A Comparison of Sparse and Dense Point Approach to Photogrammetric 3d Modeling for Stone Textured Objects (case Study: Archeological Sites)

Three dimensional (3D) modeling has been an important process in documenting archeological sites. Unlike conventional 2D drawings, 3D models provide both archeologists and future reconstruction workers with accurate geometrical data of the object in digital form, thus enabling various experiments and research. For this cause, archeological institutions usually employ either the terrestrial laser scanner, or cameras using photogrammetric techniques. Photogrammetric techniques are usually employed due to their relatively low cost, simple equipments, and quick data acquisition. Photographs can then be processed into sparse point 3D model. However, with the introduction of advanced image matching algorithms, this technique can also generate dense point clouds similar to results from laser scanners. A prerequisite for this technique however, requires that the object in question possesses a texture with patterns in order to allow automatic image markings. In nature, stone textures in general provide a perfect example of this requirement. Dense point cloud generation is very useful especially to model intricate architectural details from objects such as roofs and reliefs. This has been done to document shrine No. 72, Sewu Temple Complex in Central Java, Indonesia. The purpose of this research is to determine whether the dense point approach is more effective and generates an altogether better result when compared to the sparse point approach when used to model a stone structure which is less intricate. In this case, Cangkuang Temple in West Java was chosen. Cangkuang Temple, unlike Sewu Temple, does not have intricate reliefs, but is still adorned with a tiered roof which can be time-consuming when modeled using the sparse point approach to achieve the same level of detail with dense point results. On the other hand, wherever possible, sparse point approach can minimize noise and holes on the 3D model prevalent when creating dense point models.