A Nighttime Vehicle Detection Method Based on YOLO v8
Weichuan An, Jiageng Ruan, Tongyang Li, Jing Xia
Abstract
With the continuous development of autonomous driving technology, how to accurately and efficiently identify the target in some challenge scenarios, e.g. nighttime, has become an important task. Although, in view of cost, pure visual object detection technology has become popular with growing camera performance and image processing capabilities, the pure visual object detection technology based on cameras still faces enormous challenges in accurately identifying targets in environments with poor visibility, such as dusk, night, or rainy or foggy days. This article introduces the state-of art image processing technology, i.e. YOLOv8, to improve target detection capability in the challenge environment. The vehicle recognition accuracy of YOLOv8 is evaluated in low light environments, such as dusk and night. The results show that's the recognition and detection accuracy of YOLOv8 at night surpasses previous versions of the series, making it an important tool for improving autonomous driving safety in poor lighting environments.