摘要
针对交通标志牌识别效率低、效果差等问题,该文以点云数据为研究对象,提出一种基于几何特征的道路交通标志牌识别方法。该方法根据交通标志牌中标志杆为杆状、牌面为面状的几何特征,利用拟合筛选柱状物的方法识别标志杆,结合强度差异构建基准面进行聚类的方式识别牌面。该文对两段不同路况数据进行实验分析,结果表明该方法不仅能够完整地提取多种类型的交通标志牌,而且能够较好地区分与其他地物邻近的交通标志牌,两次实验的平均精度以及平均召回率都达到了95%左右,提取结果为道路交通标志牌维护等工程应用提供可靠的数据。
Aiming at the problems of low efficiency and poor effect of traffic sign recognition,a method of road traffic sign recognition based on geometric features was proposed in this paper,taking point cloud data as the research object.According to the geometric characteristics of the sign pole and the face of the sign board,the fitting and screening method were used to identify the sign pole,combining the strength difference to construct a datum plane for clustering to identify the face of the sign board.Two sections of different road condition data were tested and analyzed.The results showed that this method could not only completely extract multiple types of traffic signs,but also better distinguish traffic signs adjacent to other features.The average precision and average recall rate of the two experiments had reached about 95%.The extraction results provided reliable data for road traffic sign maintenance and other engineering applications.
作者
黄明
车平文
韦朋成
HUANG Ming;CHE Pingwen;WEI Pengcheng(School of Mapping and Urban Spatial Information,Beijing University of Civil Engineering and Architecture,Beijing 102616,China;Engineering Research Center of Typical Architecture and Ancient Architecture Database,Beijing 102616,China;School of Geodesy and Geomatics,Wuhan University,Wuhan 430079,China)
出处
《测绘科学》
CSCD
北大核心
2023年第2期115-123,共9页
Science of Surveying and Mapping
基金
国家自然科学基金项目(42171416)
北京建筑大学基础研究基金项目(x20088)。
关键词
点云
交通标志牌
柱状物拟合
基准面
point cloud
traffic signs
column fitting
datum plane
