基于并行小波变换的快速三维面形测量技术

Fast Three-Dimensional Shape Measurement Technique by Means of Wavelet Transform Based on Parallel Computing

为了提高小波变换三维面形测量技术中相位信息的解调速度,依据不同的并行计算原理,提出两种基于并行一维连续小波变换的光学条纹图快速解调方法。利用多核CPU运算平台进行了计算机模拟和实验,结果表明并行一维连续小波变换光学条纹图相位解调方法相比于串行一维连续小波变换光学条纹图相位解调方法,在精度不变的同时计算速度显著提高:对850pixel×1000pixel大小实测条纹图像解调速度提高了7.5倍。并行一维连续小波变换相位解调方法为小波变换三维面形测量技术在实时/瞬态过程三维面形测量中的应用奠定了基础。

In order to improve the speed of phase demodulation in three-dimensional （3D） shape measurement system, according to different theories of parallelism, two different approaches of parallel one-dimensional （1D） continuous wavelet transform fringe pattern demodulation have been proposed. Computer simulation and experiment in multicore CPU computing platform show that, compared to serial phase demodulation, parallel fringe pattern phase demodulation technique keeps the computing precision while the processing speed is significantly improved： the experiment on processing 850 pixel × 1000 pixel fringe pattern achieves 7.5 times speedup. Our solutions lay the foundation for the applying of fringe pattern demodulation technique by means of wavelet transform to real-time/ transient 3D shape measurement.