Fig 1.
(a) RGB and infrared records of Kinect v1 and corresponding depth map. The black pixels in the depth image reflect areas of unknown depth, which in this record is due to the out-of-record boundary, noise, and shadow effect on the left side. (b) RGB and IR images of Kinect v2 sensor and the corresponding estimated depth information (right panel), the black pixels in the depth image reflect areas of unknown depth.
Table 1.
Comparison of the most common motion capture system.
Fig 2.
The spectrum of emitted lights from Kinect sensors and Qualisys motion capture system.
The black curve shows the emitted light spectrum of fluorescent lights in the lab. While the spectrum of the projected IR ray from Microsoft Kinect v1 and Microsoft Kinect v2 are shown in green and red. The spectrum of emitted light from Qualisys Oqus 300/310 is shown in blue. However, a part of the curve with wavelength higher than 900nm were estimated.
Fig 3.
Geometric placement of sensors and bulletin board in the motion lab.
Microsoft Kinect v2 was utilized, and 8 Qualisys cameras are mounted in the laboratory (top view). Only five cameras pointed to the bulletin board surface (i.e., involved cameras), and other cameras were placed behind the bulletin board (i.e., uninvolved cameras). The Kinect sensor was 1.2m of the board and moved up to 5.2m with 1.0m steps. The region of interest (ROI) on the bulletin board surface is shown with a yellow color.
Fig 4.
Based on a single depth map, the optimal plane was determined based on the cloud point while trying to minimize the error rate. The blue mesh represents the cloud points of the bulletin board captured by Microsoft Kinect v2 (depth information), and the corresponding optimal plane is shown in red.
Table 2.
Median and IQR of estimated RRMSframe using Microsoft Kinect v2 in absence and presence of the motion capture as a noise source for five different distance of the Kinect sensor from the board.
Fig 5.
Measured RRMSFrame values in absence and presence of Qualisys as an interference source in Kinect v2 depth records.
The values represented the median of RRMSFrame and range stand as corresponding IQR. The line graphs were interpolated based on the measurements.
Table 3.
Bland-Altman analysis of the estimated RRMSframe using Microsoft Kinect v2 evaluating the impact of the motion capture as a noise source for five different distance of the Kinect sensor from the board.
Fig 6.
Comparing the impact of Qualisys on the estimated RRMS pixel in Microsoft Kinect depth records.
Where Kinect was placed at (a) 120cm, (b) 220cm, (c) 320cm, (d) 420cm, (e) 520cm of the bulletin board (ROI). ROI dimension where 270×270, 150×150, 100×100, 80×80, 70×70 pixels from 1.2m to 5.2m. In the figure, all ROI images are resized to provide a better presentation.
Fig 7.
Comparing the impact of Qualisys on the entropy of Microsoft Kinect depth records.
Where Kinect was placed at (a) 120cm, (b) 220cm, (c) 320cm, (d) 420cm, (e) 520cm of the bulletin board (ROI). ROI dimension where 270×270, 150×150, 100×100, 80×80, 70×70 pixels from 1.2m to 5.2m. In the figure, all ROI images are resized to provide a better presentation.
Fig 8.
Impact of reflective markers and IR beams in Kinect v2.
(a) an IR record of Microsoft Kinect v2 while Qualisys Oqus cameras were turned off and (b) corresponding depth information. (c) An IR record of Microsoft v2 while Qualisys Oqus cameras were turned on and (d) the corresponding depth information. The black areas in the depth images are representing the areas of unknown distance.
Fig 9.
Impact of reflective markers on Microsoft Kinect v2 recordings.
(a) RGB image, (b) IR image and (c) depth information. The black areas in the depth images are representing the areas of unknown distance.