Abstract:
Lack of Sight causes navigation very arduous for the Visually Impaired People. Usually they use white canes for obstacle detection. They have to physically touch the objects before recognition meanwhile they memorize the shapes of those objects. They need to take assistance from their family members. This is something very frustrating for visually impaired people that they have to depend upon others most of the time. In this world of technology and advancement, various vision systems exist that use different kind of navigation techniques to help these blind people. Some solutions require carrying heavy systems for computations while other use GPS technology which have unsatisfactory results in cloudy conditions and in the indoor environment. Research is still in progress to overcome the challenges of providing perfect accuracy in object identification. Our smart vision system comprises of stereo camera and portable Raspberry PI with architecture of Convolutional Neural Architecture (CNN). First, we acquire stereo images and then we apply CNN algorithm to identify objects in each frame. Then we apply the distance calculation algorithm to find distances of each identified object from the user. At the end, an audio feed is given to the blind person via headphone. In this way, a blind can identify the desired object without touching it.
