Hazard Detection for the Visually Impaired
Student
Jason Moericke
Advisor
Paul Schrater
Abstract
Many people think of vision in a few broad categories:
- good vision
- vision corrected with an artificial lens (glasses/contacts)
- color blind
- total blindness
However, there are many different conditions that cause a wide variety of visual impairments, such as blurred vision that cannot be corrected with an artificial lens, total loss of peripheral vision, or total loss of central vision, to name a few.
The goal of this project is to prove concepts for creating assistive technology for these individuals. The goal is to inform the person when there is a hazard in their path. Future work would suggest an action to take to avoid the hazard.
The scope of this project is to create an Object Detection model to identify uneven sidewalk and other hazards, use Dense Optical Flow to determine the current heading, and identify when the person's current heading would collide with a hazard. Uneven sidewalk, for example, present trip hazards, which could result in injury for someone with impaired vision.