Design of Potholes Detection as Road’s Feasibility Data Information Using Convolutional Neural Network(CNN)
A Wiratmoko, Afif Zuhri Arfianto, Ari Wibawa Budi Santosa, Valian Yoga Pudya Ardhana, Albiyan Wanda Syauqi, Muhamad Surya Handika, Dinni Bangkit Nurrizki, Muhammad Wafi, Mat Syai’in, Imam Sutrisno, Muhammad Khoirul Hasin, Ii Munadhif · 2019
One indicator of road's feasibility can be seen from the conditions on the road surface. Road damage that often occurs is the number of potholes or holes [1]. In this study, we made an inspection device that can detect hollow roads automatically. Where the results of the detection will be sent to the server along with the coordinates of the location of the hole that will be processed and used as data on information on roadworthiness in an area. So when conducting road inspections we do not calculate the number of manually perforated roads so that the inspection process will run faster and more efficiently. This tool works with the camera as the main sensor, the camera works with the eye to provide visual data in the form of real-time video [2]. The visual data is processed using the Convolutional Neural Network algorithm which has 480 x 320 pixels with convoluted 2 times and produces smaller pixels, so we obtain weights to be classified with datasets [3]. These results will result in a decision whether there is a hole or not on the road that has been inspected by this tool. In addition to knowing where the location of the damage is this tool is equipped with GPS. If a hole is detected, the tool will take a picture, then send the image and coordinates to the server, so that the server can find information on the hole in an area [4]. This study obtained a success percentage of 92.8% in detecting holes using the CNN (Convolutional Neural Network) method.