测绘通报 ›› 2021, Vol. 0 ›› Issue (3): 28-32,86.doi: 10.13474/j.cnki.11-2246.2021.0073

• 学术研究 • 上一篇    下一篇

利用Faster R-CNN进行立交桥自动识别与定位

马京振1, 陈换新2, 朱新铭1, 张付兵1   

  1. 1. 信息工程大学, 河南 郑州 450001;
    2. 96911部队, 北京 100011
  • 收稿日期:2020-05-28 修回日期:2021-01-09 发布日期:2021-04-02
  • 作者简介:马京振(1993—),男,博士生,研究方向为数字地图制图技术。E-mail:zb50mjz@163.com
  • 基金资助:
    国家自然科学基金(41571399;41771487;41901397)

Automatic recognition and positioning of overpass based on Faster R-CNN

MA Jingzhen1, CHEN Huanxin2, ZHU Xinming1, ZHANG Fubing1   

  1. 1. Information Engineering University, Zhengzhou 450001, China;
    2. Troops 96911, Beijing 100011, China
  • Received:2020-05-28 Revised:2021-01-09 Published:2021-04-02

摘要: 立交桥结构的自动识别对道路网多尺度建模、空间分析和车辆导航具有重要意义。传统基于矢量数据的立交桥识别方法过分依赖人工设计的特征,对复杂场景的适应性较差。本文提出了一种基于目标检测Faster R-CNN神经网络模型的立交桥识别方法,该方法利用卷积神经网络学习立交桥样本的深层次结构特征,进而实现立交桥的自动识别与准确定位。试验结果表明,该方法对立交桥的识别效果较好,能够在复杂的道路网中准确确定立交桥的位置,避免了人为干预对试验结果不确定性的影响,抗干扰性较强。

关键词: 立交桥, 目标检测, Faster R-CNN, 深度神经网络, 模式识别

Abstract: The automatic recognition of overpass structures is of great significance for multi-scale modeling, spatial analysis and vehicle navigation of road network. The traditional method of overpasses recognition based on vector data relies too much on the characteristics of manual design and has poor adaptability to complex scenes. In this paper, a method for overpasses identification based on the target detection model Faster R-CNN is proposed. This method uses convolutional neural network to learn the deep structural characteristics of samples, and then realizes the automatic recognition and accurate positioning of the overpasses. The experimental results show that the method has a good recognition effect on overpasses, and can accurately determine their positions in the complex road network, avoiding the influence of human intervention on the uncertainty of results, and has a strong anti-interference.

Key words: overpass, target detection, Faster R-CNN, deep neural network, pattern recognition

中图分类号: