Digital image-stitching techniques applied to dynamic measurement of large structures

Abstract

In this work has been discussed a propose of using three-dimensional point-tracking measuring technique for measurement of large structures in which the entire field of interest could not be captured by a unique stereo-vision system. Two pairs of a stereo-system are used to capture the whole field of measurement of interest of the model and point cloud registration techniques are exploited aiming at extending the capability of digital image measurement system for dynamic measuring (displacement) for large-scale structures. Three different image registration algorithms, principle component analysis, singular value decomposition, and iterative closest point are used in the stitching process to join the point clouds obtained with the multi-camera system. The proposal is applied to vibration measurement of a wind turbine blade of 2.3 m in length, whose field of view of the whole set of points of interest is greater than the field of view of a unique stereo-vision system. The reconstruction of the set of measured points was obtained from the junction of the points clouds of each stereo-system in a reference system and a probabilistic and statistical analysis of the error generated by the transformation of the point clouds was performed. And finally, the frequencies of the structure obtained from the digital image measured data were compared with the values obtained from a set of accelerometers.