Keywords: image processing, correlation, displacement measurement, block motion estimation, adaptive rood pattern search, particle swarm optimization.

Concrete structures are used all over the world. These structures have normally a long life but it is necessary to study how materials change over time, the importance of the reinforcement materials and the aging of the material submitting it to cyclic stress tests. Civil engineering laboratory experiments traditionally use electrical gauges to measure beams displacements. However, these sensors require a complex setup and calibration procedure with risks for their integrity when the beams break. As a result of their cost and installation difficulties, only a limited number of points are usually measured limiting the analysis to local displacements.

Using a simple consumer level camera and a computer it is possible to take photos during and after the loading of the concrete beams at regular time intervals that document the bending of the structure. After collecting the time series images we can detect the displacements between consecutive images using a mathematical correlation algorithm and extract the information about the material under study.

In this paper three different image processing algorithms are used for the measurement of displacement measurements in civil engineering load tests. To test the algorithms two experiments were carried out using different beam materials: concrete and Plexiglass. Two photographic sequences were taken during the load tests on a concrete beam and on a Plexiglass bar. The images were subjected to three different algorithms: simple and efficient search (SES), adaptive rood pattern search (ARPS) and the rood pattern particle swarm optimization (RP-PSO). The data obtained from these algorithms was compared with the data from LVDT sensors. As is shown in the paper, the image processing measurements have an accuracy at least equivalent to the LVDT sensors but with much less equipment and setup requirements.

All the algorithms have shown acceptable accuracy results when compared with the LVDT sensors and with the ground truth obtained from three different trained users. The new algorithm proposed (RP-PSO), a mix between ARPS and PSO, shows less dependency on the block size used on the image analysis and a higher dynamics that facilitates the tracking of the blocks when abrupt displacements occurs. The inconvenience of RP-PSO is a slightly higher computational time that is compensated by the best results achieved. The efficiency of all the algorithms is dependent on several factors such as the speckle pattern, the block size, the search parameters, the image resolution and the time between images. However, as we show on this study, it is not difficult to find a good compromise for these values that allows good results.

The main advantages of the image processing techniques are the easy to setup and inexpensive equipment required and the production of a displacement map that shows how the material reacts to the stress that is imposed during the experiment. As a result of the unpredictable behaviour of the beam structure, the large area covered by the images also guarantees the analyses of the all the beam that is extremely difficult or even impossible to do with traditional sensors. Moreover, the image documentation also allows the detailed analysis of small areas of interest with greater detail even after the experiment is complete.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £50 +P&P)