High-speed videogrammetry has gradually become the important measurement technique in civil engineering material testing and structure monitoring. This paper presents a robust videogrammetric approach to monitor and analyse the spatial deformations of structures in large-scale shaking table tests, which consists of three main parts as follows. (1) A videogrammetry system with multiple high-speed cameras is established to measure three-dimensional (3D) morphological changes of large structures. (2) An accurate target recognition algorithm is introduced into the measurement scheme, and then the target tracking and matching strategy is proposed in our scheme to calculate the sequential image coordinates of the target points in the high-speed image sequences. (3) The 3D coordinates of the target points can be obtained by the videogrammetric analysis algorithms, and the key structural deformation parameters can be further calculated through the spatiotemporal analysis of the sequential point coordinates. The shaking table test of largescale wooden pagoda as an empirical test is performed to demonstrate the feasibility and reliability of the whole highspeed videogrammetric technique. The experimental results show that the proposed approach can achieve sub-millimeter positioning accuracy of the artificial targets with comparison of high-accuracy total station. Moreover, the credibility of the measured displacement results is further verified by comparison with the results of the high-accuracy displacement sensor.
In recent years, high-speed videogrammetry has been used to monitor the three-dimension behavior of large vibrating structures. Traditionally, this has been accomplished by transferring image sequences, captured by several cameras, to a central computer after each test has been conducted. Only then have the images been postprocessed and the required kinematic data extracted. This process is slow and inefficient because a large amount of data (image sequences) must be transferred to the host before the data analysis can begin. In order to address this problem, we have developed a novel system that adopts a distributed data processing strategy. The system which combining the processing power built into each of the individual cameras and the processing power of a central computer, uses a wired local area network. the communication between the nodes and the host is achieved using the TCP/IP protocol and a custom application layer. In this way, the processing power of the entire system is more fully utilized and the overall performance of the video processing system is improved. We describe how we have employed two cameras, operating simultaneously, to test the proposed concept. The experimental results from a series of tests showed that the average time required to perform the necessary image processing was reduced 58.7% by using the distributed processing system instead of a traditionally configured system.