| Abstract |
People with visual impairment face various problems in doing daily activities in comparison to people without visual impairment. Much research has been done to find smart solutions using mobile devices to help them perform tasks like navigation and shopping. One of the most challenging tasks for researchers is to create a solution that offers a good quality of life for people with visual impairment. It is also essential to develop solutions that encourage them to participate in social life. The essential steps of a typical navigation system are identifying the current location, finding the shortest path to the destination, and navigating safely to the destination using navigation feedback. In this thesis, an overview is given about the various technologies that have been developed in recent years to assist people with visual impairment navigating indoors. It introduces the latest direction in this area, which will help developers to incorporate such solutions into their research. A comparison has been made between different technologies used
in developing solutions to select the best one from the available solutions. A system has been proposed to help people with visual impairment navigating indoor using markers. The system is able to detect and avoid obstacles during navigation when needed. The navigation system has been improved to detect markers from a longer distance using CNN model. The system has been improved using a deep learning model which is called Tiny-YOLOv3. Several modified versions of the original model have been implemented and compared to improve the detection accuracy.
A dataset has been created by collecting marker images from recorded videos and augmenting them using some techniques such as rotation transformation, brightness, and blur processing. After training and validating this model, the performance was tested on a testing dataset and real videos. |
