Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Session: 3
Christopher Feltner - University of Central Florida
Co-Author(s): Jonathan Guilbe, University of Central Florida, Orlando, FL ; Sharare Zehtabian, University of Central Florida, Orlando, FL ; Siavash Khodadadeh, University of Central Florida, Orlando, FL ; Ladislau B?l?ni, University of Central Florida, Orlando, FL ; Damla Turgut, University of Central Florida, Orlando, FL
Navigation of environments is a complex challenge for individuals who are visually impaired, and these individuals are typically aided by tools such as canes or guide dogs. Individuals who are both visually and mobility impaired encounter greater difficulty, since conventional aids do not integrate well with walkers. Many previous smart walker solutions designed to aid the visually impaired use expensive laser ranging devices, the cost of which is a barrier to the adoption of such walkers. RGB-D cameras have the potential to aid such individuals at a significantly lower cost.
We have developed a smart walker for the visually impaired which can convey information to the user about their proximity to obstacles using haptic feedback. We have done so using the relatively affordable Microsoft Kinect RGB-D camera as an image and depth sensor. We have implemented two algorithms for identifying the distance to the nearest obstacle, the first using depth images, and the second using point clouds. In the first, an array of distance values is generated from the depth image which represents the depth along the user?s expected path of travel. Sudden increases or decreases in the difference between adjacent values indicate the presence of an obstacle. In the second, we segment the floor plane from a point cloud of the environment and identify the closest remaining point in the cloud.
The walker was evaluated on its ability to detect four types of obstacles walker users are likely to encounter on a daily basis, along with obstacle-free pathways. Each algorithm was evaluated individually. The actual distance from the obstacles to the walker was measured using a tape measure and was compared to the distances reported by each algorithm. Using either algorithm, the walker was able to identify the distances from obstacles to within 10 centimeters of the actual distance. It was also able to identify a path free from obstacles, and a potentially dangerous drop.
Possible further research on this topic includes developing a method of identifying obstacles in sunlit environments and improving detection of downward drops.
References:
G. J. Lacey and D. Rodriguez-Losada, ‘The Evolution of Guido,’ in IEEE Robotics & Automation Magazine, vol. 15, no. 4, pp. 75-83, Dec. 2008.
doi: 10.1109/MRA.2008.929924
Huy-Hieu Pham, Thi-Lan Le, and Nicolas Vuillerme, ‘Real-Time Obstacle Detection System in Indoor Environment for the Visually Impaired Using Microsoft Kinect Sensor,’ Journal of Sensors, vol. 2016, Article ID 3754918, 13 pages, 2016. https://doi.org/10.1155/2016/3754918.
Ortigosa, N., Morillas, S. & Peris-Fajarn’s, G. J Intell Robot Syst (2011) 63: 115. https://doi.org/10.1007/s10846-010-9498-4
Panteleris P., Argyros A.A. (2015) Vision-Based SLAM and Moving Objects Tracking for the Perceptual Support of a Smart Walker Platform. In: Agapito L., Bronstein M., Rother C. (eds) Computer Vision – ECCV 2014 Workshops. ECCV 2014. Lecture Notes in Computer Science, vol 8927. Springer, Cham
Funder Acknowledgement(s): The support for this work was provided by the National Science Foundation REU program under Award No. 1560302. Any opinions, findings, and conclusions and recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
Faculty Advisor: Prof. Damla Turgut, turgut@cs.ucf.edu
Role: I contributed to the development of the project goals, the literature review for the project, and design of the walker's electronics. My main contribution was the development of software which can recognize obstacles using point clouds. I also contributed in part to the development of the software to recognize obstacles in depth images.