Designing a Smartphone-Based Assistance System for Blind People to Recognize Intersections and Obstacles in Indoor Corridors

People with visual impairment face significant difficulties when navigating in indoor corridors. As an indoor corridor may contain obstacles (e.g., boxes, chairs), blind people need to be aware of them to avoid accidents. In addition, they also face another challenge when they have to turn in an intersection. Detecting an intersection may be challenging for blind people, as they may not notice its existence even if they are in it. For example, when an obstacle is in front of an intersection, blind people may walk past it as they cannot walk along the wall. Therefore, We aim to navigate blind people safely in an indoor corridor using only one smartphone by assisting them to avoid obstacles and detect intersections. To detect intersections in indoor corridors, Garcia et al.proposed to use a quadcopter to capture RGB images and perform the detection using a convolution neural network. As photos taken by blind people may contain motion