MapFuse: Complete and Realistic 3D Modelling

Abstract

Validating a 3D indoor radio propagation model that simulates the signal strength of a wireless device can be a challenging task due to an incomplete or a faulty environment model. In this paper, we present a novel method to simulate a complete indoor environment that can be used for evaluating a radio propagation model efficiently. In order to obtain a realistic and robust model of the full environment, the OctoMap framework is applied. The system combines the result of a SLAM algorithm and secondly a simple initial model of the same environment in a probabilistic way. Due to this approach, sensor noise and accumulated registration errors are minimised. Furthermore, in this article, we evaluate the merging approach with two SLAM algorithms, three vision sensors, and four datasets, of which one is publicly available. As a result, we have created a complete volumetric model by merging an initial model of the environment with the result of RGB-D SLAM based on real sensor measurements.