Real-Time Detection of Road Objects and Lane Markings for Autonomous Vehicles

This research focuses on the development of an advanced object and lane recognition system for autonomous vehicles, leveraging the YOLOv8 Deep Learning (DL) model. The primary goal is to ensure accurate, real-time recognition of road objects, lane boundaries, and other critical features to enhance autonomous navigation. The methodology involves preprocessing and augmentation of a diverse dataset, followed by training and testing the YOLOv8 model. This approach ensures robustness under varying environmental and traffic conditions, including occlusions and lighting challenges. The final model is integrated into an intuitive user interface that provides real-time visual feedback, enabling seamless interaction and monitoring. Comprehensive evaluations were conducted to validate the system’s performance. The results demonstrated high accuracy in recognizing objects like vehicles, and pedestrians, as well as stable lane recognition across diverse scenarios. Despite minor limitations in extreme conditions, the system proved to be reliable and efficient for real-world applications. Future work will aim to expand functionality by incorporating additional sensors like LiDAR and optimizing the system for edge devices to reduce computational overhead. Additional features, such as traffic sign recognition and obstacle prediction, will also be explored. The proposed system serves as a robust foundation for enhancing Autonomous Vehicle (AV) technology, contributing significantly to road safety and intelligent transportation systems. Through innovative integration of machine learning techniques, this research addresses key challenges in autonomous navigation and lays the groundwork for further advancements in the field. This research addresses inaccurate object detection and unstable lane recognition in autonomous vehicles. The proposed system improves accuracy and reliability for safer self-driving technology.