高速立体视频测量大型建筑物模型振动台试验健康监测关键技术研究

Non-contact, high-frequency intensive measurement and three-dimensional measurement are the three advantages of high-speed stereo videogrammetry, which can solve the disadvantages of the traditional transducers, such as single point, one-dimensional measurement, easy to damage and lower monitoring frequency. For the high-speed stereo videogrammetry for shaking table test of large structure models aiming to health monitoring, massive image sequences processing, high frequency noise processing and structure models health analysis are studied in this project, including: (1) Fast, accurate identification model of elliptic targets for massive image sequences based on elliptic characteristics and space gray level matrix sub-pixel location considering the accuracy and efficiency, and the target tracking and matching model based on digital image correlation and correlation matching technique, (2) Source and propagation law of high-frequency noise and the corresponding influence mechanism, three stage high frequency noise reduction model based on the multi-point Savitzky and Golay filter and the corresponding quality evaluation index, and (3) Intrinsic mode decomposition model and criterion of input signal based on the introduction of innovative ideas of ESMD, and the damage identification model and damage assessment index of structure models by time-frequency analysis. This project will provide a set of relatively systematic theory and methods for the massive image sequences processing and the structure models health analysis for high-speed stereo videogrammetry for shaking table test of large structure models.

高速立体视频测量具有非接触、高频密集测量和三维量测的优点，可解决传统传感器单点、一维性、易破坏和监测频率低等缺点。本项目主要针对高速立体视频测量大型建筑物模型振动台试验健康监测的海量数据处理、高频噪声处理和建筑物模型健康分析展开研究，包括：（1）顾及精度和效率的基于椭圆特征和空间灰度矩亚像素定位的海量影像序列椭圆形目标点快速、准确识别模型，基于数字图像相关和相关匹配法的目标点跟踪和匹配模型；（2）高频噪声的来源和传播规律及其影响机制，基于多点Savitzky和Golay滤波的三级高频噪声降噪模型和质量评价指标；（3）创新性引入ESMD思想的输入信号的固有模态分解准则和模型，时频分析建筑物模型损伤识别模型和损伤程度评估指标。通过本项目的研究，有望为高速立体视频测量大型建筑物模型振动台试验健康监测提供一套较为系统的海量影像序列处理与建筑物模型健康分析的理论与方法。

为了保证建筑物的安全性，通常需要在建/构筑物建造之前按照模型相似关系制作缩尺建筑物模型，并将试验模型安置在振动台上行激振试验，获取建/构筑物模型关键位置的动态响应时程曲线，分析评价建/构筑物模型的抗震性能，从而采取合理的抗震措施和改进方法以提高建构筑物的抗震能力。本项目采用高速视频立体测量进行建筑物模型振动台试验健康，针对高速立体视频测量海量数据处理、高频噪声影响和建/构筑物模型健康分析展开研究，包括：（1）构建了顾及精度和效率的基于椭圆特征和空间灰度矩亚像素定位的海量影像序列椭圆形目标点快速、准确识别模型，基于数字图像相关和相关匹配法的目标点跟踪和匹配模型，可有效的提高目标点的识别和跟踪效率；（2）分析高频噪声的来源和传播规律及其影响机制，构建基于多点Savitzky和Golay滤波的三级高频噪声降噪模型和质量评价指标，有效的降低了高频噪声的影响，提高了数据解析精度；（3）创新性引入ESMD思想的输入响应信号的模态分解模型和准则，时频分析建/构筑物模型损伤识别模型和损伤程度评估指标，不仅可以分析建筑物模型的稳定性，而且可以分析建筑模型荷载的条件。高速立体视频测量具有非接触、高频密集测量和三维量测的优点，将其应用于土木工程领域的建/构筑物模型振动台试验健康监测，可解决传统接触式传感器单点、一维性和易破坏等缺点，以及其他非接触传感器价格昂贵和监测频率低等缺点，为建/构筑物模型振动台试验健康监测提供一种新的监测方法和测试手段，获取建/构筑物模型振动台试验过程的三维动态响应过程，为分析其破坏机理和提高抗震性提供数据支持。因此，本项目的研究对推动高速立体视频测量建/构筑物模型振动台试验健康监测应用的深入发展，具有重要的现实意义和应用价值。