DeepEyes: A Deep Vision Indoor Positioning System with Individual Movement Recognition Based on IoT

This paper designs and implements a deep vision indoor positioning system with individual movement recognition, called DeepEyes, based on Internet of Things localization. DeepEyes integrates deep learning with pedestrian dead reckoning, enabling the recognition of individual walking distances and moving angles. DeepEyes addresses the accuracy challenges in step length and heading direction estimation encountered by pedestrian dead reckoning methods, which often result in positioning errors that accumulate over time, impacting subsequent localization. In DeepEyes, we design deep vision indoor positioning using existing surveillance cameras for high-precision real-time localization and correction. To the best of our knowledge, this is the first centimeter-level positioning solution to combine deep neural networks with pedestrian dead reckoning methods to recognize individual movement distances and heading angles. In particular, an Android-based prototype with web cameras is implemented to verify the feasibility and performance of our DeepEyes system.