基于双目视觉的结构试件位移角点检测方法研究

发布时间：2019-03-28 10:29

【摘要】：通过振动台实验,我们可以研究结构试件的振动位移等信息。通过分析位移信息得到其抗震性能。检测结构试件的特征点作为物体测试的特定点,其位移、速度、加速度、挠度等运动参数是实验所要得到的数据。而本文的重点在于检测特征点的位移变化,通过对位移变化曲线的分析,来判断结构试件是否发生了变形。本文主要研究了视觉技术应用于结构位移检测的现状以及发展趋势。了解了振动台试验的位移检测的基本理论,主要包括数字图像的成像原理及其表现形式与特征、图像用于测量工作时所涉及的坐标系、图像预处理的相关算法、目标点的定位以及相机标定算法。提出了基于双目立体视觉技术的振动台试验的结构试件位移测量的算法,并对该算法的关键技术做了具体的研究。其核心内容包括,1)本文采用基于圆形阵列组成的标定板与张氏标定算法对系统进行标定。2)通过基于彩色空间转换与Hough变换检测直线,降低了直线漏检率。再根据Harris角点检测算法得到图像的特征点,通过控制阈值来控制角点的数目和质量。3)本文研究了特征点匹配的约束条件以及sift特征描述子,采用基于极线约束的双向匹配算法进行实验匹配,得到较好的结果。研究了双目视觉测量位移的原理。并将其应用于结构试件位移检测中。本文重点研究了双目视觉系统的标定、特征点的检测及匹配。通过设计静态试验验证了本文算法的可行性,再将此算法应用到振动台的动态试验中,得到结构试件的动态位移曲线。针对本文提出的测量方法,设计了相应的动、静态两种实验方案,并在我校的振动台实验室中进行了实验。首先通过静态试验来验证本文算法测量的可行性以及测量精度。然后进行了正弦波激励实验,并通过本文算法计算结构试件特征点的位移,得到的位移曲线基本与正弦激励信号变化趋势是一致的。
[Abstract]:Through the shaking table experiment, we can study the vibration displacement and other information of the structural specimen. The seismic performance is obtained by analyzing the displacement information. The movement parameters such as displacement, velocity, acceleration, deflection and so on are the data to be obtained by testing the characteristic points of the structural specimens as the special points of the object test. The emphasis of this paper is to detect the displacement change of the characteristic points and to judge the deformation of the structural specimen by analyzing the curve of the displacement change. This paper mainly studies the present situation and development trend of the application of visual technology in structural displacement detection. In this paper, the basic theory of displacement detection in shaking table test is understood, including the imaging principle of digital image, the form and characteristics of digital image, the coordinate system involved in the measurement work of the image, and the related algorithm of image pre-processing. Target location and camera calibration algorithm. An algorithm based on binocular stereo vision for measuring the displacement of structural specimen in shaking table test is presented and the key technology of the algorithm is studied in detail. The core contents of this paper are as follows: 1) the calibration algorithm based on circular array is used to calibrate the system. 2) through color space conversion and Hough transform to detect straight line, the detection rate of straight line is reduced. 2) the calibration board based on circular array and Zhang's calibration algorithm are used to calibrate the system. Then the feature points of the image are obtained according to the Harris corner detection algorithm, and the number and quality of the corners are controlled by the control threshold. 3) in this paper, the constraint conditions of feature point matching and the sift feature descriptor are studied. Two-way matching algorithm based on polar constraint is used for experimental matching, and good results are obtained. The principle of measuring displacement with binocular vision is studied. It is applied to the displacement detection of structural specimens. This paper focuses on the calibration of binocular vision system, the detection and matching of feature points. The feasibility of the proposed algorithm is verified by the static test. Then the dynamic displacement curve of the structural specimen is obtained by applying the algorithm to the dynamic test of the vibration table. In view of the measurement method proposed in this paper, the corresponding dynamic and static experimental schemes are designed, and the experiments are carried out in the shaking table laboratory of our school. First of all, the feasibility and accuracy of this algorithm are verified by static test. Then the sinusoidal excitation experiment is carried out and the displacement of the characteristic points of the structural specimen is calculated by this algorithm. The displacement curve obtained is basically consistent with the changing trend of the sinusoidal excitation signal.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TB534.2

【参考文献】