YOLO Algorithm-Based Surrounding Object Identification on Autonomous Electric Vehicle

Abstract

An autonomous vehicle must be equipped with a camera, which works by providing visual input that is used to detect objects around the autonomous electric vehicle. Currently, no method has been implemented in real-time. Thus, this study utilized the You Only Look Once (YOLO) algorithm to detect objects in real-time around the autonomous electric vehicle. The objects were limited to humans, motorcycles, and cars. The results showed that the most compatible YOLO model for the system was the Tiny YOLOv4 model which was built with the darknet framework. The simulation experiment showed that detection accuracy was 80% and was able to transmit information in a form of data location of the object to the microcontroller. A success rate of 100% was obtained from 10 tests. Hence, it showed that the YOLO was able to detect objects and provided input to the steering control system. Meanwhile, the depth information method was used to measure the distance of the object to the vehicle in real-time with an accuracy of 60%. Real-time testing was conducted to test whether the autonomous electric vehicle can avoid objects in front of it by providing input from the detection results of the Tiny- YOLOv4 model object. The success rate of the system in real-time experiments was 100%.