Non-contact Vital Signs Measurement by Thermal Imaging Technology
-
摘要: 针对目前临床上监测生命体征设备的不便携带、接触人体等问题实现了一种将红外热成像仪作为信息采集设备,通过分析人体面部血管模型及鼻孔位置温差变化得到心率值和呼吸信息的方法。首先对获取的热像图序列提取前景目标以缩短在整幅图像中进行人脸检测的时间,再利用各向异性扩散滤波法增强感兴趣区域内血管位置的对比度,并利用形态学处理获得人脸血管部位的灰度均值形成初始心率信号。最终通过趋势消除、小波阈值去噪方法去除时间序列中的趋势项和随机噪声获取最终的心率波形图和动态心率、呼吸值。与医院专用设备对比试验得出该方法可控制心率误差小于4%,平均的均值误差为$ \left| {\bar d} \right| $=0.718次/min。呼吸误差在1次/min内,具有较高的准确性和鲁棒性,能够满足实际需求。Abstract: To solve the problems of inconvenient carrying and contact with the human body when using current clinical vital signs monitoring equipment, a method of estimating heart rate and respiratory information is proposed by analyzing the facial vascular model and temperature difference of the nostril position using an infrared thermal imager as a transmission device. First, the foreground target is extracted from the obtained thermal image sequence to shorten the time of face detection in the entire image. Anisotropic diffusion is then used to enhance the contrast of the vascular position in the region of interest, and the gray mean of the vascular position in the face is obtained by morphological processing to form the initial heart rate signal. Finally, trend elimination, wavelet threshold denoising, and other filtering methods were used to remove the trend item and random noise in the time series to obtain the final heart rate waveform, dynamic heart, and respiration values. Compared with specialized equipment in the hospital, it was found that the method exhibited a heart rate error of less than 4%, and the average error of the average value was 0.718 beats/min. The breathing error is within 1 beat/min, showing high accuracy and robustness and that the method can meet actual needs.
-
Key words:
- vital signs /
- thermal imaging /
- face detection /
- wavelet threshold denoising /
- image processing /
- non-contact detection
-
表 1 检测生命体征的非接触式方法对比
Table 1. Comparison of non-contact methods for vital signs detection
Non-contact measurement method Advantage Shortcoming Radar Strong penetrability and accurate measurement results Long-term radiation is harmful, susceptible to electromagnetic wave interference IPPG Strong portability, low cost, comfortable and non-invasive Easily affected by light, cannot detect at night Thermal imaging technology Non-invasive, all-day, all-weather detection; night-time detection; support for telemedicine The image details are not clear and the resolution is poor 表 2 残差序列ADF检验
Table 2. Residual sequence ADF test
Value t-statistic P-value Augmented Dickey-Fuller
test statistic−5.640742 0.000001 Critical value 1% level
5% level
10%level−3.605565
−2.937069
−2.606986表 3 男性测试者动态心率结果对比
Table 3. Comparison of dynamic heart rate results in male
Time/s 1 2 3 4 5 M/bpm R/bpm M/bpm R/bpm M/bpm R/bpm M/bpm R/bpm M/bpm R/bpm 5 65 65 70 72 75 77 65 63 69 70 10 66 64 70 71 74 75 65 65 69 69 15 64 64 69 72 73 75 67 65 70 68 20 62 64 72 70 73 73 66 65 68 67 25 65 63 72 72 72 72 65 64 66 65 30 63 64 71 73 71 73 63 63 65 65 35 65 64 70 71 72 73 64 63 67 65 40 65 65 69 70 72 72 64 65 65 67 45 63 64 69 69 73 72 63 63 65 65 50 64 65 70 70 73 71 63 63 66 66 55 64 65 70 71 72 72 64 63 68 67 60 65 64 71 72 71 72 63 64 66 67 表 4 女性测试者动态心率结果对比
Table 4. Comparison of dynamic heart rate results in female
Time/s 6 7 8 9 10 M/bpm R/bpm M/bpm R/bpm M/bpm R/bpm M/bpm R/bpm M/bpm R/bpm 5 78 78 80 81 68 70 82 85 80 78 10 76 77 79 81 69 69 85 86 80 79 15 77 76 78 80 68 68 81 84 81 78 20 78 75 79 79 67 68 82 83 82 78 25 79 76 78 79 68 67 84 84 80 82 30 78 78 80 78 68 67 83 85 82 81 35 80 78 81 80 65 67 81 85 84 83 40 79 77 80 82 64 65 82 83 83 81 45 78 76 79 80 65 65 82 84 85 83 50 78 76 78 79 68 66 82 85 83 82 55 77 76 79 79 67 65 82 82 82 82 60 77 75 79 79 66 65 81 83 81 82 表 5 对10个测试者心率均值偏差分析
Table 5. Deviation analysis of the mean heart rate of 10 testers
Number $ {\bar M_{{\text{estimate}}}} $/
(times/min)$ {\bar M_{{\text{true}}}} $/
(times/min)d/
(times/min)1 64.25 64.25 0 2 70.25 71.08 −0.83 3 72.58 73.08 −0.5 4 64.33 63.83 0.5 5 67 66.75 0.25 6 77.92 76.5 1.42 7 79.16 79.75 −0.59 8 66.92 66.83 0.09 9 82.25 84.08 −1.83 10 81.92 80.75 1.17 表 6 测试者呼吸平均值结果对比
Table 6. Comparison of the average results of the tester's breath
Number MBR/
(times/min)MBR′/
(times/min)RBR/
(times/min)error/
(times/min)1 13.92 14 13 +1 2 13.58 14 14 0 3 16.5 17 16 +1 4 12.33 12 12 0 5 13.5 14 14 0 6 15.91 16 16 0 7 17.83 18 18 0 8 12.33 12 13 -1 9 19.33 19 19 0 10 17.66 18 17 +1 -
[1] Chekmenev S Y, Farag A A, Essock E A. Multiresolution Approach for Non-Contact Measurements of Arterial Pulse using Thermal Imaging[C]//2006 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW'06), 2006: 129-129, doi: 10.1109/CVPRW.2006.139 [2] Garbey M, Sun N, Merla A, et al. Contact-free measurement of cardiac pulse based on the analysis of thermal imagery[J]. IEEE Trans. Biomed. Eng. , 2007, 54(8): 1418-1426. doi: 10.1109/TBME.2007.891930 [3] Gault T R, Farag A A. A fully automatic method to extract the heart rate from thermal video[C]// 2013 IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2013: 336-341. [4] 梁智敏, 陈骐, 肖书明, 等. 利用热成像技术对心率进行无接触检测的研究[J]. 中国体育科技, 2018, 54(1): 136-145. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTY201801019.htmLIANG Zhimin, CHEN Qi, XIAO Mingshu. Research on non-contact heart rate detection using thermal imaging[J]. China Sport Science and Technology, 2018, 54(1): 136-145. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTY201801019.htm [5] 景斌, 李海云. 基于红外序列图像的心率无损检测方法研究[J]. 中国生物医学工程学报, 2010, 29(6): 943-946. doi: 10.3969/j.issn.0258-8021.2010.06.024JING Bin, LI Haiyun. Study on nondestructive detection method of heart rate based on infrared sequence image[J]. Chinese Journal of Biomedical Engineering, 2010, 29(6): 943-946. doi: 10.3969/j.issn.0258-8021.2010.06.024 [6] 王建明, 王彦召. 基于红外热像图估计心率和心率波形的新方法[J]. 北京师范大学学报: 自然科学版, 2019(3): 324-328. https://www.cnki.com.cn/Article/CJFDTOTAL-BSDZ201903005.htmWANG Jianming, WANG Yanzhao. A new method for estimating heart rate and heart rate waveform based on infrared thermography[J]. Journal of Beijing Normal University: Natural Science, 2019(3): 324-328. https://www.cnki.com.cn/Article/CJFDTOTAL-BSDZ201903005.htm [7] 张栋, 高惠合, 温宝珠, 等. 人体正常红外面部图像和温度分布[J]. 自然杂志, 1992, 15(5): 398-399. https://www.cnki.com.cn/Article/CJFDTOTAL-ZRZZ199205027.htmZHANG Dong, GAO Huihe, WEN Baozhu. Normal infrared facial image and temperature distribution of human body[J]. Nature, 1992, 15(5): 398-399. https://www.cnki.com.cn/Article/CJFDTOTAL-ZRZZ199205027.htm [8] TANG Q, DAI S G, YANG J. Object tracking algorithm based on Camshift combining background subtraction with three frame difference[J]. Applied Mechanics & Materials, 2013, 373-375: 1116-1119. [9] Akula A, Khanna N, Ghosh R, et al. Adaptive contour-based statistical background subtraction method for moving target detection in infrared video sequences[J]. Infrared Physics & Technology, 2014, 63: 103-109. [10] SUN D, Roth S, Black M J. a quantitative analysis of current practices in optical flow estimation and the principles behind them[J]. International Journal of Computer Vision, 2014, 106(2): 115-137. doi: 10.1007/s11263-013-0644-x [11] 谢刚. 红外图像人脸识别方法研究进展[J]. 计算机工程与设计, 2008, 29(18): 4801. https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ200818051.htmXIE Gang. Research progress of infrared image face recognition method[J]. Computer Engineering and Design, 2008, 29(18): 4801. https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ200818051.htm [12] 曾华. 红外人脸图像识别研究[D]. 武汉: 武汉工程大学, 2011.ZENG Hua. Research on Infrared Face Image Recognition[D]. Wuhan: Wuhan Institute of Technology, 2011. [13] 刘鸿宇. HP滤波方法原理及其应用情况浅析[J]. 课程教育研究, 2018(34): 150-151. https://www.cnki.com.cn/Article/CJFDTOTAL-KCJY201834152.htmLIU Hongyu. HP filter principle and its application[J]. Research on Curriculum Education, 2018(34): 150-151. https://www.cnki.com.cn/Article/CJFDTOTAL-KCJY201834152.htm [14] 陈海龙, 王钧婷, 张岩. 时间序列平稳性检验方法分析及应用研究[EB/OL]. 中国科技论文在线[2015-09-02] [2015-09-29]. http://www.paper.edu.cn/releasepaper/content.CHEN Hailong, WANG Junting, ZHANG Yan. Analysis and application of time series stationarity test method[EB/OL]. science & technology magazine online. [2015-09-02] [2015-09-29]. http://www.paper.edu.cn/releasepaper/content. [15] 王瑞. 小波变换在红外图像处理中的应用研究[D]. 淮南: 安徽理工大学, 2016.WANG Rui. Application of Wavelet Transform in Infrared Image Processing[D]. Huainan: AnHui University of Science and Technology, 2016. [16] 李庆华, 山拜·达拉拜, 孙全富. 基于自适应小波阈值的心电信号去噪算法[J]. 计算机仿真, 2013, 30(10): 368-371. doi: 10.3969/j.issn.1006-9348.2013.10.084LI Qinghua, Senbai Dalabaev, SUN Quanfu. A denoising algorithm of ECG signal based on adaptive wavelet threshold[J]. Computer Simulation, 2013, 30(10): 368-371. doi: 10.3969/j.issn.1006-9348.2013.10.084 [17] QI C, WANG Z, HAN J, et al. Wavelet threshold denoising of thermal image from transmission joints[C]//2011 International Conference of Information Technology, Computer Engineering and Management Sciences, 2011: 108-111. [18] 王盛波. 基于视频实时持续心率检测及可视化[D]. 杭州: 杭州师范大学, 2015.WANG Shengbo. Continuous Heart Rate Detection Visualization Based on Video[D]. Hangzhou: Hangzhou Normal University, 2015. [19] 石超, 王永锋, 王正军. 复杂背景下前景目标提取算法研究[J]. 中国水运, 2019, 19(6): 58-59. https://www.cnki.com.cn/Article/CJFDTOTAL-ZSUX201906032.htmSHI Chao, WANG Yongfeng, WANG Zhengjun. Research on foreground target extraction algorithm under complex background[J]. China Water Transport, 2019, 19(6): 58-59. https://www.cnki.com.cn/Article/CJFDTOTAL-ZSUX201906032.htm