ViNav: A Vision-Based Indoor Navigation System for Smartphones

Abstract

Smartphone-based indoor navigation services are desperately needed in indoor environments. However, the adoption of them has been relatively slow, due to the lack of fine-grained and up-to-date indoor maps, or the potentially high deployment and maintenance cost of infrastructure-based indoor localization solutions. This work proposes ViNav, a scalable and cost-efficient system that implements indoor mapping, localization, and navigation based on visual and inertial sensor data collected from smartphones. ViNav applies structure-from-motion (SfM) techniques to reconstruct 3D models of indoor environments from crowdsourced images, locates points of interest (POI) in 3D models, and compiles navigation meshes for path finding. ViNav implements image-based localization that identifies users' positions and facing directions, and leverages this feature to calibrate dead-reckoning-based user trajectories and sensor fingerprints collected along the trajectories. The calibrated information is utilized for building more informative and accurate indoor maps, and lowering the response delay of localization requests. According to our experimental results in a university building and a supermarket, the system works properly and our indoor localization achieves competitive performance compared with traditional approaches: in a supermarket, ViNav locates users within 2 seconds, with a distance error less than 1 meter and a facing direction error less than 6 degrees.

A Hybrid Marker-Based Indoor Positioning System for Pedestrian Tracking in Subway Stations

“…Dong et al [27] proposed a system called ViNav that could be used without installing indoor maps, radio maps, or additional hardware. It collected photos or videos through crowdsourcing and 3-D modeling of the interior structure using structure-from-motion technology.…”

“…Methods that utilize vision, such as [26,27], often process queries on the server because the required memory and computational requirements of the image data are large. This makes it difficult to scale up the system and reduce the positioning response time.…”

“…Therefore, because the positioning targets are not assumed to be users and positioning is performed through a public camera, privacy issues may arise. Much the same way that selecting partitions for image matching using a Wi-Fi fingerprint reduces complexity in [27], the movement distance and direction are estimated using an IMU sensor and Wi-Fi channel state information (CSI) for feature matching in [31]. When a user takes two pictures of the same place at slightly different locations, spatial information is derived by obtaining data on the location and direction in which the two pictures were taken through the IMU and CSI.…”

Cell-Free Massive MIMO for Wireless Federated Learning

“…One reason for this trend is the fast growing number of mobile devices, wearable devices and autonomous vehicles. They are generating a vast amount of data by using in-built sensors, e.g., microphones, GPS and camera, for critical applications such as traffic navigation, indoor localization, image recognition, natural language processing, and augmented reality [3]. In addition, the computational capabilities of these devices also grow significantly with dedicated hardware architecture and computing engines, e.g., the energyefficient Qualcomm Hexagon Vector eXtensions on Snap-dragon 835 [4].…”

A Learning-Based Credible Participant Recruitment Strategy for Mobile Crowd Sensing

“…Mobile crowd sensing has promising applications in many domains, e.g., indoor maps reconstruction, where participants collect sensing data (such as photos or videos) to reconstruct 3D maps [8]. However, a common challenge for most mobile crowd sensing applications is to identify participants who can contribute sensing data that meets the requirement of the task, then to motive them to collect and contribute high quality data [9].…”