Solving the speed problem in vlab "visual localization aid for the blinds

Abstract

VLAB is a system designed to help the sight impaired people to determine the position within indoor env· · · · · ironment using image processing. This system uses a mobile camera that captures images from the environment, analyze it and give a voice message for the blind person. The system has three phases: simulation, solving the speed problem and deploying the system onto a mobile device. The first phase was a simulation for the system to test if the system can compete with the other non-image processing system, the results for this phase was accurate but there was the speed problem, the system gives a result in 73 seconds, that means the system can not work in the real-time unless the speed is increased. The second phase, our project, aims to solve the speed problem without affecting the accuracy of the system, the speed problem arises from the sequential comparison between the features in the query image and the database; the proposed solution is using a KD-Tree to index the database. This has decreased the time of comparisons to less than 0.082 second per query and the accuracy was maintained