Page 92 - 无损检测2024年第六期

缺陷的智能识别与分类专题
缺陷的智能识别与分类专题

DOI：10.11973/wsjc202406010

基于点云数据实时配准算法的桥梁结构复杂

缺陷形态识别

李沛东
（平顶山市公路交通勘察设计院，平顶山 467000）

摘 要：桥梁结构损伤检测过程中，大多依托于二维图像完成结构复杂缺陷的形态识别，只考
虑表面的灰度和纹理等信息，使得识别结果的平均精度均值（m AP ）较低。因此，提出基于点云数据
实时配准算法的桥梁结构复杂缺陷形态识别方法。首先采用激光扫描仪和传感器设备，采集桥梁
结构三维点云数据，并运用基于密度的聚类算法进行点云数据聚类分割，实现点云数据去噪处理；
然后利用柔性动作-评价（SAC）配准算法、改进迭代最近点（ICP）配准算法进行点云数据实时配准，
考虑深度、高度等第三维属性完成复杂缺陷区域检测；最后针对缺陷区域的测量点分别计算结构位
移函数，基于此识别出缺陷的具体形态。试验结果表明，所提方法得出的复杂缺陷形态识别结果的
m AP 值大于0. 92，基本满足了桥梁检测要求。
关键词：三维点云数据；点云配准；去噪处理；形变特征
中图分类号：TH133.3；TG115.28 文献标志码：A 文章编号：1000-6656(2024)06-0054-06

Complex defect morphology identification of bridge structures based on point cloud data real
time registration algorithm

LI Peidong
（Pingdingshan Highway Transportation Institute of Survey and Design, Pingdingshan 467000, China）

Abstract: In the process of bridge structural damage detection, most of the recognition of complex defect morphology
relies on two-dimensional images, only considering information such as surface grayscale and texture, resulting in low
Mean Average Precision(m ) of the recognition results. Therefore, a real-time registration algorithm based on point cloud
AP
data was proposed for the recognition of complex defect morphology in bridge structures. Firstly, laser scanners and sensor
equipment were applied to collect three-dimensional point cloud data of bridge structures, and density-based clustering
algorithms was used for point cloud data clustering and segmentation to achieve point cloud data denoising processing.
The Soft Assignment Cost (SAC) registration algorithm and improved Iterative Closest Point (ICP) registration algorithm
were used for real-time registration of point cloud data, complex defect area detection was completed by considering
three-dimensional attributes such as depth and height. Finally, the structural displacement function separately for the
measurement points in the defect area was calculated, and the specific defect morphology was identified based on this. The
experimental results showed that the m value obtained from the application of the proposed method for complex defect
AP
morphology recognition was higher than 0.92, which basically met the requirements of bridge detection.
Key words: 3D point cloud data; point cloud registration; noise reduction processing; deformation characteristic

在现代综合交通体系中，桥梁是不可或缺的一
收稿日期：2023-09-28
项基础设施 。在长期运营后，受到基础不均匀沉降、
[1]
作者简介：李沛东（1982—），男，本科，高级工程师，主要从事道
外部荷载的共同影响，桥梁结构会出现复杂缺陷，这
路、桥梁工程的检测工作
些缺陷会直接影响桥梁的承载能力，甚至威胁桥梁
通信作者：李沛东，dangpangyijc@163. com
54
2024 年 第 46 卷 第 6 期
无损检测