用于原位滑坡机理分析的抗扰动自标定相机群DIC方法研究

At present, the study of landslide mechanism mainly focuses on theoretical simulation and qualitative analysis, and in-situ experimental study lacks quantitative full-field characterization means. According to its practical measurement requirements, this project intends to explore a novel camera group digital image correlation method (DIC) based on self-calibration and atmospheric turbulence rejection, to achieve high precision deformation measurement and mechanism analysis of landslide at long distance. The main research contents include: set up the camera group system and study on images synchronous acquisition technology of uniform clock under different devices; It is developed an imaging constraint optimization method by single and full time domain scene matching for camera group self-calibration, and it is explored the error matching and the disturbance factor elimination during the calibration process; It is proposed the concept of local plane constraint optimization, and is studied the theory of camera group image points correspondence, theoretical model of error criterion is constructed, and it is developed the iteration process based on re-projection and reconstruction to suppress and eliminate the influence of atmospheric turbulence on imaging drift finally; Based on energy space-time evolution law, the macroscopic evolution mechanism of landslide is revealed with the in-situ full-field mechanical effects characterization. It solves the problem of difficult to calibrate and the imaging drift caused by atmospheric turbulence in the long distance deformation measurement, at the same time, the high-precision full-field in-situ deformation results provide the experimental and scientific basis for the study of the landslide mechanism and reinforcement theory, which is of important significance in both theory and practice.

目前滑坡机理研究多集中在理论计算与定性分析，而原位实验研究缺乏定量全场力学效应表征手段。本项目根据其实际测试需求，拟探索一种以自标定与抗空气扰动为基础的新型相机群数字图像相关方法（DIC），实现滑坡远距离高精度变形测量与机理分析。主要研究内容包括：搭建相机群系统并开展异备控制模式下时钟一致图像同步采集技术研究；提出通过单一及全时域场景对应的成像约束优化鲁棒自标定方法，探究标定过程中误匹配及扰动因素消除算法；提出局部平面约束优化的思想，研究相机群图像点对应理论，构建误差判据的理论模型，发展以重投影与重构建为基础的迭代过程，最终实现抑制及消除空气扰动对成像漂移的影响；根据能量时空演化规律，利用滑坡原位全场力学效应表征揭示其宏观演化机理。解决了远距离变形测量过程中无法标定和空气扰动的难题，同时高精度全场变形数据为滑坡机理乃至边坡加固理论的研究提供了实验基础和科学依据，具有重要的理论与实际意义。

针对滑坡机理的分析研究缺乏定量原位实验全场力学效应表征手段，本项目提出一种基于相机群的抗扰动自标定新型数字图像相关方法。主要研究内容包括：采用树莓派主机、高分辨率工业相机、低畸变镜头等加工并搭建了相机群硬件系统，编写了相关的软件，实现了远程相机控制、图像采集与无线传输等功能，基于网络授时方法实现了相机群优于1ms的时钟同步采集精度。从理论上推导了相机群系统在不同夹角情况下的三维测量误差公式，并且进行了计算机数值仿真模拟，其结果表明，相比于传统双目立体视觉系统，相机群测试系统的三维整体测量精度至少可以提高15%以上。进行了60°、80°和100°的热流扰动实验，随着温度的升高，引起的像素坐标误差越大，三维坐标重构的误差也越大。在热扰动环境下，相机群测试系统比双目立体视觉系统精度提高约15%-20%，证明了相机群测试系统的抗扰动特性。发展了迭代成像约束参数优化自标定模型和迭代坐标映射方法，实现了相机群系统自标定、标定板三维坐标优化以及非重叠阵列相机标定。针对滑坡机理实验研究，设计并加工了玻璃制边坡模型箱，搭建了1:1黑白砂边坡模型，分别进行了静态和线性加载等试验，基于数字图像相关方法对边坡位移进行了分析，结果表明竖向位移大致呈线性分布且随荷载的增加逐步增大。通过FLAC 3D的Mohr-Coulomb准则对边坡进行了差分元模拟分析。比较模拟和实验数据发现：数字图像相关方法可以准确预测边坡的变形状况和发展趋势，改变了现有滑坡演化机理的研究以理论计算和数值仿真为主，以定量的原位实验分析揭示滑坡宏观稳定及破坏的力学机制情况，实现了原位滑坡变形的可视化。