图 文 导 读

该文亮点：

1.发现了可变形三维网格模型控制点坐标和结构姿态参数的关联性和映射关系；

2.提出了基于可变形三维网格和运动光流表示学习的结构稠密位移识别方法；

3.分析了所提方法对结构稠密位移识别的空间分辨率和准确性;

4.验证了所提方法对复杂背景和光照变化等环境干扰的鲁棒性。

Highlights:

1. The correlation and mapping relationship between control point coordinates of deformable 3D mesh models and structural pose parameters are found.

2. A structural dense displacement recognition method is proposed based on deformable 3D mesh model and motion representation learning of optical flow using monocular video.

3. The spatial resolution and accuracy of the proposed method for structural dense displacement recognition is analyzed.

4. The robustness of the proposed method to environmental disturbances, such as complex background and illumination changes, is verified.

位移是表征机械系统和土木结构健康状态的重要变量。传统接触式位移感知技术和视觉特征跟踪技术存在稀疏测点和固定靶点的测量限制，高空间密度的大型结构稠密位移准确识别仍面临挑战。采用可变形三维网格模型作为结构振动视频和稠密位移之间的载体，将结构稠密位移识别问题转化为控制点坐标光流表示学习和结构姿态参数识别问题，是一种新颖思路，可在低计算成本条件下保证位移识别的稠密性。

哈尔滨工业大学李惠教授团队在《国际机械系统动力学学报（英文）》（International Journal of Mechanical System Dynamics, IJMSD）发表题为“基于可变形三维网格和运动光流表示学习的结构稠密位移识别方法”的研究论文。该文提出了基于可变形三维网格模型和运动光流表示学习的结构稠密位移识别网络Nodes2STRNet，其中，控制点估计子网络NodesEstimate以结构振动视频帧为输入，通过稠密光流计算可变形三维网格模型的控制点坐标；姿态参数识别子网络Nodes2PoseNet以控制点坐标为输入，通过多层感知机自监督学习，获得结构姿态参数识别结果并转换为结构稠密位移。此外，通过四层钢筋混凝土框架结构振动台试验，验证了所提结构稠密位移识别方法的高空间分辨率、准确性及对复杂背景和光照变化等环境干扰的鲁棒性。该研究为基于单目视频的大型结构稠密位移识别提供了新方法，可为结构动力特性辨识提供高空间分辨率的基础数据。

Abstract：Displacement is a critical indicator for mechanical systems and civil structures. Conventional vision-based displacement recognition methods mainly focus on the sparse identification of limited measurement points, and the motion representation of an entire structure is very challenging. This study proposes a novel Nodes2STRNet for structural dense displacement recognition using a handful of structural control nodes based on a deformable structural three-dimensional mesh model, which consists of control node estimation subnetwork (NodesEstimate) and pose parameter recognition subnetwork (Nodes2PoseNet). NodesEstimate calculates the dense optical flow field based on FlowNet 2.0 and generates structural control node coordinates. Nodes2PoseNet uses structural control node coordinates as input and regresses structural pose parameters by a multilayer perceptron. A self-supervised learning strategy is designed with a mean square error loss and L2 regularization to train Nodes2PoseNet. The effectiveness and accuracy of dense displacement recognition and robustness to light condition variations are validated by seismic shaking table tests of a four-story-building model. Comparative studies with image-segmentation-based Structure-PoseNet show that the proposed Nodes2STRNet can achieve higher accuracy and better robustness against light condition variations. In addition, NodesEstimate does not require retraining when faced with new scenarios, and Nodes2PoseNet has high self-supervised training efficiency with only a few control nodes instead of fully supervised pixel-level segmentation.

Keywords:

structural dense displacement recognition, deformable structural mesh model, deep-learning-based monocular vision, self-supervised learning

DOI: 10.1002/msd2.12083

Share this article:

https://onlinelibrary.wiley.com/doi/10.1002/msd2.12083（点击链接直达原文）‍

Cite this article: Zhao J, Li H, Xu Y. Nodes2STRNet for structural dense displacement recognition by deformable mesh model and motion representation. IntJ Mech Syst Dyn. 2023; 3(3):229‐250.

ResearchGate: 点击  直达该文ResearchGate页面

作 者 简 介

赵  今    博士毕业于哈尔滨工业大学，发表论文5篇，主要从事基于计算机视觉的结构位移识别方法研究。

李  惠    哈尔滨工业大学教授。国家自然科学基金创新群体负责人、国家杰青、长江学者特聘教授，曾任国际结构控制与监测学会理事长、任亚太智能结构技术研究中心网络主席。主持国家自然科学基金创新群体项目、重大项目课题、重点项目、联合基金项目、国家重点研发计划项目等国家级项目20项。在Science和Nature等期刊发表论文200余篇，出版专著3部，授权发明专利60余项，主编了交通行业公路桥梁结构监测技术规范，成果成功应用于百余座大型桥梁。获国家科技进步二等奖4项、美国土木工程师协会George W Housner奖章和Robert H Scanlan奖章等。

徐  阳    哈尔滨工业大学副教授。入选博新计划，承担国家自然基金青年基金等国家级和省部级项目10项。发表论文20篇，授权发明专利10项。主要从事结构损伤识别计算机视觉方法研究。