Abstract
Abstract. Many current indoor localisation approaches need an initial location at the beginning of localisation. The existing visual approaches to indoor localisation perform a 3D reconstruction of the indoor spaces beforehand, for determining this initial location, which is challenging for large indoor spaces. In this research, we present a visual approach for indoor localisation that is eliminating the requirement of any image-based reconstruction of indoor spaces by using a 3D model. A deep Bayesian convolutional neural network is fine-tuned with synthetic images generated from a 3D model to estimate the camera pose of real images. The uncertainty of the estimated camera poses is modelled by sampling the outputs of the Bayesian network fine-tuned with synthetic images. The results of the experiments indicate that a localisation accuracy of 2 metres can be achieved using the proposed approach.
Highlights
Indoor localisation is the key enabler of many applications like navigation guidance, location-based services and augmented reality
In Addition, we model the uncertainty of camera pose estimations by adopting a Bayesian convolutional neural network (CNN) (Kendall and Cipolla, 2016)
The second experiment was performed to measure the performance of the network fine-tuned with different types of synthetic images on real images
Summary
Indoor localisation is the key enabler of many applications like navigation guidance, location-based services and augmented reality. In the absence of global navigation satellite system signals (GNSS) in indoor environments, several approaches have emerged in the past two decades that include Wifi, ultra-wideband or radio frequency identification (Mautz, 2012). These approaches are dependent on a dedicated network of sensors and are often expensive. The key constraint for these approaches is the requirement of an initial location at the start of localisation (Se et al, 2002). Structure-from-motion (SfM) requires capturing a large number of overlapping images to estimate the camera poses and reconstruct the environment. The SfM approaches are computationally expensive and are susceptible to errors
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have