Life detection and non-contact respiratory rate measurement in cluttered environments
- PDF / 2,528,466 Bytes
- 13 Pages / 439.642 x 666.49 pts Page_size
- 94 Downloads / 184 Views
Life detection and non-contact respiratory rate measurement in cluttered environments Shiqi Li1 · Haipeng Wang1
· Shuze Wang1 · Shuai Zhang1
Received: 12 April 2020 / Revised: 15 July 2020 / Accepted: 31 July 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract A method is proposed in this paper for life detection and non-contact respiratory rate measurement in cluttered environments. Only an RGB video of the detection area is required. In the method, spatial filtering is firstly applied to each frame of the video for image denoising. Gray level compensation follows to compensate for the change of gray level caused by the environment light. Thirdly, the gray levels of each pixel over time are filtered separately by a low-pass filter. At last, the human is located and the respiratory rate is measured. Tests on a self-made dataset show that an accuracy of 76.7% is achieved by the proposed method, which is better than that of the Convolutional Neural Networks (30%) and the histogram of oriented gradients (3.3%). Keywords Life detection · Respiratory rate measurement · Spatial filtering · Gray level compensation · Temporal filtering
1 Introduction The increasing extent of natural disasters, particularly earthquakes, hurricanes, and tsunamis, motivate research in search and rescue. Life detection and vital signs (the respiratory rate and body temperature, etc.) measurement are critically important and challenging in this field. Body temperature can be obtained easily through the infrared camera [9], while the respiratory rate is relatively hard to obtain. In this paper, both the life detection and Haipeng Wang
[email protected] Shiqi Li [email protected] Shuze Wang [email protected] Shuai Zhang [email protected] 1
School of Mechanical Engineering, Science, Huazhong University of Science and Technology, Wuhan 430074, China
Multimedia Tools and Applications
respiratory rate measurement are studied, and the respiratory rate should be measured in a non-contact way [18]. The radio [28], radar [15], and commercial Wi-Fi devices [26] can be used for life detection and respiratory rate measurement. However, there are many restrictions on the use of such devices. For example, the radar [8] is hard to obtain, and the Wi-Fi device requires the human to be close to the device [27]. In this paper, only the visual (RGB or RGB-D) processing methods are concerned. The reason is that the RGB/RGBD cameras are most widely used in robotics and computer vision. In earlier studies, the histogram of the oriented gradient (HOG) [6] is popularly used to detect full-body or different parts [14, 20] (face, hand, etc.) of humans. However, such methods are easy to fail when the human is in a messy background. Convolutional Neural Networks (CNN) attracts a lot of attention in recent years [2, 23]. CNN can learn features of humans from massive annotation datasets (the ImageNet [7], Microsoft coco [19], and MPII [21], etc.). The detection accuracy improved greatly compared with that of the HOG. How
Data Loading...
