End-to-end Image-based Indoor Localization for Facility Operation and Management

Recent research on facility management has focused on leveraging location-based services (LBS) to assist on-demand access to model information and on-site documentation. Researchers highlight that fast and robust indoor localization is of great importance for location-based facility management services, especially considering when using facility management services in a mobile computing context. Despite the importance of location data, most existing facility management systems do not support LBS due to the following rea- sons: 1) Signal-based indoor localization methods, such as WIFI, RFID, Bluetooth and Ultrasound, require installation of extra infrastructure in a building to support localization, 2) Visual-based indoor localization methods, such as LiDAR and camera, still depend on feature point detection and matching that require heavy computation and can be impacted by environmental conditions. In this paper, the authors present an end-to-end image-based localization framework to support facility operation and management using a convolutional neural network (CNN). The proposed framework contains two modules: mapping and localization. The mapping module takes a set of training images with their pose as input and trains a CNN model for the localization module. The localization module takes a single image and the trained model as input and outputs estimated camera pose (position and view angle). Compared to conventional methods, the proposed end-to-end image-based indoor localization framework does not require any infrastructure installed in a building and can achieve real-time 6-DoF localization that is robust to different lighting conditions and scenes with poor texture. The proposed framework was evaluated on publicly available datasets and the results show that end-to-end imaged-based method can achieve real-time 6-DoF localization with acceptable accuracy and a small map size.