Real-Time Obstacle Detection for People with Visual Disabilities on Mobile Phone Using Optical Flow and Delaunay Triangulation

HitAlert is a collision detection system for people with visual disabilities, designed to be run on smartphone devices. This allows the system to be affordable and provides the user with society assimilation. Using video processing techniques, the system will alert its user to possible obstacle collision threats that may lead to physical injuries. The system incorporates Delaunay triangulation to create a representation of obstacles and Sparse Optical Flow algorithm to track obstacles across time frames. These cost-efficient algorithms enable HitAlert to analyze incoming obstacles in real-time. Users could choose to receive an alert via sound, vibration, or HitAlert Band, a phone mounting peripheral belt that provides multi-channel, multi-level haptic feedback for precise notification. From our real scenario testing, HitAlert could detect incoming obstacles with a recall of 84%.