Automatic Generation of Three-Dimensional Model with Reduced Data Size Based on Point Cloud Measurement

3D modeling by 3D scanning of real objects using laser scanners is effective for surveying disaster sites to confirm the extent of damage and for digital archiving of tangible cultural assets to preserve their shapes. In recent years, smartphones and tablets are equipped with distance measurement sensors, making it possible to easily perform 3D scanning. However, generating a 3D model by extracting only the necessary data that constitutes an object is not easy. The point cloud acquired by a ranging sensor consists of a very large number of points and includes many points other than the object and noise. It has generally been necessary for a person to visually designate target objects and unnecessary areas. However, the large data size, such as data from a wide area of the surrounding environment, increases the amount of work required. In addition, if the object has a complex shape, the process of defining the area of the shape itself is complicated. This requires careful attention to delete unnecessary areas and save necessary areas, making it difficult to specify the area appropriately. A 3D model with a large data size is undesirable for data transfer over information networks and for the execution of various processes. In this paper, we design an algorithm for generating lightweight 3D models with reduced data size based on point cloud surveying and demonstrate its effectiveness.