High-precision 3D models of buildings and industrial plants generated via contactless laser surveying are increasingly in demand. However, current data processing still requires numerous manual steps. Subjective decisions and interpretations by analysts lead to premature generalizations, errors and reduced accuracy. Advances in technology have led to a steady increase in scan resolution. The resulting data is too large to be analyzed by hand. Although restricting measurements to fewer locations reduces the effort required for analysis, it also decreases the level of detail of the final 3D model. Instead, a methodology is required for reducing manual labor and coping with the large data volume of high-resolution images.
Automated processing meets both of these goals. In contrast to manual analysis, computer algorithms guarantee the desired accuracy of the resulting 3D models and achieve the prescribed tolerance. Furthermore, a reduction of manual effort enables the processing of a much larger data volume and thereby permits more detailed modeling of large buildings and industrial plants. Once recorded, the extensive and high-precision data can be re-processed at will without the need for additional measurements.
The software scene·Capture offers a database driven solution for all stages of laser surveying. Methods for optimal site planning for laser scanners are particularly important for minimizing the time required for measurements and guaranteeing the desired accuracy. Such methods aim to reduce the number of laser scans, as each measurement corresponds to an increase in the data volume and processing effort. Nonetheless, sufficient laser scans must be performed in order to resolve occlusion and to compensate for the diminished sensor accuracy as distance to the scanner increases. A high-resolution digital camera delivers color information for subsequent texturing. Fixed points surveyed with a tachymeter additionally serve to link with an existing reference system such as national coordinate systems.
A typical scan has an accuracy of ≈3 mm (in 10 m distance), consists of 175 to 800 million points (depending on the resolution), and covers up to 300 m² of ground area.
Generalization of laser point clouds into high-precision 3D models is achieved in multiple automated steps by the scene·Model software. First, a structural analysis of the individual point clouds makes it possible to generate highly detailed triangular mesh structures. Their fusion yields the entire model with the desired accuracy. Algorithms for object reconstruction then automatically generalize the 3D model. This enables the export of objects, e.g. industrial components, in standard formats for use in CAD software.
All modules uphold prescribed tolerances and degree of detail of detected objects. Photos from the camera contribute to texturing an immersive virtual reality which is generated from the 3D model.
Several augmented-reality tools such as a virtual tape measure are available in the viewing program ≈InterView. Floor plans, cross sections, snapshots and videos can also be generated.