Investigating the impact of a motion capture system on Microsoft Kinect v2 recordings: A caution for using the technologies together

Abstract

Microsoft Kinect sensors are considered to be low-cost popular RGB-D sensors and are widely employed in various applications. Consequently, several studies have been conducted to evaluate the reliability and validity of Microsoft Kinect sensors, and noise models have been proposed for the sensors. Several studies utilized motion capture systems as a golden standard to assess the Microsoft Kinect sensors, and none of them reported interference between Kinect sensors and motion capture systems. This study aimed to investigate possible interference between a golden standard (i.e., Qualisys) and Microsoft Kinect v2. The depth recordings of Microsoft Kinect sensors were processed to estimate the intensity of interference. A flat non-reflective surface was utilized, and smoothness of the surface was measured using Microsoft Kinect v2 in absence and presence of an active motion capture system. The recording was repeated in five different distances. The results indicated that Microsoft Kinect v2 is distorted by the motion capture system and the distortion is increasing by increasing distance between Kinect and region of interest. Regarding the results, it can be concluded that the golden standard motion capture system is robust against interference from the Microsoft Kinect sensors.

Evaluation of the Pose Tracking Performance of the Azure Kinect and Kinect v2 for Gait Analysis in Comparison with a Gold Standard: A Pilot Study

“…As mentioned by Naeemabadi et al [ 28 ], retro-reflective markers, like those of the Vicon system, appear as noise in the IR images, so their 3D position is unknown to the Kinect sensors. To determine the 3D position of the Vicon markers, nevertheless, the intermediate intersection points on the chessboard were identified in the 2D color images using the OpenCV library, followed by the coordinate mapper functions provided by the Kinect SDKs to obtain the actual 3D positions from the depth camera space, denoted as points in the point set P .…”

Use of the Azure Kinect to measure foot clearance during obstacle crossing

“…In our preliminary recordings and a previous study (Naeemabadi et al, 2018), it was observed that infrared markers placed on the ankle and knee for optical motion capture could hinder the recording of depth images by the Azure Kinect sensor. Instead of placing a marker on the knee and ankle, we created a rigid body model for the shank segment before the recording session.…”

Markerless 3D Human Pose Tracking in the Wild with Fusion of Multiple Depth Cameras: Comparative Experimental Study with Kinect 2 and 3

“…In addition, the synchronization between the sensors constrains the Qualisys system to work at 30 Hz instead of 300 Hz. Naembadi et al studied this question with Kinect 2, but Kinect 3 seems to be much more sensitive to interference [21]. Another point to highlight is the computational cost for running Kinect 3.…”