Holistic multi-camera deflectometry (MultiDeflect)

点击翻译按钮获取中文摘要

Geometric-optical metrology enables fast, robust, non-contact and accurate 3D shape measurement. Hence, its application is preferential in quality control of many manufacturing applications. Deflectometry is particularly suited as a measurement technique for specular surfaces such as varnished car bodies and optical components. Deflectometry uses specific patterns on a display area which are reflected on the object surface and observed by an imaging system as such as a CCD-camera. Phase Measuring Deflectometry (PMD) uses sinusoidal fringe patterns to yield a unique coding by applying phase shifting. The geometry of the surface under test is assessed by identifying surface positions and local slopes using ray tracing. From this approach some fundamental properties of deflectometry arise: i) Position and local slope are ambiguous. Uniqueness is achieved by introducing prior knowledge or additional sensors. ii) The sensitivity with respect to the local slope is significantly higher compared to the position. Shape acquisition is gained by integrating over the slopes. This leads to particularly high demands on the system calibration since calibration errors will propagate through this process. The project aims at making best use of the huge amount of sensor information and prior knowledge as well as various methods to solve ambiguities in order to improve the system calibration and to lower the measurement uncertainty. This is approached by holistic treatment of all system parameters analogous to the Bundle Adjustment used for auto-calibration of cameras. In the literature, only rudimentary studies on investigating the improvement of the accuracy during calibration and measurement by introducing additional independent information in a holistic approach have been presented yet. Cameras were mainly treated in the photogrammetric model although the model-free Vision Ray Calibration is capable of providing additional information to the holistic approach. The proposed project therefore envisages the assembly of a PMD setup with access to 4 cameras and a distance sensor for the determination of a reference point. The following sub goals must be archived: i) Development and qualification of holistic optimization procedures considering 1 to 4 cameras; ii) solving the ambiguity by combining the multi camera technique and the knowledge of a reference point; iii) improved optimization by additional application of the vision ray model to the cameras. The procedures to be developed here are applicable to existing PMD systems to reduce their measurement uncertainty and the time required for measurement and calibration due to a more effective usage of available sensor information. Furthermore, confinements and possibilities for improving the accuracy by additional sensors are demonstrated.