| As one of the key components of Aero-engine power source,the three-dimensional(3D)measurement of Aero-engine blade profile is one of the important aspects of verifying the machining quality and ensuring the overall performance of Aero-engine.How to quickly and flexibly scan the 3D profiles of blade has always been a frontier problem in the field of aviation industry.Taking the fringe projection profilometry with the monocular vision as theories foundation,the dissertation further carries out systematically several key issues,such as the principle modeling of the measurement system,the calibration of system parameters,absolute phase recovery of fringe images and invalid phase values removal.Main research and achievements are as follows:A simplified and flexible phase-to-depth conversion model(PDCM)is presented for the monocular system of fringe projection profilometry of blade measurement.The detailed procedure of the PDCM is deduced in 2D and 3D geometric principles.Combined with the pinhole camera for area scan CCD with the first two terms of radial distortion coefficient,a 3D coordinate computing frame of space points is established in the camera coordinate system.At last,two kinds of structure situations and numerical simulation results of the PDCM in space geometry layout are given.Experiments have been performed to test the feasibility and flexibility of the proposed PDCM.A 2D checkerboard-based and phase-based measurement system parameters calibrating and optimizing system is proposed.The presented system covers four sub-systems: 1)camera intrinsic parameters calibration,2)camera extrinsic parameters calibration,3)depth parameters calibration and 4)a nonlinear global refinement.Firstly,the template of orthogonal vanishing points is modified through the checkerboard pattern and a least-squares optimization function.And an over-determined linear homogeneous equation is established with the improved orthogonal vanishing points and the camera intrinsic parameters.The Singular Value Decomposition method and the Cholesky method are adopted for accurately solving the intrinsic parameters.Secondly,with the help of camera intrinsic parameters,an over-determined linear equation is got by amending Tsai’s two-steps algorithm for partial extrinsic parameters.Depending on the property of rigid body transformation,the camera extrinsic parameters are calibrated via the least-squares approach.And then,taking the camera extrinsic parameters,sub-pixel corner point coordinates and absolute phase values of the sub-pixel corner points as input conditions,an over-determined linear homogeneous equation with depth parameters is constructed.After the normalization of the coefficient matrix,the Moore–Penrose pseudo-inverse method is nominated to calculate every depth parameter.Finally,the first two terms of radial distortion coefficients are introduced,and a nonlinear optimization is constructed by minimizing the functional of the residuals of coordinates.The unconstrained Levenberg-Marquardt algorithm is employed to calculate the distortion coefficients and refine all parameters.Many experiments have validated the feasibility,effectiveness and flexibility of the presented calibrating and optimizing system.An absolute phase recovery algorithm based on orthogonal vector diagram and bi-mapping rule is proposed.At first,inspired by the theory of force orthogonal decomposition and synthesis in physics,the phase shift principle of orthogonal vector diagram based on pixel point,intensity value(or gray value)and phase shift is expressed.Subsequently,after analyzing the classic double-three phase-shifting algorithm,the denoted source and complementary fringe images are fused for expanding the gray-scale values of the fringe maps.And a wrapped phase map is extracted precisely by eliminating the phase fluctuation errors caused by the Gamma distortion and the saturation.Finally,the bi-mapping rule based on two fringe orders with the same prime number is designed from the rounded phase values to the fringe order of each pixel.Absolute phase retrieval can be implemented accurately pixel by pixel by traversal searches.Simulated and experimental results have been carried out to validate the superior performances of the proposed method.An approach based on fringe order maps for distinguishing and removing invalid phase values is put forward.After analyzing the components and the connectivity of fringe order maps,the types and cases of invalid phase values,and the commonalities and shortcomings of the existing removal algorithms,the fringe order maps are affirmed as the processing object.The first step is to crudely refine the initial fringe order map to generate a milder fringe order map,which employs the square of the amplitude of the image gray-scale gradient,the moment of order(p,q)for the maximum region,Boolean negation and rounding operations.Due to the involvement of varietal region splitting and merging,next the milder fringe order map is precisely refined to produce a pure fringe order map.Finally,the background pixels,shadows,and abnormal discrete points are effectively identified and automatically removed by Boolean intersection from the original absolute phase map with the pure fringe order map.Experimental results showed that the proposed method was stable and reliable,and worked well in significantly eliminating invalid phase values.A prototype system of the monocular vision of fringe projection profilometry is developed for blade measurement.The performance of the system is evaluated in four aspects of scanning speed,measurement accuracy for one time,resolution and stability.A high pressure turbine blade is measured by the prototype system.The performance analysis and comparison experiments have verified that the system could obtain more complete surface point cloud data and possess high measurement efficiency and precision. |
