Deep Learning in Autonomous Vehicles: A Comprehensive Review of Object Detection, Lane Detection and Scene Perception

With the advancements in transportation today, deep learning (DL) plays an important role in enhancing thinking and decision-making capabilities. This paper presents a comprehensive review of state-of-the-art deep learning for object detection, lane detection, and autonomous vehicle scene detection. First, we explore the evolution of deep learning-driven object detection methods, focusing on the use of convolutional neural networks (CNNs) and multimodal sensor fusion techniques that provide lidar data, cameras, and radar. The paper continues on line detection problems, focusing on traditional methods and recent innovations such as entropy-based fusion models and deep CNNs. Special attention is paid to overcoming the disadvantages of the environment that hinder the search process, such as poor lighting and occlusions. The role of deep learning in this sense is also investigated, interpreting driving difficulties by driving itself. By analyzing these components (object detection, line detection, and situational awareness), this analysis provides insight into the current capabilities, limitations, and future directions of deep learning for autonomous driving, paving the way for improved autonomous driving efficiency and autonomous driving safety.