In combination with standard transducers and data acquisition systems, computer vision can be adopted in order to perform the analysis of the behaviour of structures during dynamic tests such as earthquake simulations on shake tables. The paper describes the design and implementation of a machine vision system aimed at providing bi-dimensional position measurement of reflective markers directly placed on test specimens. The developed solution is composed of a scalable set of acquisition units, each consisting of a high definition digital camera and a personal computer. A sequence of images is acquired by the cameras and the position of the markers in the scene is estimated by means of a software application running on the computers. Each unit can perform measurements in a single plane which is defined in a previous calibration phase. The method has many advantages over the most commonly used acquisition devices such as accelerometers and potentiometers: first, the absence of contact between the acquisition device and the tested structure, which allows the non-invasive deployment of an arbitrary number of measurement targets, which is even more important in destructive tests, for preventing the loss of expensive transducers; second, the direct calculation of the position of an object in length units, without the need of post processing like integration and conversion, as required when using accelerometers in shake table tests. Besides, in the selected plane, thanks to the adoption of infrared illumination and filters to reduce environmental lighting interferences, each unit can follow the movements of a large number of markers (up to 50 for each camera in the performed tests) with a precision of around 0.05 mm. On the other hand, the method is by itself unable to overcome problems deriving from the three-dimensional movement of the selected markers. The paper also explains the different approaches and the corresponding results obtained while solving this issue
Computer Vision System for Monitoring in Dynamic Structural Testing
Lunghi, Francesco;Pavese, Alberto;Peloso, Simone;Lanese, Igor;
2012-01-01
Abstract
In combination with standard transducers and data acquisition systems, computer vision can be adopted in order to perform the analysis of the behaviour of structures during dynamic tests such as earthquake simulations on shake tables. The paper describes the design and implementation of a machine vision system aimed at providing bi-dimensional position measurement of reflective markers directly placed on test specimens. The developed solution is composed of a scalable set of acquisition units, each consisting of a high definition digital camera and a personal computer. A sequence of images is acquired by the cameras and the position of the markers in the scene is estimated by means of a software application running on the computers. Each unit can perform measurements in a single plane which is defined in a previous calibration phase. The method has many advantages over the most commonly used acquisition devices such as accelerometers and potentiometers: first, the absence of contact between the acquisition device and the tested structure, which allows the non-invasive deployment of an arbitrary number of measurement targets, which is even more important in destructive tests, for preventing the loss of expensive transducers; second, the direct calculation of the position of an object in length units, without the need of post processing like integration and conversion, as required when using accelerometers in shake table tests. Besides, in the selected plane, thanks to the adoption of infrared illumination and filters to reduce environmental lighting interferences, each unit can follow the movements of a large number of markers (up to 50 for each camera in the performed tests) with a precision of around 0.05 mm. On the other hand, the method is by itself unable to overcome problems deriving from the three-dimensional movement of the selected markers. The paper also explains the different approaches and the corresponding results obtained while solving this issueI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
