Automatic Object Detection to assist Blind Persons

Abstract

The aim of this article is to present a prototype of assistance to blind people. It uses the devices of a standard mobile phone and its design minimizes the computational requirement. The work presented here is the result of a series of activities covering video collection, pre-processing to reduce streaming, image segmentation and processing with Data Mining. The derived models are used in the detection of objects, profiling of the behavior of the individual, detection and
evaluation of movements in the environment and the interpretation of relevant events in real time in order to activate alarms specifically designed to assist the blind. Likewise, the prototype that implements through these models the techniques of image processing, detection of areas and classification of purified objects is introduced in order to infer if there is any type of risk in the
environment. This device is intended to assist blind people by determining the nature of objects
and circumstances in the environment, in order to warn in advance of relevant eventualities that deserve special attention. The scope of work includes background of the sector, first findings, variables and characteristics of the derived prototype, peculiarities detected in the interaction with blind people and the overall functionality of the device, as well as a small analysis of the results