基于异构多核构架的双目散斑3维重建

本文提出一种基于异构多核构架的双目立体散斑三维重建方法。该方法在构建双目立体视觉模型基础上，辅助投射白色散斑结构光，然后通过采集高分辨率双目散斑变形图像，经过极线校正后，运用零均值归一化互相关函数(ZNCC)作为相关算法的匹配代价函数，对传统的ZNCC快速计算方法进行修改并将其移植到异构多核处理器Myriad2上，实现了高精度物体三维重建效果。使用该方法进行三维物体重建可有效克服传统立体视觉算法对弱纹理区域重建效果较差的缺点，并借助异构多核构架处理器的自身特点，重充分发挥其强劲的并行运行能力，在不损失系统重建精度的前提下，使系统运行时间极大缩短，对系统的重建效率具有较大提升。

3D Reconstruction Method with Binocular Speckle Based on Heterogeneous Multi-core Processor

This paper presents a 3D reconstruction method with binocular speckle images based on heterogeneous multi-core processor. Depth images of the target are estimated from the correlation of high resolution binocular speckle projection image. Zero-mean normalized cross correlation (ZNCC) is adopted as the correlation function. Traditional fast-ZNCC calculation method is modified to implement on Myriad2 platforms. The shortcoming of low precision and efficiency of the previous stereo vision algorithm for weak texture regional reconstruction is overcome by this method with the parallel characteristics of heterogeneous multicore architecture .The experiments indicate that the heterogeneous computational architecture significantly improves the efficiency and performance in 3D objects reconstruction with binocular stereo speckle.