| With the rapid development of manufacturing technology,the processing of components is moving towards miniaturization,integration and precision,has been widely applied in integrated circuit,microelectromechanism systems,and many other fields.The rapid and precise acquisition of microdevice morphology and defect detection is of great significance for the development of these industries.The traditional 1D signal or 2D images based detection approaches can no longer meet the requirements of modern industrial measurement.The importance of 3D microscale measurement technology,which can rapidly and accurately perform quantitative analysis and characterization of the3 D structure of microdevices,is becoming increasingly prominent.Microscopic Fringe Projection Profilometry(MFPP)approaches are very suitable for high-precision online detection of 3D morphology of microdevices due to its significant advantages such as fullfield measurement,non-contact,simple system structure,fast measurement speed,high reliability,and insensitivity to environmental vibration and background light.This thesis concentrates on the research of the problems in the existing MFPP approaches,such as complex structure,difficult calibration,and insufficient measurement accuracy.It explores the integrated system calibration algorithm and improves the related measurement theory to achieve automated high-precision 3D micro-measurements.Also,it simplifies the measurement device structures to the flexibility of 3D morphology detection,making it meet the requirements of automated measurement and promote its application in largescale industrial automation.In summary,this dissertation mainly focus on expanding the industrial online detection application tasks of structured light 3D measurement,exploring methods to simplify system structure,system calibration procedure,and improve measurement accuracy in the microscopic scene.The main work of this thesis includes:Firstly,we design and set up a simplified and practical MFPP system consists of a pinhole Scheimpflug projector and bi-telecentric camera,and develop a set of accurate system calibration algorithms.Usually,the final calibration parameters of cameras are obtained through nonlinear optimization.However,because of the instability of the optimization algorithms and the lack of appropriate constraints,the optimization result often deviates from the actual situations.To address these issues,this thesis proposes a reliable optimization approach used for bi-telecentric camera calibration.The approach is built on the closed-form solution of the bi-telecentric camera.It uses a bundle adjustment optimization approach with two constraints to finely optimize all the parameters of the bi-telecentric camera,therefore can achieve more accurate results.For the projector in the MFPP system,we view it as a reverse pinhole camera.By using phase labels on image rows and columns,a mapping relationship between the bi-telecentric camera and the“image” of the pinhole Scheimpflug projector is established.Then,the feature points of the projector “image” are succesfully extracted for calibrating the Scheimpflug projector.To eliminate the distortion of the projector,and obtain accurate calibration results,this thesis designs an iterative optimization strategy for generating the inverse distortion fringe pattern based on the Scheimpflug distortion model and initial calibration parameters.By projecting these optimized inverse distortion phase-shifting fringe patterns,the distortion of the project is compensated,thereby improving the calibration accuracy of the pinhole Scheimpflug projector.Secondly,we propose an integrated calibration approach for MFPP systems that use telecentric cameras.Existing approaches for calibrating the MFPP systems with telecentric cameras typically uses a system calibration approach based on phase height mapping(PHM).The method typically requires the calibration of a bi-telecentric camera with a translation stage first,and then the phase-height relationship calibration is achieved with the assistance of the translation stage and reference planes.The entire calibration process is cumbersome and requires additional equipment such as the tranlation stage and reference plane.Additionally,the calibration of the camera and the mapping of the phase-height relationship are separated,reducing the flexibility of the system calibration.In this dissertation,by fully utilizing the pose parameters of the calibration target obtained by the pinhole projector calibration,the determination of the sign of the third component of the rotation matrix of the bi-telecentric camera is achieved,eliminating the dependence of the bi-telecentric camera calibration on the translation stage.Furthermore,based on the complete pose of the target obtained from the projector calibration and the parameters of the bi-telecentric camera,the 3D model of all pose targets is calculated for phase-height mapping coefficient calibration,eliminating the dependence of the phase-height calibration on the translation stage and reference plane.This method also solves the problem of the separation of the calibration processes for the bi-telecentric camera and the phase-height mapping,achieving flexible and efficient calibration for the MFPP system integration.Experimental results show that the accuracy of this method is comparable to the current highest precision PHM method,but this method is more flexible,efficient and easy to operate,with great potential for industrial applications.Thirdly,we propose a flexible and efficient 3D micro-measurement algorithm based on projector ray tracing.In existing FPP algorithms,geometric calibration based on stereo vision is flexible but not suitable for MFPP systems that use telecentric lenses.The PHM method,which uses a precise translation stage,has the highest accuracy but is cumbersome and inflexible.Most existing flexible phase 3D mapping methods are based on the pinhole model and not suitable for MFPP systems that use telecentric lenses(affine model).To achieve a flexible and efficient phase 3D mapping method for MFPP systems that use telecentric lenses,this dissertation fully utilizes the advantages of existing phase 3D mapping methods,the PHM method,and geometric calibration methods,and rigorously derives the accurate phase 3D relationship based on the ray tracing principle of the constructed MFPP system.The complete pose parameters of the target are obtained by using a pinhole Scheimpflug projector calibration,and the 3D model of the target is reconstructed using the parameters of the bi-telecentric camera.On this basis,the coefficients for the direct mapping of phase to 3D coordinates are calculated based on the coupled optical geometric structure and phase label decoding light propagation process.In addition,to remove phase artifacts and improve the accuracy of the projector calibration and the quality of the dot target phase,an algorithm is designed to optimize the phase of the dot target.The experiments show that the proposed method not only achieves higher 3D measurement accuracy than the PHM method but is also more efficient. |
