测绘通报 ›› 2023, Vol. 0 ›› Issue (11): 173-176.doi: 10.13474/j.cnki.11-2246.2023.0348

• 测绘地理信息应用案例 • 上一篇    下一篇

手持三维激光扫描室内停车场地面标识要素提取方法

刘云彤1,2,3, 黄金亭1, 王家耀2,3   

  1. 1. 黄河水利职业技术学院, 河南 开封 475004;
    2. 河南大学地理与环境学院, 河南 开封 475004;
    3. 河南大学时空大数据产业研究院, 河南 郑州 450046
  • 收稿日期:2023-07-10 出版日期:2023-11-25 发布日期:2023-12-07
  • 作者简介:刘云彤(1977—),男,博士生,副教授,主要研究方向为遥感与地理信息技术。E-mail:yuntongliu@163.com

Method for extracting indoor parking lot ground marking elements based on handheld laser scanning point cloud

LIU Yuntong1,2,3, HUANG Jinting1, WANG Jiayao2,3   

  1. 1. Yellow River Conservancy Technical Institute, Kaifeng 475004, China;
    2. The College of geography and environment, Henan University, Kaifeng 475004, China;
    3. Henan Industrial Technology Academy of Spatio-temportal Big Data, Zhengzhou 450046, China
  • Received:2023-07-10 Online:2023-11-25 Published:2023-12-07

摘要: 为了满足停车场快速建图的需求,本文提出了基于手持激光点云的室内停车场地面标识要素的提取方法。首先,为减少要素提取对内存空间的需求,将整个点云以规则网格进行划分;其次采用RANSAC平面拟合的方法提取每个网格内的地面点云;然后为提取地面标识要素,根据地面点云生成地面图像,并在地面图像的基础上,采用BiSeNet网络对不同的标识要素进行语义分割,得到车道线、车位线和导向箭头标识的像素;最后针对车道线和车位线,采用基于霍夫变换的直线提取方法对其进行提取,对于地面导线箭头,采用模板匹配的方法对其进行提取。试验证明,本文提出的方法能够对扫描的结构要素和标识要素进行快速提取,可大大减少制图的人工工作量,有效提高室内停车场建图的效率。

关键词: 手持三维激光扫描, 室内停车场, 地面标识要素, 三维点云, 深度学习

Abstract: In order to meet the demand for rapid mapping of indoor parking lots, a method is proposed for extracting ground marking features in indoor parking lots based on handheld LiDAR point clouds. Firstly, to reduce the memory space requirements for feature extraction, the entire point cloud is divided into regular grids. Then, the RANSAC plane fitting method is used to extract the ground point cloud within each grid. In order to extract ground marking features, a ground image is generated based on the ground point cloud. On this basis, the BiSeNet network is employed for semantic segmentation of different marking features, such as lane lines, parking lines, and directional arrow markings, to obtain the corresponding pixels. For lane lines and parking lines, a line extraction method based on Hough transform is used, while for ground arrow markings, a template matching method is applied for extraction. Experimental results demonstrate that the proposed method can quickly extract both structural elements and marking features from scanned data, significantly reducing manual mapping workload and improving the efficiency of indoor parking lot mapping.

Key words: handheld 3D laser scanning, parking garage, ground marking elements, 3D point cloud, deep learning

中图分类号: