This study introduces ViTrack, an advanced vision-and-wireless-based navigation system that assists visually impaired individuals in indoor environments in a user-controlled, privacy-preserving manner. Traditional visual navigation methods often encounter challenges due to obstructed views, hindering effective tracking and identification. Similarly, wireless positioning systems are prone to signal interference, leading to inaccuracies. ViTrack seamlessly blends vision-based and wireless positioning technologies to provide comprehensive, unobtrusive navigation support. A key element of our approach is using existing surveillance cameras, expanding our observation capabilities while minimizing the need for additional devices for the user. Our system is designed to improve navigation while maintaining simplicity for the user. We exhibit ViTrack within an already-established surveillance camera network. It automatically employs low-cost wireless sniffers to detect users within the observation guiding zone. Simultaneously, vision-based methods ensure collision-free path planning. Our findings highlight the superiority of Long Short-Term Memory (LSTM) neural networks over other neural network architectures regarding localization and tracking accuracy. We also demonstrate the crucial role of data augmentation in mitigating device failure impacts, significantly boosting the system’s reliability.
Sou et al. (Sun,) studied this question.