Temporal Matrices Mapping-Based Calibration Method for Event-Driven Structured Light Systems
Temporal Matrices Mapping-Based Calibration Method for Event-Driven Structured Light Systems
复制标题
事件驱动结构光系统的基于时间矩阵映射的校准方法
DOI：
--
复制
发表时间：
2021
影响因子：
4.3
通讯作者：
Huazhong Yang
中科院分区：
文献类型：
--
作者：
Guijin Wang;Chenchen Feng;Xiaowei Hu;Huazhong Yang
由于传感器带宽和光源功率的限制，强环境光照严重降低了传统结构光三维成像系统的性能。相比之下，事件驱动结构光技术充分利用了激光振镜扫描和事件检测的特性，可以在这种具有挑战性的场景下实现鲁棒的3D重建。然而，这种系统的低测量精度严重阻碍了它们的广泛应用，因为还没有为它们开发精确的校准方法。在这项工作中，我们提出了一种新的时间矩阵映射（TMM）的事件驱动的结构光系统的校准算法。我们的方法的关键步骤是建立检流计和事件相机图像平面与两个时间矩阵之间的像素对应关系。具体地说，我们1）垂直和水平扫描一个前平行平面，得到两个时间矩阵; 2）通过时间矩阵和相应的扫描速度估计振镜像平面上特征点的坐标。为了充分利用我们的校准方法，我们提出了一个视差校正方法的深度计算。我们开发了一个原型系统来验证所提出的算法。实验结果表明，该标定算法可以达到亚像素精度，系统的测量误差可以达到0.2%，优于典型的1.0%的最先进的。
Strong ambient illumination severely degrades the performance of conventional structured light 3D imaging systems due to the limited sensor bandwidth and light source power. In contrast, event-driven structured light techniques fully take advantage of laser-galvanometer scanning and event-detection property, which can achieve robust 3D reconstruction under such challenging scenarios. However, the low measurement accuracy of such systems severely hinders their extensive applications, as no accurate calibration method has yet been developed for them. In this work, we propose a novel Temporal Matrices Mapping (TMM) based calibration algorithm for event-driven structured light systems. The crucial step of our method is establishing the pixel correspondences between the galvanometer and event camera image planes with two temporal matrices. Specifically, we 1) scan a front-parallel plane vertically and horizontally to attain two temporal matrices; 2) estimate the coordinates of feature points on the galvanometer image plane through the temporal matrices and corresponding scanning speeds. In order to make the most of our calibration method, we present a disparity correction approach for depth calculation. We developed a prototype system to validate the proposed algorithms. Experimental results demonstrate that the calibration algorithm can reach sub-pixel precision, and the system’s measurement error can achieve 0.2%, which outperforms the typical 1.0% of the state-of-the-art.
DOI：
10.1007/978-3-030-01219-9_2
发表时间：
2018
期刊：
European Conference on Computer Vision
影响因子：
--
作者：
Wang, Jian;Bartels, Joseph;Whittaker, William;Sankaranarayanan, Aswin C.;Narasimhan, Srinivasa G.
通讯作者：
Narasimhan, Srinivasa G.