Scholarlink

<p>Autonomous Vehicle System (AVS) are reshaping the future of transportation by leveraging computer vision, deep learning, and advanced sensor technologies to achieve intelligent, self-reliant mobility. This review presents a comprehensive analysis of key vision-based components essential to AVS, with a focus on pedestrian detection, vehicle recognition, and traffic light and signal interpretation—stages that directly impact road safety and decision-making. The study aims to examine the current research landscape, highlighting technological methodologies, datasets, performance metrics, and the role of deep learning models. It also explores trends in AV development led by major automotive and tech industries. A detailed comparison of feature-based, deep learning-based, and optimization-based approaches is conducted for each detection stage. The role of simulation environments and annotated datasets in training robust AV perception systems is also discussed. Furthermore, the paper reflects on ethical, regulatory, and deployment challenges associated with AVS in complex urban environments. It emphasizes the need for adaptable, context-aware models capable of performing reliably in diverse and unpredictable conditions. While advancements continue, realizing full autonomy demands collaborative innovation across domains such as AI, embedded systems, and human–machine interaction.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Vision technologies in autonomous vehicles: progress, methodologies, and key challenges

  • 期刊论文

  • 发表日期：

    2025年8月5日

查看全文
  • Hemant Kumar,
  • Pushpa Mamoria,
  • Deepak Kumar Dewangan

摘要

Autonomous Vehicle System (AVS) are reshaping the future of transportation by leveraging computer vision, deep learning, and advanced sensor technologies to achieve intelligent, self-reliant mobility. This review presents a comprehensive analysis of key vision-based components essential to AVS, with a focus on pedestrian detection, vehicle recognition, and traffic light and signal interpretation—stages that directly impact road safety and decision-making. The study aims to examine the current research landscape, highlighting technological methodologies, datasets, performance metrics, and the role of deep learning models. It also explores trends in AV development led by major automotive and tech industries. A detailed comparison of feature-based, deep learning-based, and optimization-based approaches is conducted for each detection stage. The role of simulation environments and annotated datasets in training robust AV perception systems is also discussed. Furthermore, the paper reflects on ethical, regulatory, and deployment challenges associated with AVS in complex urban environments. It emphasizes the need for adaptable, context-aware models capable of performing reliably in diverse and unpredictable conditions. While advancements continue, realizing full autonomy demands collaborative innovation across domains such as AI, embedded systems, and human–machine interaction.