On the Utilization of SLAM for Obstacle Detection in Commodity Mobile Devices

Abstract

Immersive Technologies are an increasingly prevalent field, employed by a plethora of portable and stationary solutions. Ongoing research continues to unlock new possibilities for their use, improving Human-Machine Interaction. In the health sector, such technologies have the potential to optimize the living of individuals with special needs, like the visually impaired. SLAM is a technique used in robotics and computer vision for achieving environmental understanding by building a vector map of the frontal space. It is the fundamental for developing applications that achieve immersive experiences to the user, such as Augmented Reality applications. The ability of these applications to understand their surroundings and the diverse methodologies studied over the years has led to the proposal of efficient techniques for extracting depth data from a camera feed, without the need of a depth sensor. This work introduces a system that exploits the Depth Estimation capabilities of SLAM to detect obstacles and cliffs in the user’s frontal environment. A mobile application was developed, that retrieves the camera feed and generates scene Depth-Maps, before importing them to a newly designed algorithm for obstacle and cliff identification. Audio and haptic feedback is used for warnings and usability notifications. The system was fine-tuned and tested in both indoor and outdoor spaces and quantitative and qualitative results were captured. The goal of the study is to present the development of a tool that can be executed on commodity mobile devices in real-time and it can enhance safety facilitating movement in both indoor and outdoor environments.