Scholarlink

<p>This paper proposes a novel two-step method combining classical and quantum computation to accurately detect crosswalk traffic lights for visually impaired pedestrians. The first step utilizes the improved YOLOv5 model to identify crosswalk traffic light locations. In the second step, the detected lights are processed using a single-layer quantum convolutional neural network with a newly designed parameterized quantum circuit. The proposed quantum circuit reduces the number of qubits and circuit depth required while maintaining high accuracy, making it suitable for noisy intermediate-scale quantum devices. Experimental results demonstrate that our method has better performance compared to existing methods, achieving detection accuracy of 94.9%, precision of 95.4%, and the AUC reaches 0.98. These results also demonstrate the model’s robustness in complex traffic environments.</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

Crosswalk traffic light detection for the visually impaired based on hybrid classical–quantum neural networks

  • 期刊论文

  • 发表日期：

    2025年6月19日

查看全文
  • Jinghan Zhang,
  • Zhengan Tian

摘要

This paper proposes a novel two-step method combining classical and quantum computation to accurately detect crosswalk traffic lights for visually impaired pedestrians. The first step utilizes the improved YOLOv5 model to identify crosswalk traffic light locations. In the second step, the detected lights are processed using a single-layer quantum convolutional neural network with a newly designed parameterized quantum circuit. The proposed quantum circuit reduces the number of qubits and circuit depth required while maintaining high accuracy, making it suitable for noisy intermediate-scale quantum devices. Experimental results demonstrate that our method has better performance compared to existing methods, achieving detection accuracy of 94.9%, precision of 95.4%, and the AUC reaches 0.98. These results also demonstrate the model’s robustness in complex traffic environments.