A Customized Weather Adaptive Blind Crutch Based on Target Recognition Research Using Deep Learning

Abstract

Object detection problem is one of the important topics in the field of computer vision. This technology can be especially used to help blind people identify the objects with details. In these cases, the preciseness of detection of objects is extremely important for blind people, since no other references can be made for them. In this project, we create a blind crutch that can not only alarm users about presenting obstacles, but also identify objects with clear labels using the convolutional neural network (CNN), a neural network especially designed for image identification, to train the classifications needed. Among the process of creating the detection function, in order to improve the accuracy of real scene detections, the snow and rain scenes preprocessing of the image data is completed by the algorithms of dark channel prior and the generative adversarial network. To make up the insufficiency of the GPS module of detecting exact locations, we also create a function of detecting guideposts on streets and broadcast the current location for the blind people. After testing, the blind crutch can precisely classify objects and complete the broadcast function under different environments caused by different weathers. This electronic document is a “live” template and already defines the components of your paper [title, text, heads, etc.] in its style sheet.