Fusion of laser-scan and image data for deformation monitoring – Concept and perspective

Abstract

Areal measurements, like laser scans or camera images, are more and more frequently used for deformation monitoring.Each single acquisition method has its advantages and disadvantages in detecting displacements, however.In dense point clouds of laser scan, distance changes in line of sight are easy to identify.In contrast, high-resolution image data is sensitive to displacements perpendicular to the viewing direction of the camera, where the analysis of scan data weakens.A promising solution of combining the advantages and reducing the drawbacks is the fusion of laser-scan and image data in the form of RGB+D images.Point clouds are converted into a depth image, resulting in an additional D-channel to the colour (RGB) image data.In the combined RGB+D image, each pixel can be directly converted into 3D coordinates.The necessary mutual orientation of both data sets can be solved via a-priori sensor calibration or a-posteriori data registration.By identifying corresponding points in subsequent measurement epochs (RGB+D images), it is further possible to directly determine 3D displacement vectors.The results of the RGB+D method can be integrated in a rigorous geodetic deformation analysis with tests of significance.Corresponding points in the images can be found by identifying key points and by describing them with image features, for example.The prominent representatives of image features are the SIFT algorithm, which uses floating point values, or the binary descriptor BRISK.These algorithms are used to match two images at an abstract numerical level, instead of in the original image domain.In this analysing strategy, sensor data of any kind of acquisition system can be used, such as modern total station/multistations, mobile mapping systems, unmanned aerial vehicles or robot platforms.It will exploit the full potential of sensor data of such systems, which already provide both kind of data, for the first time.