Object Detection and Distance Estimation Tool for Blind People Using Convolutional Methods with Stereovision
Rais Bastomi, Annas Singgih Setiyoko, Budi Herijono, Efrita Arfah Zuliari, Mardlijah Mardlijah, Firza Putra Ariatama, Lucke Yuansyah Arif Tryas Putri, Septian Wahyu Saputra, Mohammad Rizki Maulana, Mat Syai’in, Ii Munadhif, Agus Khumaidi, Mohammad Basuki Rahmat · 2019
In this research, a tool that can provide information about object around is made. This tool can also estimate distance of detected object through camera which is combined with glasses, to ease blind people who use it. This tool is certainly can help them to identify object around and improve their skill and ability. This tool use camera as main sensor, which works like human eyes, to provide real time video as visual data. The RGB visual data is processed using Convolutional Neural Network which has 176 × 132 pixels by convoluting 2 times. It produces smaller pixels with size 41 × 33 pixels, so weights is obtained for classification using back propagation and determined dataset. After getting detection result, the next step is a find centroid value as center point for measuring the distance between objects and cameras with Stereo Vision The results is converted into sound form and connected to earphones, so blind people can hear the information. The test results show that this tool can detect predetermined objects, namely humans, tables, chairs, cars, bicycles and motorbikes with an average accuracy of 93.33%. For measurements of distances between 50 cm to 300 cm it has an error of around 6.1%.