基于图像矫正与特征融合的结构振动特性监测

Monitoring of Structural Vibration Characteristics Based on Image Correction and Feature Fusion

针对机器视觉测量在实际监测环境中存在的相机标定难、测点追踪稳定性差等问题,建立了一套基于机器视觉的结构模态参数识别方法。提出一种以归一化正交投影模型简化相机成像模型的简化标定方法,以多点局部标定代替整体标定,降低标定难度;提出一种基于图像矫正与特征融合的特征点匹配方法,先对测点进行变形矫正,然后采用梯度信息和位置信息加权平均值作为特征点匹配依据,提高特征点匹配精度。通过一个3层框架振动试验验证方法的精度和有效性。结果表明,机器视觉方法所提取的位移和识别的模态参数与真实值基本吻合,且该方法在不同拍摄角度下仍然具有较高精度和稳定性。该方法在超高层建筑、长大跨桥梁等远距离动态位移监测及模态参数提取中具有较好应用前景。

In order to deal with the difficulties of camera calibration in field measurement,a structural modal parameters identification method is established based on machine vision measurement.A normalized orthogonal projection model is proposed to simplify the imaging model.In addition,the multi-point local calibration instead of the overall calibration is applied to lower the difficulty of cameral calibration.A method is proposed to improve the stability of feature points matching,i.e.,the distortion of the imaging of artificial target is rectified,and the weighted average of the gradient information and position information is used as the basis for feature points matching.The accuracy and effectiveness of the proposed method is verified by a vibration test of a three-floor frame structure.The results show that the displacement and modal parameters obtained by the machine vision method are basically consistent with the real values.In addition,the proposed method still has high accuracy and stability under different shooting angles.This method is promising in the dynamic displacement monitoring of super high-rise buildings and long-span bridges.