Nguyen Hieu, Wang Zhaoyang, Jones Patrick, Zhao Bing
Appl Opt. 2017 Nov 10;56(32):9030-9037. doi: 10.1364/AO.56.009030.
Consumer-grade red-green-blue and depth (RGB-D) sensors, such as the Microsoft Kinect and the Asus Xtion, are attractive devices due to their low cost and robustness for real-time sensing of depth information. These devices provide the depth information by detecting the correspondences between the captured infrared (IR) image and the initial image sent to the IR projector, and their essential limitation is the low accuracy of 3D shape reconstruction. In this paper, an effective technique that employs the Kinect sensors for accurate 3D shape, deformation, and vibration measurements is introduced. The technique involves using the RGB-D sensors, an accurate camera calibration scheme, and area- and feature-based image-matching algorithms. The IR speckle pattern projected from the Kinect projector considerably facilitates the digital image correlation analysis in the regions of interest with enhanced accuracy. A number of experiments have been carried out to demonstrate the validity and effectiveness of the proposed technique and approach. It is shown that the technique can yield measurement accuracy at the 10 μm level for a typical field of view. The real-time capturing speed of 30 frames per second makes the proposed technique suitable for certain motion and vibration measurements, such as non-contact monitoring of respiration and heartbeat rates.
消费级红绿蓝和深度（RGB-D）传感器，如微软Kinect和华硕Xtion，因其低成本和对深度信息进行实时传感的稳健性而成为有吸引力的设备。这些设备通过检测捕获的红外（IR）图像与发送到IR投影仪的初始图像之间的对应关系来提供深度信息，其主要局限性在于三维形状重建的精度较低。本文介绍了一种利用Kinect传感器进行精确三维形状、变形和振动测量的有效技术。该技术涉及使用RGB-D传感器、精确的相机校准方案以及基于区域和特征的图像匹配算法。从Kinect投影仪投射的红外散斑图案极大地促进了感兴趣区域内数字图像相关分析，提高了精度。已进行了多项实验以证明所提出技术和方法的有效性。结果表明，该技术在典型视场下可实现10μm级别的测量精度。每秒30帧的实时捕获速度使所提出的技术适用于某些运动和振动测量，如呼吸和心率的非接触监测。
