Wearable visual impairment assistance system
Loading...
Date
2017-06
Authors
Wong, Hong Loon
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In year 2014, the World Health Organization (WHO) reported that there are 285 million people are estimated to be visually impaired worldwide, include low vision and blind. About 90% of this group of people are from low income family. Visually impaired person encounter a lot of challenges during their daily life, such as navigation and recognize object. When carry out daily activities, they just can rely on other sensory information to perceive the environment which is quite inconvenient for them. Todate, many engineers and designers have designed a lot of assistance system for visually impaired person to improve their quality of life. However, some of the systems are uncomfortable, medium accuracy or too expensive for the low income blind people. In this study, a low cost wearable assistance system for visually impaired people is developed to help them detect obstacle during navigation. The proposed system implements the concept of stereo vision to achieve the purpose of obstacle detection. A stereo-camera system, which is fabricated from two single cameras is used to capture a stream of stereo image pairs of the environment in real time manner. Stereo matching will be performed on stereo image pairs to generate disparity map where the obstacle is spotted and extracted from. An obstacle detection algorithm is developed by using disparity map and U-disparity map. Raspberry Pi 3 is used as controller to operate stereo camera and implement obstacle detection algorithm. A wearable assistance system prototype is designed and built for blind users to wear at their waist level. The wearable assistance system able to inform blind users about the obstacle condition in front of them through vibration. The frame refreshing rate of the developed system is approximately one second and the system cost approximately RM600. The whole system and obstacle detection algorithm is tested in indoor and outdoor for static and dynamic environment. Experimental results show that the obstacle detection algorithm has more than 70% of detection rate in indoor dynamic environment and more than 90% of detection rate in indoor static environment with false alarm rate of around 20%. However, although the algorithm has high detection rate in outdoor environment, it also has very high false alarm rate, hence, this algorithm need further improvement for outdoor use.