Contribution to the development of a traceable stereovision system for the 3D scanning of large and complex parts

This thesis is focused on the investigation and development of a 3D vision system coupled with a robot for contactless 3D scanning on mechanical parts in the aeronautic domain. Unlike probe contact-based coordinate measurement machines, 3D vision systems lead to scan parts in a short time. All these scanning operations, initially performed on simple parts, have recently been extended to complex surfaces while catching the new industrial needs in terms of in-line measurement automation, as specified in the industry of the future (or Industry 4.0). Therefore, the quality assurance of multi-camera scanning systems and their traceability to the SI meter definition represents a challenging objective. The full measurement system combining a camera-projector system and one industrial robot is dedicated to in-line 3D scanning on mechanical large volume parts with complex shapes. The system is designed, developed and assembled in-house in order to ensure a complete traceability chain of the measurement process.One 3D vision system based on the principle of structured light has been developed and calibrated in-house. The calibration of 3D vision systems is a crucial step prior to any 3D scanning operations. It allows to identify the requested internal, external and distortion parameters used later to collect a dense and accurate point-cloud of the mechanical part. In this context, calibration techniques of 3D vision systems have been studied and one novel optimisation method is proposed to improve the calibration accuracy. A synthetic and experimental evaluation was conducted to prove the efficiency of the optimization method where the convergence has been proven to be faster. The calibration of the developed 3D vision system is carried out with a traceable ceramic checkboard that has been measured by an optical CMM internally using MicroVu excel optical CMM machine.Finally, a large volume part quite similar to that of aeronautics with complex shapes is developed, measured by a traceable Zeiss UPMC carat CMM machine, and used for the evaluation of the developed 3D vision system. A local scanning strategy is adopted to cover the entire surface of the large volume part. It consists of independently scanning several areas of the part and then aligning the measurement in a single reference frame using validated registration techniques. To obtain a reliable and accurate measurement result, 3D data processing and fusion algorithms are studied. The result of large volume part measurement has shown a maximum fitting error of about 150 µm.

photogrammetry, 3D measurement, programming, image processing, C++ / Matlab, metrology