Digital Fringe Projection 3D Measurement with Improved Flexibility and Accuracy


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date3 Dec 2018


Digital Fringe Projection (DFP) techniques are commonly utilized in non-contact 3D shape measurement due to their fast speed and high accuracy. We improve several aspects of the flexibility and accuracy of DFP technique in this work.

The first aspect is binary fringes generation. Phase-shifted sinusoidal fringes are often projected in a DFP system to encode object height information. Grayscale sinusoidal fringes encounter the problems of gamma non-linearity and low projector refresh rate. Binary fringes projected by a defocused projector to imitate the sinusoidal fringes are thus very popular. They add flexibility to the system because no non-linear gamma correction is needed, and the refresh rate for the projector can be higher. To produce high-quality binary fringes, researchers have proposed to dither the sinusoidal fringes and optimize them in the intensity domain or phase domain. Phase-based optimization is direct and efficient because the fringes’ phase quality is directly related to the 3D measurement quality. However, phase-optimized fringes are very sensitive to projector defocusing levels, which are hard to quantize and adjust in practical use. Intensity-based optimization is more robust against the projector defocusing level, but it does not fully optimize the fringes’ phase quality. In this research project, we combine the merits of both intensity and phase-based optimization, which include a preliminary intensity optimization and a further optimization based on the synthesized error function. This technique is implemented and tested in two frameworks—the whole-fringe optimization and the best-patch optimization—to generate binary fringes.

After projecting and capturing the fringes, we focus on object segmentation from shadows and background. Shadows and background are two common factors in digital fringe projection that lead to ambiguity in three-dimensional measurement. Preprocessing is often needed to segment the object from shadows and background. Normally, existing segmentation approaches based on modulation perform well in purely dark background circumstances but lose accuracy in situations of white background. In our research, an accurate shadow and background removal technique is proposed, which segments the shadows by one threshold from the modulation histogram and segments the background by the threshold in the intensity histogram of an extra coding map.

After we obtain the object phase map, we need to transform the phase information into 3D coordinates through system calibration, including phase-to-height (Z) calibration and in-plane (X-Y) calibration. A lot of research can be found for phase-to-height calibration. Some of the existing methods require rigid structure for the relative position of the projector, camera and reference plane; some of them require the projector to be correctly focused, which is not flexible for an arbitrarily arranged system; some of them calibrate the system for each pixel and save the results in a look-up-table, which is very complicated and time-consuming. The in-plane coordinates are often calculated from the rotation and transformation matrix of each calibration plane, while the error accumulates from the camera calibration. In our research, a flexible global calibration technique is proposed for an arbitrarily arranged digital fringe projection system. The height information of each pixel is calculated from the phase map. The in-plane (X-Y) coordinates are calculated from height and phase information and optimized by a least-squared algorithm. The calibration method is flexible enough to be utilized in all kinds of phase based DFP system, including the one with a defocused projector. Simulation analysis and experiments are presented to verify the above techniques.