Fringe projection is an extensively applied technique for optical three-dimensional (3-D) shape measurements. Although showing favorable performance for motionless objects, it tends to have difficulties to retrieve surfaces globally or locally varying over time. The reason is that common methods developed for static scenes are prone to fail when measuring dynamic processes. Therefore, to facilitate the application of high-speed real-time measurements, we suggest considerations from four aspects to improve conventional fringe projection methods. The first two aspects are related to raising the measuring efficiency, which can be achieved by encoding the measured object robustly yet with less required patterns, and by increasing the rate of pattern projection which is a bottleneck restricting the measuring speed. The third consideration is to obtain accurate 3-D reconstructions by removing unreliable points induced by system and random errors during dynamic measurements. The last one is to handle moving shiny objects as it is supposed to be a time-consuming process for traditional approaches. We believe the mentioned considerations will help ease the efforts to achieve desired results for fast real-time measurements.
|