Autonomous Mapping and Navigation using Fiducial Markers and Pan-Tilt Camera for Assisting Indoor Mobility of Blind and Visually Impaired People
Authors:
Dharmateja Adapa,
Virendra Singh Shekhawat,
Avinash Gautam,
Sudeept Mohan
Abstract:
Large indoor spaces have complex layouts making them difficult to navigate. Indoor spaces in hospitals, universities, shopping complexes, etc., carry multi-modal information in the form of text and symbols. Hence, it is difficult for Blind and Visually Impaired (BVI) people to independently navigate such spaces. Indoor environments are usually GPS-denied; therefore, Bluetooth-based, WiFi-based, or…
▽ More
Large indoor spaces have complex layouts making them difficult to navigate. Indoor spaces in hospitals, universities, shopping complexes, etc., carry multi-modal information in the form of text and symbols. Hence, it is difficult for Blind and Visually Impaired (BVI) people to independently navigate such spaces. Indoor environments are usually GPS-denied; therefore, Bluetooth-based, WiFi-based, or Range-based methods are used for localization. These methods have high setup costs, lesser accuracy, and sometimes need special sensing equipment. We propose a Visual Assist (VA) system for the indoor navigation of BVI individuals using visual Fiducial markers for localization. State-of-the-art (SOTA) approaches for visual localization using Fiducial markers use fixed cameras having a narrow field of view. These approaches stop tracking the markers when they are out of sight. We employ a Pan-Tilt turret-mounted camera which enhances the field of view to 360° for enhanced marker tracking. We, therefore, need fewer markers for mapping and navigation. The efficacy of the proposed VA system is measured on three metrics, i.e., RMSE (Root Mean Square Error), ADNN (Average Distance to Nearest Neighbours), and ATE (Absolute Trajectory Error). Our system outperforms Hector-SLAM, ORB-SLAM3, and UcoSLAM. The proposed system achieves localization accuracy within $\pm8cm$ compared to $\pm12cm$ and $\pm10cm$ for ORB-SLAM3 and UcoSLAM, respectively.
△ Less
Submitted 16 October, 2023;
originally announced October 2023.
