Research On Portable Whole-Profile Measurement System Based On Digital Fringe Projection

The digital fringe projection profilomentry(DFPP) is non-conact and full-field 3D coordinate technique. To implement portable whole-profile measurement for color object, in this thesis a novel profile measurement method is developed based on the DFPP. The research efforts being made in this thesis include sensor design, system calibration, directional pattern design, global system calibration and 3D registration technique and etc. This is a promising method, which can be used in the fields of industrial design, quality control, medical inspection and virtual reality and etc. The major contributions of this thesis are listed as follows,1. The structure of sensor is established using DLP projection and double-CCD cameras. The unicolor CCD and DLP take charge of the 3D coordinate measuring, and the color CCD captures the color information and renders the coordinated model. The model of coordinate-phase relation is analyzed. And the color rendering algorithm based on stereo vision theory is proposed.2. The phase modulation and unwrapping algorithm of phase shifting technique is analyzed theoretically. The noise-reducing method of phase extraction is introduced as the decisional standard of phase shifting algorithm. The temporal phase unwrapping technique is emphasized in the paper because it can extract the absolute phase easily.3. The parameters of coordinate-phase and color rendering model are divided into two kinds, system parameters and variable. According to the proposed analysis method based on measuring model, the simple space mapping model is established as a replacer with n-order polynomial. A novel virtual space pattern is proposed to provide the given points with only one-plane pattern.4. A novel directional pattern is designed, and the mapping location algorithm for the proposed pattern is presented at the same time. The mapping relations of pixel coordinate and world coordinate can be defined by the collinear constrain. The process algorithm and flow for the novel pattern are designed.5. The whole-scale measuring technique is researched. The emphasis is placed on the whole-scale measuring with multi-sensors, whose position relations are calibrated with one-plane globally. The overlap definition of measured data based on pattern is presented to improve the efficiency of matching in 3D registration algorithm. And the registration strategy based on Levenberg-Marquardt optimization is analyzed to merge data.6. Automatic system decision and manual non-simple polygon selection methods are applied to eliminate unwanted and noise data successfully, which possibly makes improper color models reconstructed. A data reduction method is presented based on Grid reduction method considering color-boundary preservation. The reconstruction method is introduced in details, and the display and processing platform is developed based on OpenGL.The creative work of this thesis includes:1. The simple space mapping calibration model is established with n-order polynomial according to the proposed analysis method of measuring model. The novel virtual space pattern is designed to calibrate system rapidly only with one-plane pattern.2. A novel directional pattern is presented with the mapping location algorithm based one the collinear constrain. The process algorithm and flow for the novel pattern are designed.3. The same plane pattern is also applied to the global calibration of multi-sensors. The overlap definition of measured data based on pattern is presented for improving the efficiency of matching in 3D registration algorithm.4. Automatic system decision and manual non-simple polygon selection method is applied to eliminate unwanted and noise data. A data reduction method is presented based on Grid reduction method considering color-boundary preservation in RGB color space.