Scholarlink

<p>To improve the performance of recognizing surrounding objects in autonomous driving, deep learning-based object detection models such as You Only Look Once (YOLO) have been widely used recently. In this YOLO model, bounding boxes are displayed on objects to provide various information such as object type, confidence score, distance and angle to the object. However, overlapping bounding boxes can reduce the accuracy of distance estimation to objects. Therefore, in this paper, we propose a new object detection model that includes a bounding box separation method. To achieve this, coordinate alignment process is performed for each of the light detection and ranging (LiDAR) and camera sensors. Through this sensor fusion, even if the two sensors are installed in different locations, the LiDAR points can be simultaneously displayed on the camera image in real time. An experiment was conducted by installing a sensor fusion system based on robot operating system 2 (ROS2) communication in an actual passenger vehicle. Finally, it was verified that the proposed YOLO model with the bounding box separation method precisely derives both distance and angle to objects around the experimental vehicle in a low-speed driving environment.</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

Development of YOLO-Based Object Detection Model With Bounding Box Separation Techniques

  • 期刊论文

  • 发表日期：

    2025年8月28日

查看全文
  • Hyeongmuk Kang,
  • Giseo Park

摘要

To improve the performance of recognizing surrounding objects in autonomous driving, deep learning-based object detection models such as You Only Look Once (YOLO) have been widely used recently. In this YOLO model, bounding boxes are displayed on objects to provide various information such as object type, confidence score, distance and angle to the object. However, overlapping bounding boxes can reduce the accuracy of distance estimation to objects. Therefore, in this paper, we propose a new object detection model that includes a bounding box separation method. To achieve this, coordinate alignment process is performed for each of the light detection and ranging (LiDAR) and camera sensors. Through this sensor fusion, even if the two sensors are installed in different locations, the LiDAR points can be simultaneously displayed on the camera image in real time. An experiment was conducted by installing a sensor fusion system based on robot operating system 2 (ROS2) communication in an actual passenger vehicle. Finally, it was verified that the proposed YOLO model with the bounding box separation method precisely derives both distance and angle to objects around the experimental vehicle in a low-speed driving environment.