Visual Object Detection System for Autonomous Vehicles in Smart Factories

Abstract

Autonomous transport vehicles are very important for smart factories. Computer vision studies for autonomous vehicles in industrial environments are considerably less than that of outdoor applications. Recognition of safety signs has an important place in safe movement of vehicles and safety of humans in factories. In this study, we built a test environment for smart factories and collected a visual data set including some important safety signs for the safe and comfortable movement of the vehicles in smart factories. Then, we developed a visual object detection system using YOLOv3 deep learning model and tested it by using autonomous robots. In our tests, an accuracy of 76.14% mAP (mean average precision) score was obtained in the dataset we collected.