Structured light systems have been used in increasingly more applications for 3D shape measurement due to their fast measurement speed, good accuracy, non-contact characteristic, and portability. This dissertation is focused on improving the performance of the 3D shape measurement systems based on digital fringe projection, phase shifting and stereovision techniques. New camera and projector models and calibration algorithms as well as a novel system design based on a combined phase shifting and stereovision method are introduced.
The first part of this dissertation introduces systems based on digital fringe projection and phase shifting techniques. In this research, a color fringe pattern is generated by software and projected onto the object being measured by a digital-light-processing (DLP) projector working in the black and white (B/W) mode. The fringe images are captured by a high speed CCD camera, which is synchronized with the projector by software. The 3D model is reconstructed by using every three consecutive fringe images.
The previously developed linear calibration method does not take lens distortion into consideration and as a result, has very limited measurement accuracy. In this research, the effect of lens distortion on both the camera and projector is modeled based on careful calibration. Radial and tangential distortion parameters of different orders are analyzed and the right combination of parameters is chosen to provide an optimal performance. Experimental results show that the measurement accuracy has been improved by more than 75 percent (the RMS from 1.4 mm to 0.35 mm) after the implementation of the proposed nonlinear calibration method.
In the second part of this dissertation, a novel design, which combines the phase shifting and stereovision techniques, is proposed to eliminate errors caused by inaccurate phase measurement, for example, periodic errors due to the nonlinearity of the projector’s gamma curve. This method uses two cameras, which are set up for stereovision and one projector, which is used to project phase-shifted fringe patterns onto the object twice with the fringe patterns rotated by 90 degrees in the second time. Fringe images are taken by the two cameras simultaneously, and errors due to inaccurate phase measurement are significantly reduced because the two cameras produce phase maps with the same phase errors. One side effect of this method is that the projector calibration is not necessary, which simplifies the calibration of the entire system.
The use of a visibility-modulated fringe pattern is proposed as well to reduce the number of images required by this combined method. This new fringe pattern is sinusoidal in the horizontal direction as in a conventional fringe pattern, but is visibility-modulated in the vertical direction. With this new pattern, we can obtain the phase information in one direction and fringe visibility information in the other direction simultaneously. Since no pattern changing is necessary during the image acquisition process, the image acquisition time can be reduced to less than half of the time previously required, thus making the measurement of dynamically changing objects possible.
A color system is designed to further improve the speed of this system. Color cameras and color projector are introduced in this system. By utilizing these color devices, one color fringe image is sufficient to reconstruct a 3D model instead of three black and white fringe images. The three phase shifting fringe patterns are encode into the R, G and B channels of the color pattern which is projected onto the object. And the color fringe images taken by color cameras can be separated into three black and white images. By using this technique, we can further improve the speed of the structure light system, and the system will be more resistant to fast moving objects. (Abstract shortened by UMI.)