Building an IndoorGML model in (near) real-time

Abstract

Afkomstig van de abstract van de scriptie:

Navigating indoor spaces requires indoor models. Depending on the use, such indoor models may need to be created quickly, such as in an emergency. Creating these models in real-time could help emergency responders to better interpret the situation by applying location-based services on the models that are created on-the-fly, such as navigation and spatial measurements. These services also requires a sufficient quality of the models. This work presents a methodology for automatically creating an IndoorGML model from a benchmark dataset in near real-time using a simulation of the scanning process. The timestamps of the dataset was used to simulate the scanning of the building. The rooms were identified and reconstructed by detecting the doorways and extracting the points scanned prior to detection. The results demonstrate the models' completeness but highlight challenges such as limitation in accuracy of the geometry and a delayed reconstruction of the doorways and navigation graph. This work contributes to science in a unique way by exploring how to create an indoor spatial model in real-time, and highlighting the challenges involved. At the same time, it also considers how to evaluate such a model.