Grammar-based reconstruction of building interiors by mobile participants

点击翻译按钮获取中文摘要

Maps are an essential prerequisite for location-based services. Although street maps and even detailed 3D city models are already available today, indoor models are still largely missing due to the great manual effort of creating detailed indoor models. Collaborative approaches for indoor modelling try to solve this problem by deriving indoor models automatically from sensor data like movement trajectories and camera images gathered by a crowd of mobile users. Therefore, the major goal of the second funding period is to design methods for the automatic creation of 3D indoor models from collaboratively gathered sensor data. Starting point are the methods designed during the first funding period for the creation of 2D indoor models from movement trajectories gathered by crowds of mobile devices. A second important foundation is the concept of indoor grammars designed in the first funding period, which describe structural rules of 2D floor plans. During the first funding period, we could show that indoor grammars are highly beneficial to improve model quality as well as energy efficiency of mobile devices participating in collecting sensor data. Therefore, also the second funding period focuses on grammar-based methods, now to facilitate the generation of indoor models in three dimensions. The transition from the grammar-based modelling of 2D floor plans to inherently more complex 3D models including corridors and room structures of different heights as well as interior objects contained in rooms requires enhanced grammar concepts and methods for the generation of such 3D indoor models. A second task is the extension of grammars by semantic information and the utilization of this information for model generation. Similarly to the first funding period, energy efficiency remains an important goal also of the second period. Additionally to the energy efficient gathering of sensor data by mobile devices, the energy efficient processing of gathered data becomes a great challenge. This raises the question for an optimal distribution of model generation functions between mobile devices and server infrastructure. Closely related are investigations for the energy efficient communication of information required for the distributed gathering and processing of sensor data.