Research On Non-contact Heart Rate And Respiration Measurement Method Based On Thermal Infrared

With the high incidence of cardiovascular and cerebrovascular diseases such as coronary heart disease,hypertension,and arrhythmia,people are paying more and more attention to the detection and prevention of these diseases.The current contact measurement method requires direct or indirect contact with a specific part of the tester’s skin and keeps still,which brings inconvenience and limitation to the heart rate detection of special populations.In the non-contact measurement method,the thermal image obtained by the method of measuring vital signs based on infrared thermal imaging technology has the problems of poor resolution,blurred details,and poor stability and accuracy of measurement results.In this regard,this paper realizes a method that can accurately and real-time obtain the subject’s dynamic heart rate and respiration value.The main research work is as follows:1)Based on face detection and tracking.In this paper,the method based on target contour extraction and background difference is used to extract the slow moving dynamic foreground target.The combination of the extraction of the foreground target and the face detection will result in pedestrians with undetected faces in the background and non-targets with high temperature in the background.Objects are culled.Then,the method of face detection and face alignment based on feature points is used to locate the face area in real time,and the deep learning network can be used to predict the location of the feature points even if some of the features are occluded.2)Enhance the edges and filter out noises for problems such as low contrast and blurred edges of the thermal image.Use histogram equalization to achieve the effect of image enhancement,and use anisotropic diffusion filter to treat the pixel value as heat flow to improve the contrast of the blood vessel edge,after morphological processing to segment the edge of the blood vessel area in the image and the gray value in the area Stored in an array;finally,median filter,Butterworth bandpass filter,VMD and wavelet multi-resolution threshold denoising are combined to eliminate noise in the image and improving the accuracy of life parameter detection.3)Smoothing of dynamic heart rate value.For the phenomenon of violent oscillations and instability of the acquired dynamic heart rate value,analysis of the acquired initial heart rate time series contains certain regularly changing trend items and non-stationary data,and HP filtering and stationarity test are used to remove the trend items.The experimental results show that the heart rate and respiration values obtained by the methods based on foreground moving target extraction and face detection and tracking,face alignment,and gray value extraction of blood vessel regions used in the article are relative to the reference values obtained by professional equipment used in hospitals.The error is less than 4%,and the average error is 0.718bpm.In the Bland-Altman analysis,the method in this paper is in good agreement with the results obtained by professional equipment,which proves the feasibility of the method.