With the recent advancements in digital technology, three-dimensional (3-D) shape measurement has played an increasingly important role in fields including manufacturing, homeland security, medical sciences, and entertainment. Over the past decades, numerous 3-D shape measurement techniques have been developed. Among these existing techniques, fringe analysis based on phase-shifting sinusoidal structured patterns stands out because of its numerous advantages. However, there are still some major challenges of the existing digital fringe projection system for accurate 3-D shape measurement and for future speed improvement. They are: (1) projector nonlinearity problem, (2) synchronization problem, and (3) exposure time limitation problem. There are currently two approaches to generate sinusoidal fringe patterns with a digital-light-processing (DLP) projector: defocusing binary patterns (DBP) and focusing sinusoidal patterns (FSP). The focus of this dissertation research is to compare these methods for high-quality 3-D shape measurement.

We developed a system based on a digital fringe projection and phase-shifting technique to perform various comparison tests. The system utilizes a DLP projector to project computer generated fringe patterns onto the object and a charged-coupled-device (CCD) camera to acquire the fringe images. Conventionally, sinusoidal fringe patterns are usually supplied to a focused projector, and the DBP method is used to properly defocus the projector to generate sinusoidal patterns from binary structured patterns. We compare the performance of the new DBP approach against the traditional FSP method by analyzing the phase errors introduced by following factors: (1) defocusing degree, (2) exposure time, (3) synchronization, and (4) projector nonlinear gamma.

The traditional FSP involves some practical issues for high-quality measurement. Our experiment found it is possible to generate ideal sinusoidal fringe patterns by the DBP method, and when the projector is defocused to a certain degree, the phase error induced by the DBP method is very close to that produced by the FSP approach. With the DBP method, 3-D reconstruction was shown to be feasible.

Short exposure time is especially needed when measuring fast motion. For the FSP method, the minimum exposure time of the camera is limited by the projector's fringe projection rate, and the phase error is very large when a very short exposure time is needed. The experimental results show that the phase error does not change very much when the exposure time alters, and if a very short exposure time is needed, the DBP method clearly outperforms the FSP method for 3-D shape measurement. It also provides a potential way to develop fast 3-D shape measurement technique.

For the DLP projector, if it is supplied with sinusoidal fringe patterns, the synchronization between the projector and the camera is critical. When the projector is not synchronized with the camera, the phase error for the DBP method is much smaller than that for the FSP method when the exposure time is not multiples of projection cycle. By implementing the DBP method in our system, we could achieve 3-D reconstruction without synchronization between the projector and the camera.

Projector gamma correction, which is usually a time-consuming procedure, is mandatory for the FSP method. In this research, we found no projector gamma correction is needed for the DBP method. Our experimental results demonstrated it can achieve high-quality 3-D reconstruction by the DBP method without projector nonlinearity calibration.

Compared with the FSP method, the possible shortcomings of the DBP method are: (1) seemingly sinusoidal fringe patterns are still composed of high-frequency harmonics, which results in measurement error, and (2) the depth range of high-contrast fringe patterns is small. Even with these drawbacks, this new technique still has the potential to replace the conventional fringe generation technique.