the silhouette of the 3D model projected onto each image, an error controlling algorithm by comparing both silhouettes of the projected model and the image is proposed. The second and third algorithms are formulated as an iterative procedure for optimizing the 3D model. The final 3D model would be closer to the shape of the real object after the above-mentioned optimization process. Three object examples are employed for demonstrating the entire procedures and the feasibility of the proposed method.
|
|---|
