Multi-Sensor Deformation Measurement System

The increasing number of objects involved in deformation processes in highly populated areas has increased the demand for rapidly working and easily usable deformation measurement systems. This demand is associated with a higher degree of automation of monitoring, analysis and interpretation. For monitoring of an object involved in a deformation process the object and its surrounding has to be modeled, which means dissecting the continuum by discrete points in such a way that the points sufficiently characterize the object, and that the movements of the points represent the movements and distortions of the object. The framework targeted in this project uses two types of sensor systems and/or a combination of them to address the deformation problem: image assisted theodolites (videotheodolites) and raster laser scanners. In a complex measurement system, the selection of a suitable sensor or suitable measurement modes is a highly non-trivial task. To provide user decision support, information about the object respectively the scene has to be collected in an automated way. For this task we suggest cognitive vision techniques. A process which sets up a description of the movements and distortions of the object can be coupled to deformation monitoring, followed by an assessment of the deformation. This process must be outlined in a framework of local-to-global information integration, by grouping locally measured deformation into a more informative deformation pattern. Especially when automatic monitoring is in procedure, the deformation classification can lead to important decisions. Such a measurement system consists of different components: the sensors, a system control component, a system for cognitive vision, a system for deformation analysis and assessment, and the knowledge base. The first phase of the project will contain an extensive analysis of the problem and the knowledge necessary for developing solutions. In the next phase a suitable model for the knowledge-based and cognitive vision system will be investigated. Tests, experiments and evaluations while the whole project phase shall demonstrate the advantages of a deformation measurement system which is supported by knowledge-based and cognitive vision techniques.

In this research project a new kind of image-based measurement system is under development. This system is based on new techniques (originally developed in the area of Artificial Intelligence) which shall be used for the tasks of deformation measurement, -analysis and -interpretation. On the sensor side the system is based on a combination of the `Image Assisted Total Station` (IATS) prototype and a Terrestrial Laser Scanner (TLS). Such a system provides an immense number of 3D data, both from the IATS system and from the laser scanner. This point cloud may be reduced by filtering, even if not very effective. Our approach builds on cognitive vision techniques. These methods can be used as well for finding regions of interest as for point filtering. The system first generates a scene overview on the basis of one or more images taken by the IATS (by the wide- angle camera or by a georeferenced image-mosaic) or laser scanner data. In a next step, a scene/object description is generated by an image understanding tool, using three new developed methods (HMF algorithm, window detection based on Adaboost and laserscanner data). On the basis of this description regions of interest (ROI, smallest measurement area in the current framework) are selected. This subdividing of the surface enables a local-to-global information integration strategy (allows a detection of changes in the inner and outer geometry of the surface). After defining the ROIs an automatic point detection is done by using so-called interest operators. Subsequent the detected points are measured by the IATS automatically. For a deformation investigation of an object these steps have to be done repeatedly. To ensure the measurement of the same points in each epoch a matching algorithm is used. The next step in the procedure is the deformation assessment, a combination of a classical deformation analysis and a new developed deformation characterization. Building on the deformation patterns an interpretation of the deformation will be done.