Research On Digital Projection Based Three-dimensional Fringe Pattern Profilometry Technique

Fringe pattern profilometry is a popular and non-contact technique for three-dimensionalprofile reconstruction. Because of its simplicity for the equipment, fast computation and highmeasurement accuracy, fringe pattern profilometry technique has been widely utilized into theareas of 3-D sensing, machine vision, industry monitoring, mechanical engineering, etc. Inrecent years, in order to realize digitalized three-dimensional profilometry, digital projectorscontrolled by computers have been widely introduced into the FPP system, so that fully automaticprofilometry can be achieved.However, different from traditional optical projection equipments, digital projectors willintroduce visible distortions when they are used to generate fringe patterns, including nonlineardistortion and colour distortion, which consequently result in measurement errors forthe reconstructed surface when using conventional fringe pattern analysis methods. Therefore,it becomes the main problem to adaptively reduce the influence from nonlinear andcolour distortion and improve the reconstruction accuracy of digital projection based fringepattern profilometry. In this thesis, the key techniques regarding digital projection based threedimensionalfringe pattern profilometry are thoroughly studied. The research programme issupported by the international cooperative research project of Hubei Key Laboratory of IntelligentInternet Technology of Huazhong University of Science and Technology and the Universityof Wollongong, and funded by Australia Research Council Discovery Project titled"Multi-resolution phase measuring profilometry for dynamic 3D digital imaging".The thesis begins with tracing the history and development of fringe pattern profilometrytechnique and briefly introduces its principles, relevant traditional fringe pattern analysisalgorithms and typical applications. Subsequently, focusing on digital projection based fringepattern profilometry, the author has theoretically analyzed the measurement errors for the traditionalmethods when digital projection are utilized to generate fringe patterns and discoveredthat with digital FPP nonlinear distortion is the primary problem that influences the accuracyof profile reconstruction, which is unavoidable in practice and makes the conventional fringepattern analysis methods no longer available. In order to solve the problem the author presents a generalized analysis model (GAM) for fringe pattern profilometry (FPP), which is a moregeneral mathematical model for the fringe pattern profilometry system. This model does notdepend on the characteristics of the projection and acquisition systems. Consequently, theerrors caused by the nonlinear distortion of the system is completely prevented. The general-ized analysis model converts the problem of the profile measurement into an equivalent taskof calculating the distribution of a shift function. Further, based on the generalized model, aseries of algorithms are presented to precisely reconstruct three-dimensional surfaces by utiliz-ing digitally projected fringe patterns. At first, the author presents two direct shift estimationalgorithms, inverse function based shift estimation (IFSE) algorithm and gradient based shiftestimation (GSE) algorithm, both of which directly calculate the values of the shift function forall the points by using the captured fringe patterns. However, as the height distribution of thepoints on the object surface highly depends on each other, it is not economical to independentlycalculate height distribution for every point. Therefore, the author proposes two parametricalshift estimation algorithm, including optimal filtering based shift estimation (OFSE) algorithmand DCT based shift estimation algorithm. Parametrical shift estimation algorithms expressthe shift function by vectors with fewer dimensions and estimate the parameters based onthe generalized analysis model, so that they can also accurately reconstruct three-dimensionalobject surfaces and at the same time faster computation can be achieved. The measurementresults are compared with traditional algorithms and showing that the proposed GAM and al-gorithms are capable of significantly improving the reconstruction accuracy for digital fringepattern profilometry system.Additionally, colour channel based fringe pattern profilometry is also in the scope of thethesis. The author mathematically models so-called colour cross-talk problem, which intro-duces main measurement errors for colour fringe pattern profilometry system, and presents ablind colour isolation (BCI) algorithm to achieve adaptive and effective colour separation andimprove the reconstruction accuracy of colour fringe pattern profilometry.Finally, the author concludes the thesis and looks forward to the future work.