Content of research

We are researching technologies for the automatic recognition and modeling of objects and structures in indoor and outdoor environments from point clouds obtained by ground-based and vehicle-mounted 3D laser scanning systems, as well as basic point cloud data processing methods. The objects to be recognized and modeled include a wide range of arbitrarily shaped objects, rooms, pole-like objects such as utility poles and street lights, trees, road surfaces, and buildings. In addition to the technology of generating mesh models, polygon models, and CAD models from point clouds, we also conduct research on point cloud registration, segmentation, shape feature extraction, machine learning, and procedural object recognition that serves as the foundation for the technology. This technology enables detailed recognition and analysis of the environment and structures, maintenance and management, various simulations and improvement plans using 3D models that faithfully reflect the current state.

Potential for social implementation

• ・3D map generation and 2D drawing generation
• ・Analysis and simulation of the living environment
• ・Support for the maintenance and management of structures and infrastructure
• ・Automatic driving and autonomous robot behavior support

Appealing points to industry and local governments

We are researching the most important point cloud processing methods for the use of 3D laser scanning technology, which will become increasingly popular in many fields in the future. We develop theories and algorithms to robustly and efficiently extract and model the information on objects required for operations from point clouds of various qualities, according to the actual environmental conditions and needs, and develop software that implements these theories and algorithms.