A Smartphone-Based Affordable NavigationSystem using Visual SLAM for Blind and VisuallyImpaired People

Navigation guidance is crucial for blind and visually impaired (BVI) people. However, current navigation tools face challenges such as cumbersome portability, and limited cognitive capabilities. With the advancement of smartphone technology, it is convenient to utilize visual sensors to perceive the surrounding environments. Furthermore, the vision-based simultaneous localization and mapping (SLAM) makes it possible to build high-accuracy navigation maps. Therefore, we propose a smartphone-based navigation method for the BVI, which leverages the SLAM by purely adopting a stereo camera as a sensor. Our system can be decomposed into two stages: (1) Initially, we conduct an offline mapping, where RGB images are utilized to build semantic maps including 3D point cloud maps and 2D grid maps. (2) Subsequently, the user is allowed to select the desired object as the goal, and then the online navigation provides accurate positioning and optimal path planning while avoiding obstacles in real time. In addition, the navigation adopts an incremental strategy, enabling the user to easily reach goals step by step. Through extensive experiments, we evaluate the effectiveness of our system in mapping, object localization, and path planning, which demonstrates its comprehensive ability to enhance the independent mobility of the BVI. We release our implementation as open source software at https://github.com/Takerhythm/Indoor navigation for blind.