High-resolution, High-speed 3-D Dynamically Deformable Shape Measurement Using Digital Fringe Projection Techniques

Abstract

With recent advancements of 3-D geometric shape analysis, acquiring 3-D dynamic scenes in real time becomes increasingly important in enormous fields, such as manufacturing, medical sciences, computer sciences, homeland security, and entertainment, etc. Over the last decades, 3-D shape measurement techniques have been improving dramatically rapidly. Numerous techniques have been developed including time of flight, stereo vision, structured light, and digital fringe projection. Because of the development of digital display technology, the structured-light-based methods have the potential to be the most important technology for 3-D shape measurement in both scientific research and industrial practices in the long run. Real-time 3-D shape measurement has recently emerged as a quickly expanding field with the enhancements of computation power of an ordinary personal computer. For real-time 3-D shapemeasurement, 3-D shapes have to be acquired rapidly, processed quickly, and displayed in real time (Zhang, 2010). In the past few years, a number of technologies have been developed. Among which, the optical method is one of the core techniques. However, due to its fundamentally difficult nature, capturing 3-D dynamic scenes with high quality remains challenging, and even fewer systems are capable of processing 3-D geometry in real time because of the complexity of the problem Our group has devoted a vast amount of effort in real-time 3-D shape measurement field. Over the past few years, we have been developing technologies using a technique called digital fringe projection and phase-shifting technique, a variation of the structured light technique with structured patterns being sinusoidally changing intensity. We have developed a number of algorithms to improve the processing speed, and taken advantage of the most advanced hardware technologies, graphics processing unit (GPU), to increase the speed of 3-D geometry reconstruction and rendering. With these technologies, we have successfully developed a high-resolution, real-time 3-D shape measurement that achieved simultaneous 3-D absolute shape acquisition, reconstruction, and display at a speed of faster than 26 frames / sec with 300K points per frame under an ordinary personal computer (Zhang et al., 2006). Experiments have demonstrated that the system can accurately capture dynamic changing 3-D scenes, such as human facial expressions. The data acquired by such a system have already been applied to 2