Spatial information system for the collection, documentation and analysis of industrial archaeological objects (RIO, Geometry sub-project)
The data fusion of image and scan data is used to link visual and geometric information and is intended to create an improved interpretation and modeling basis for 3D objects. The aim is to develop a method in which the recording positions of the camera and laser scanner do not have to be identical. This allows increased flexibility in object detection to be implemented.
Due to the complementary properties of image and scan data, both digital photogrammetry and terrestrial laser scanning are increasingly being used simultaneously for capturing and documenting three-dimensional objects. Their cooperative use requires both the creation of correspondence between the different, heterogeneous data types and the linking of similar data types from different recording points of view.
The procedure developed uses specially designed passport objects to create better conditions for establishing correspondence. The focus is on linking the scan data. Cubes are used, whose corner points are used as control points for determining the transformation parameters. The cube is broken down into its individual elements (levels, edges, vertices) and the cube vertices are extracted using a RANSAC algorithm and the adjustment and intersection of the cube levels based on a topological model. Due to its uniqueness, this allows the determination of the spatial alignment of the cube.
The accuracy of the calculated cube geometry depends primarily on the shooting distance. For a distance of <10m, maximum length deviations of the room and surface diagonals of <4mm are obtained.
The calculated transformation parameters when linking the scan data are comparable with the results of existing methods for registering scan data. Depending on the recording configuration, the residual gaps are up to 5mm. In contrast to existing methods, good approximation values for the transformation parameters are already obtained with just one cube.
At the same time, the determination of the cube corner points from the scan data creates good conditions for a link with the image data, which can be recorded independently of the position of the scanner standpoint.