An elderly state analysis method based on image recognition

Abstract

The present invention relates to the technical field of elderly status monitoring, and specifically to an elderly status analysis method based on image recognition, which includes the following steps: S1. Obtain the first image data of the elderly in real time from multiple angles, as well as the physiological data of the elderly including body size data, heart rate record, and blood pressure record; S2. Identify the identity of the elderly, extract the initial key points in the first image data, and determine whether there are errors; if so, proceed to step S3; if not, mark the initial key points as determined key points, mark the first image data as the second image data, and proceed to step S4; S3. Correct the initial key points in the first image data to generate the second image data; The present invention takes into account that the changes in heart rate and blood pressure may occur when the elderly fall, so the calculation accuracy of the comprehensive risk degree value is further improved, so that medical staff can take corresponding treatment measures in a timely manner according to the comprehensive risk degree value.

Description

Technical Field

[0001] The present invention relates to the technical field of elderly status monitoring, and specifically provides a method for analyzing the status of the elderly based on image recognition. Background Art

[0002] The main responsibility of a nursing home is to provide comprehensive care and services for the elderly to ensure their quality of life, health, and safety. However, due to the limited number of medical staff in the nursing home, it is often impossible to cover everything. Therefore, it is necessary to install multiple monitoring devices in the nursing home to take real-time pictures of the elderly and analyze the status of the elderly based on the taken pictures.

[0003] However, the common analysis method is to directly use the openpose algorithm to extract key points, and then directly predict whether the elderly person has fallen based on the changes in the key points and the physiological data of the elderly. However, this method does not take into account the impact of the elderly person's fall on the physiological data, so there are errors in calculating the actual status and risk level of the elderly, which is not conducive to medical staff taking corresponding measures accurately based on it. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a method for analyzing the status of the elderly based on image recognition, which solves the technical problem that the prior art does not consider the impact of the elderly person's fall on the physiological data, resulting in errors in the analysis of the actual status of the elderly.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for analyzing the status of the elderly based on image recognition, comprising the following steps:

[0007] S1. Obtain the first image data of the elderly from multiple angles in real time, as well as the physiological data of the elderly including body size data, heart rate records, and blood pressure records;

[0008] S2. Identify the identity of the elderly, extract the initial key points in the first image data, and determine whether there are errors;

[0009] If so, proceed to step S3;

[0010] If not, mark the initial key points as determined key points, mark the first image data as the second image data, and proceed to step S4;

[0011] S3. Correct the initial key points in the first image data to generate the second image data;

[0012] S4. Analyze the current status of the elderly based on the continuous second image data and the physiological data of the elderly;

[0013] S5. The comprehensive risk degree value Provide real-time feedback to the staff.

[0014] Further, in step S2, it specifically includes the following steps:

[0015] S21. Construct initial key edges for representing the bones and torso of the elderly according to the initial key points;

[0016] S22. Obtain the calibration parameters of the monitoring device, including the focal length, principal point coordinates, rotation matrix, and translation vector;

[0017] S23. Construct a world coordinate system with one of the monitoring devices as the origin, select any two pieces of first-image data, and calculate the three-dimensional coordinates of each initial key point according to the calibration parameters ;

[0018] S24. Calculate the length of each initial key edge , and its calculation formula is:

[0019]

[0020] In the formula, represents the length of the i-th initial key edge; represents the three-dimensional coordinates of one of the initial key points of the i-th initial key edge; represents the three-dimensional coordinates of the other initial key point of the i-th initial key edge;

[0021] S25. Preset the tolerance range , and its expression is:

[0022]

[0023] In the formula, represents the tolerance range of the i-th initial key edge; represents the actual length of the limb corresponding to the i-th initial key edge; represents the tolerance percentage;

[0024] S26. Judge whether the length is within ;

[0025] If so, it means that there is no error in the initial key point;

[0026] If not, it means that there is an error in the initial key point.

[0027] Further, in step S23, it specifically includes the following steps:

[0028] S231. Select any one initial key point and obtain its positions in the two pieces of first-image data and ;

[0029] S232. According to and construct the first linear equations, and their expressions are respectively:

[0030]

[0031]

[0032] Simplify them to:

[0033]

[0034]

[0035] where ;

[0036] S233. Define the normalized coordinates , and its expression is:

[0037]

[0038]

[0039] S234. According to the normalized coordinates construct the second linear equations, and its expression is:

[0040]

[0041]

[0042] Expand it to:

[0043]

[0044]

[0045] S235. Combine and to obtain the third linear equations, and its expression is:

[0046]

[0047] where

[0048]

[0049]

[0050]

[0051] S236. Solve the third linear equation set by the least square method to obtain the three-dimensional coordinates of the initial key points. .

[0052] Furthermore, in step S3, it specifically includes the following steps:

[0053] S31. Divide the initial key points into the first initial key points and the second initial key points;

[0054] S32. Use the edge detection algorithm to extract the actual limb direction vector of the elderly , ;

[0055] S33. Calculate the estimated direction vector of the initial key edge , and its calculation formula is:

[0056]

[0057] In the formula, represents the estimated direction vector of the initial key edge between the i-th and j-th initial key points; and respectively represent the three-dimensional coordinates of the i-th and j-th initial key points;

[0058] S34. Calculate the included angle between the actual limb direction vector and the estimated direction vector , and its calculation formula is:

[0059]

[0060]

[0061] S35. Calculate the offset influence degree of the key point in each direction, and its calculation formula is:

[0062]

[0063]

[0064]

[0065] In the formula, , and respectively represent the offset influence degrees of the i-th key point on the x, y, and z axes; represents the length of the initial key edge between the i-th and j-th initial key points; represents the actual length of the elderly limb corresponding to the initial key edge between the i-th and j-th initial key points;

[0066] S36. Correct the three-dimensional coordinates of the second initial key points according to the offset influence degree ( ), and its calculation formula is:

[0067]

[0068]

[0069]

[0070] S37. Repeat the above steps until all the second initial key points are corrected. Denote all the key points as determined key points, denote the corrected initial key edges as determined key edges, and construct the second image data according to them.

[0071] Further, in step S4, it specifically includes the following steps:

[0072] S41. Calculate the limb angle change score of the elderly , and its calculation formula is:

[0073]

[0074] In the formula, represents the limb angle change score indicating the degree of angle change of all the determined key edges of the elderly; represents the total number of determined key edges; represents the weight coefficient for ; represents the angle between the i-th determined key edge and the vertical direction at time t; represents the angle between the i-th determined key edge and the vertical direction at time ; represents the time interval between two adjacent frames of the first image data;

[0075] S42. Calculate the speed change score of the elderly , and its calculation formula is:

[0076]

[0077] In the formula, represents the speed change score indicating the degree of speed change of all the determined key points of the elderly; represents the speed of the i-th determined key point at time t; represents the speed of the i-th determined key edge at time ; represents the weight coefficient for ;

[0078] S43. According to the limb angle change score and speed change score Calculate the fall risk value ;

[0079] S44. Evaluate the current physical health of the elderly based on their heart rate records and blood pressure records ;

[0080] S45. Calculate the comprehensive risk degree value based on the fall risk value and the physical health level .

[0081] Furthermore, in step S43, the formula for calculating the fall risk value is as follows:

[0082]

[0083] In the formula, represents the probability of the elderly falling; , and respectively represent the first, second, and third weight parameters regarding the fall risk value.

[0084] Furthermore, in step S44, it specifically includes the following steps:

[0085] S441. Obtain the blood pressure records and heart rate records of several test elderly people within the time period T as training samples;

[0086] S442. Calculate the physical condition score of each test elderly person on the day after time T , and use this as the sample label;

[0087] S443. Use the training samples and sample labels to train the physical condition evaluation model to obtain the target model;

[0088] S444. Input the heart rate record and blood pressure record, and output the physical health of the elderly .

[0089] Furthermore, in step S442, it specifically includes the following steps:

[0090] S4421. Calculate the heart rate deviation score , and its calculation formula is:

[0091]

[0092] In the formula, represents the weight coefficient regarding the heart rate; represents the heart rate of the test elderly person on the day after T; represents the normal heart rate level value; is the standard deviation of heart rate;

[0093] S4422. Calculate the systolic blood pressure deviation score and the diastolic blood pressure deviation score , and its calculation formula is:

[0094]

[0095]

[0096] In the formula, and respectively represent the weight coefficients for systolic blood pressure and diastolic blood pressure; and respectively represent the systolic blood pressure and diastolic blood pressure of the tested elderly on the day after T; and respectively represent the normal systolic blood pressure and diastolic blood pressure level values; and are the standard deviations of systolic blood pressure and diastolic blood pressure;

[0097] S4423. According to the heart rate deviation score , the systolic blood pressure deviation score and the diastolic blood pressure deviation score calculate the physical condition score , and its calculation formula is:

[0098]

[0099] In the formula, represents the physical condition score of the tested elderly on the day after T.

[0100] Furthermore, in step S45, it specifically includes the following steps:

[0101] S451. Calculate the average heart rate deviation score of the elderly after falling , and its calculation formula is:

[0102]

[0103] In the formula, represents the average heart rate of the elderly after falling; represents the weight parameter for the average heart rate deviation score;

[0104] S452. Calculate the average systolic blood pressure deviation score of the elderly after falling , and its calculation formula is:

[0105]

[0106] In the formula, represents the average systolic blood pressure of the elderly after a fall; represents the weight parameter regarding the deviation score of the average systolic blood pressure;

[0107] S453. Calculate the deviation score of the average diastolic blood pressure of the elderly after a fall , and its calculation formula is:

[0108]

[0109] In the formula, represents the average diastolic blood pressure of the elderly after a fall; represents the weight parameter regarding the deviation score of the average diastolic blood pressure;

[0110] S454. Calculate the comprehensive risk degree value .

[0111] Furthermore, in step S454, the calculation formula of the comprehensive risk degree value is:

[0112]

[0113] In the formula, , , , and respectively represent the first, second, third, fourth, and fifth weight parameters regarding the comprehensive risk degree value.

[0114] Compared with the prior art, the present invention provides a method for analyzing the state of the elderly based on image recognition, having the following beneficial effects:

[0115] 1. The present invention takes into account that when the elderly fall, not only will the angle of the limbs change significantly, but also their moving speed will suddenly change. Therefore, the calculation of the fall risk value is more accurate compared to other calculation methods. And when calculating the comprehensive risk degree value, considering that the fall of the elderly may cause changes in heart rate and blood pressure, the calculation accuracy of the comprehensive risk degree value is further improved, so that medical staff can take corresponding treatment measures in a timely manner according to the comprehensive risk degree value.

[0116] 2. When detecting the limb position and movement of the elderly, the present invention takes into account that thick clothing will affect the detection accuracy of the openpose algorithm, thus affecting the later judgment of the state of the elderly. Therefore, by introducing image processing technology to extract the actual limb direction and considering the angle difference for correction, the actual posture of the human body can be more accurately reflected. This method not only utilizes the length information provided by the three-dimensional coordinates but also combines the direction information in the two-dimensional image, thereby providing a more comprehensive pose estimation.

[0117] 3. Based on the first image data obtained by the monitoring devices from two angles, the present invention calculates the three-dimensional coordinates of each initial key point in the world coordinate system, which facilitates calculating the distances between the initial key points in the later stage and thereby evaluating whether there are deviations in the initial key points. BRIEF DESCRIPTION OF THE DRAWINGS

[0118] The drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0119] Figure 1 is a flowchart of the method for analyzing the state of the elderly based on image recognition according to the present invention;

[0120] Figure 2 is a schematic diagram of the initial key points of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0121] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments. Thereby, the implementation process of how the present application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0122] Those of ordinary skill in the art can understand that all or part of the steps in the following embodiments can be completed by instructing relevant hardware through a program. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0123] Nursing homes are equipped with professional medical staff and caregivers, who can provide comprehensive nursing services for the elderly, including daily life care, health monitoring, drug management, etc. This is particularly important for those elderly people who require special medical attention. However, the number of caregivers in nursing homes is limited. When the elderly move freely in the nursing home, it is impossible to ensure real-time monitoring of the status of each elderly person. Therefore, if an elderly person suddenly falls ill or trips and falls in the nursing home, the elderly person cannot receive timely care and treatment. For this reason, as Figure 1 shown, the present invention proposes a method for analyzing the state of the elderly based on image recognition, including the following steps:

[0124] S1. Obtain the first image data of the elderly in real time from multiple angles and the physiological data of the elderly including body size data, heart rate records, and blood pressure records. Specifically, in nursing homes, multiple monitoring devices are generally set up to obtain photos of each elderly person from multiple angles (at least two angles) in real time. The image data that can identify the elderly can be extracted through a convolutional neural network model. In addition, the body size data, heart rate records, and blood pressure records can be obtained through daily measurements. It should be noted that the convolutional neural network belongs to common existing technologies and will not be elaborated here.

[0125] S2. Identify the identity of the elderly, extract the initial key points in the first image data, and determine whether there are errors.

[0126] If so, go to step S3.

[0127] If not, mark the initial key points as determined key points, mark the first image data as the second image data, and go to step S4. Specifically, the openpose algorithm and the face recognition algorithm are adopted. The face recognition algorithm can directly identify the identity of the elderly to facilitate matching the corresponding physiological data of the elderly. The openpose algorithm can extract the initial key points in the first image data. In the present invention, each initial key point, as Figure 2 shown, includes 13 initial key points, namely the head initial key point, neck initial key point, left shoulder initial key point, right shoulder initial key point, left elbow initial key point, right elbow initial key point, left wrist initial key point, right wrist initial key point, hip initial key point, left knee initial key point, right knee initial key point, left ankle initial key point, and right ankle initial key point. Since the elderly are more afraid of cold than the young, they generally wear more and thicker clothes in winter, which will affect the detection accuracy of the initial key points and thus affect the later analysis of the elderly's state. It should be noted that both the openpose algorithm and the face recognition algorithm belong to existing technologies and will not be elaborated here. Therefore, in step S2, it specifically includes the following steps:

[0128] S21. Construct the initial key edges for representing the bones and torso of the elderly according to the initial key points. Specifically, the initial key edges, as Figure 2 shown, include the head and neck initial key edge, left shoulder and neck initial key edge, right shoulder and neck initial key edge, left shoulder and upper arm key edge, right shoulder and upper arm key edge, left forearm initial key edge, right forearm initial key edge, torso initial key edge, left thigh initial key edge, right thigh initial key edge, left calf initial key edge, and right calf initial key edge, a total of 12 initial key edges.

[0129] S22. Obtain the calibration parameters of the monitoring device including the focal length, principal point coordinates, rotation matrix, and translation vector of the monitoring device. Specifically, the expression of the projection equation of the monitoring device is:

[0130]

[0131] Among them, represents the two-dimensional coordinates of the i-th initial key point in an image data; represents the three-dimensional coordinates of the i-th initial key point; represents the internal parameter matrix of the monitoring device; represents the rotation matrix of the monitoring device; represents the translation vector of the monitoring device;

[0132]

[0133] Among them, and respectively represent the focal lengths of the monitoring device on the x-axis and y-axis; and respectively represent the coordinates of the principal point of the monitoring device in the x-axis and y-axis directions;

[0134]

[0135] Among them, represents the rotation amount of the monitoring device relative to the world coordinate system; represents the translation component of the monitoring device relative to the world coordinate system.

[0136] S23. Construct a world coordinate system with one of the monitoring devices as the origin, select any two first image data, and calculate the three-dimensional coordinates of each initial key point according to the calibration parameters ; Specifically, since there is an angle between different monitoring devices, the actual coordinates of each initial key point can be calculated by triangulation. It should be noted that the acquisition time points of these two first image data are the same. Therefore, in step S23, it specifically includes the following steps:

[0137] S231. Select any one initial key point and obtain its positions in the two first image data and ;

[0138] S232. Construct a first linear equation system according to and , and its expressions are respectively:

[0139]

[0140]

[0141] Simplify it to:

[0142]

[0143]

[0144] Among them, ;

[0145] S233. Define the normalized coordinates , and its expression is:

[0146]

[0147]

[0148] S234. Construct the second linear equation system according to the normalized coordinates , and its expression is:

[0149]

[0150]

[0151] Expand it to:

[0152]

[0153]

[0154] S235. Combine and , and obtain the third linear equation system, and its expression is:

[0155]

[0156] Among them,

[0157]

[0158]

[0159]

[0160] S236. Solve the third linear equation system by the least square method to obtain the three-dimensional coordinates of the initial key points.

[0161] In step S23 of the present invention, according to the first image data obtained by the monitoring devices of two angles, then calculate the three-dimensional coordinates of each initial key point in the world coordinate system, which is convenient for calculating the distance between the initial key points later, and thereby evaluating whether there is a deviation in the initial key points.

[0162] S24. Calculate the length of each initial key edge , and its calculation formula is:

[0163]

[0164] Wherein, represents the length of the i-th initial key edge; represents the three-dimensional coordinates of one of the initial key points of the i-th initial key edge; represents the three-dimensional coordinates of the other initial key point of the i-th initial key edge;

[0165] S25. Preset tolerance range , and its expression is:

[0166]

[0167] Wherein, represents the tolerance range of the i-th initial key edge; represents the actual length of the limb corresponding to the i-th initial key edge; represents the tolerance percentage; in the present invention, is 5%;

[0168] S26. Determine the length Whether it is in ;

[0169] If so, it means that there is no error in the initial key point;

[0170] If not, it means that there is an error in the initial key point.

[0171] It should be noted that if there is no error in the initial key point, the initial key edge is also marked as a determined key edge.

[0172] S3. Correct the initial key points in the first image data to generate second image data; specifically, if there are large errors in the positions of the initial key points, then the lengths and angles of the initial key edges constructed based on the initial key points will also have large errors, so that the actions and postures of the elderly cannot be accurately obtained, affecting the accuracy of the subsequent analysis of the elderly's state. Therefore, in step S3, the following steps are specifically included:

[0173] S31. Divide the initial key points into first initial key points and second initial key points; specifically, since some initial key points are not easily blocked and can be directly photographed by the monitoring device, and usually have small errors, such as the head, wrists, and ankles, the initial key points corresponding to these parts can be used as the first initial key points, and other initial key points as the second initial key points. Therefore, the first initial key points include the head initial key point, the left wrist initial key point, the right wrist initial key point, the left ankle initial key point, and the right ankle initial key point, and the other initial key points are the initial key points to be corrected;

[0174] S32. Extract the actual limb direction vector of the elderly using an edge detection algorithm , ; where i and j respectively represent the i-th and j-th initial key points, and there is an initial key edge between the i-th and j-th key points; specifically, although the openpose algorithm is affected by clothing and causes offsets, its calculation of the general direction of the limb is still relatively accurate. Therefore, the general position of the elderly's limb can be determined according to the direction of the initial key edge, and then the actual limb direction of this position can be determined according to the edge detection algorithm, so as to facilitate the correction of the position of each initial key point in the later stage. It should be noted that the edge detection algorithm belongs to the prior art and will not be elaborated here;

[0175] S33. Calculate the estimated direction vector of the initial key edge , and its calculation formula is:

[0176]

[0177] In the formula, represents the estimated direction vector of the initial key edge between the i-th and j-th initial key points; and respectively represent the three-dimensional coordinates of the i-th and j-th initial key points;

[0178] S34. Calculate the angle between the actual limb direction vector and the estimated direction vector , and its calculation formula is:

[0179]

[0180]

[0181] S35. Calculate the offset influence degree of the key point in each direction, and its calculation formula is:

[0182]

[0183]

[0184]

[0185] In the formula, , and respectively represent the offset influence degrees of the i-th key point on the x, y, and z axes; represents the length of the initial key edge between the i-th and j-th initial key points; Represents the actual length of the elderly person's limb corresponding to the initial key edge between the $i$-th and $j$-th initial key points;

[0186] S36. Correct the three-dimensional coordinates of the second initial key point according to the offset influence degree ( ), and its calculation formula is:

[0187]

[0188]

[0189]

[0190] Specifically, the second initial key point can be corrected by the first initial key point. For example, the neck initial key point can be corrected according to the head initial key point, and then the left shoulder initial key point and the right shoulder initial key point can be corrected according to the neck initial key point. And to avoid reducing the accuracy, the first initial key point is preferably used to calculate the adjacent second initial key points. From Figure 2 the content, it can be seen that only the left shoulder initial key point, the right shoulder initial key point, and the hip initial key point are not adjacent to the first initial key point. Therefore, these three initial key points are calculated last.

[0191] S37. Repeat the above steps until all the second initial key points are corrected. Denote all the key points as determined key points, denote the corrected initial key edges as determined key edges, and construct the second image data according to them.

[0192] In the present invention, when detecting the limb position and movement of the elderly person, considering that thick clothing will affect the detection accuracy of the OpenPose algorithm, which in turn affects the later judgment of the elderly person's state, so by introducing image processing technology to extract the actual limb direction and considering the angle difference for correction, the actual posture of the human body can be more accurately reflected. This method not only utilizes the length information provided by the three-dimensional coordinates but also combines the direction information in the two-dimensional image, thereby providing a more comprehensive pose estimation.

[0193] S4. Analyze the current state of the elderly person according to the continuous second image data and the physiological data of the elderly person; specifically, the daily actions of the elderly person are generally slower than those of young people, so the acceleration of their limbs is generally smaller. When the elderly person suddenly falls ill or is tripped under the feet, etc., the acceleration of the limbs will increase significantly. This change will start from the legs and spread to the whole body. And the basic physical conditions of each elderly person are different, so when different elderly people have the same falling situation, their conditions may be different. Therefore, in step S4, it specifically includes the following steps:

[0194] S41. Calculate the limb angle change score of the elderly person , and its calculation formula is:

[0195]

[0196] In the formula, represents the limb angle change score indicating the degree of change in the angles of all the determined key edges of the elderly; represents the total number of determined key edges; represents with respect to the weight coefficient; represents the angle of the i-th determined key edge with respect to the vertical direction at time t; represents the angle of the i-th determined key edge with respect to the vertical direction at time t; represents the time interval between two adjacent frames of the first image data; in the present invention, is determined according to different limb positions, wherein, for the head and neck determined key edge, left shoulder and neck determined key edge, right shoulder and neck determined key edge, left upper arm key edge, right upper arm key edge, left lower arm determined key edge, and right lower arm determined key edge, the are all 0.1; for the trunk determined key edge, the is 0.3; for the left thigh determined key edge and right thigh determined key edge, the are all 0.5; for the left lower leg determined key edge and right lower leg determined key edge, the are all 0.6;

[0197] S42. Calculate the speed change score of the elderly , and its calculation formula is:

[0198]

[0199] In the formula, represents the speed change score indicating the degree of speed change of all the determined key points of the elderly; represents the speed of the i-th determined key point at time t; represents the speed of the i-th determined key edge at time t; represents with respect to the weight coefficient; wherein,

[0200] ;

[0201] Among them, represents the three-dimensional coordinates of the i-th determined key point at time t; represents the three-dimensional coordinates of the i-th determined key point at time t;

[0202] It should be noted that According to different positions, the initial key points of the head, neck, left shoulder, right shoulder, left elbow, right elbow, left wrist and right wrist are determined. All are 0.1; the initial key point of the hip is 0.3; the initial key point of the left knee and the initial key point of the right knee Both are 0.5; the initial key points of the left ankle and the right ankle Both are 0.6;

[0203] S43. Scoring based on limb angle changes and speed change score Calculate the risk of falling Specifically, when an elderly person falls, not only the angle of the limbs will change significantly, but also the speed of movement will change suddenly. Therefore, considering the angle change and speed change comprehensively can more effectively identify potential falls. For example, in the early stage of a fall, the speed change of the legs will appear before the angle change of the upper body, and then the angle change of the whole body will increase sharply. Therefore, in step S43, the fall risk value The calculation formula is:

[0204]

[0205] In the formula, Indicates the likelihood of the elderly falling; , and Respectively represent the first, second and third weight parameters about the fall risk value; In the present invention, , and They are 0.4, 0.5 and 0.1 respectively;

[0206] S44. Assess the current physical health of the elderly based on their heart rate and blood pressure records Specifically, since the heart rate record and blood pressure record can reflect the physical condition of the elderly, the common method for calculating the physical health of the elderly is to calculate directly based on the heart rate record and blood pressure record, and this method does not take into account the complex interactive relationship between the heart rate record and the blood pressure record. For example, under normal circumstances, an increase in heart rate will lead to an increase in cardiac output, which in turn leads to an increase in systolic blood pressure. However, under certain pathological conditions (such as heart failure), an increase in heart rate may not be accompanied by a significant increase in systolic blood pressure. Therefore, in step S44, the following steps are specifically included:

[0207] S441, obtaining blood pressure records and heart rate records of several test elderly people in a time period T as training samples;

[0208] S442. Calculate the physical condition score of each tested elderly person one day after time T , and use this as the sample label; specifically, in step S442, it specifically includes the following steps:

[0209] S4421. Calculate the heart rate deviation score , and its calculation formula is:

[0210]

[0211] In the formula, represents the weight coefficient regarding the heart rate; represents the heart rate of the tested elderly person one day after T; represents the normal heart rate level value; is the standard deviation of the heart rate; in the present invention, is 10; is 80; is 0.8;

[0212] S4422. Calculate the systolic blood pressure deviation score and the diastolic blood pressure deviation score , and its calculation formula is:

[0213]

[0214]

[0215] In the formula, and respectively represent the weight coefficients regarding the systolic blood pressure and the diastolic blood pressure; and respectively represent the systolic blood pressure and the diastolic blood pressure of the tested elderly person one day after T; and respectively represent the normal systolic blood pressure and diastolic blood pressure level values; and are the standard deviations of the systolic blood pressure and the diastolic blood pressure; in the present invention, and are 0.7 and 0.6 respectively; and are 115 and 70 respectively; and are 15 and 10 respectively;

[0216] S4423. According to the heart rate deviation score , the systolic blood pressure deviation score and the diastolic blood pressure deviation score calculate the physical condition score , and its calculation formula is:

[0217]

[0218] In the formula, represents the physical condition score of the tested elderly person one day after T.

[0219] S443. Train a physical condition evaluation model using training samples and sample labels to obtain a target model; specifically, the physical condition evaluation model is constructed based on LSTM. LSTM belongs to the prior art and will not be elaborated here.

[0220] S444. Input the heart rate record and blood pressure record, and output the physical health degree of the elderly person .

[0221] S45. Calculate the comprehensive risk degree value and the physical health degree ; specifically, when an elderly person falls, it may cause changes in heart rate and blood pressure. For example, falling may cause pain, anxiety, or physical stress reactions, which will all affect heart rate and blood pressure. Therefore, when evaluating the overall health risk of the elderly person, it is very important to consider the impact of falling on physiological parameters (such as heart rate and blood pressure). This interactive effect can more accurately reflect the true health status of the elderly person and help take necessary medical measures in a timely manner. For this reason, in step S45, it specifically includes the following steps: ; specifically, when an elderly person falls, it may cause changes in heart rate and blood pressure. For example, falling may cause pain, anxiety, or physical stress reactions, which will all affect heart rate and blood pressure. Therefore, when evaluating the overall health risk of the elderly person, it is very important to consider the impact of falling on physiological parameters (such as heart rate and blood pressure). This interactive effect can more accurately reflect the true health status of the elderly person and help take necessary medical measures in a timely manner. For this reason, in step S45, it specifically includes the following steps:

[0222] S451. Calculate the average heart rate deviation score of the elderly person after falling, and its calculation formula is:

[0223]

[0224] In the formula, represents the average heart rate of the elderly person after falling; represents the weight parameter regarding the average heart rate deviation score; in the present invention, ; is 0.8;

[0225] S452. Calculate the average systolic blood pressure deviation score of the elderly person after falling, and its calculation formula is:

[0226]

[0227] In the formula, represents the average systolic blood pressure of the elderly person after falling; represents the weight parameter regarding the average systolic blood pressure deviation score; in the present invention, ; is 0.75;

[0228] S453. Calculate the average diastolic blood pressure deviation score of the elderly after a fall , and its calculation formula is:

[0229]

[0230] In the formula, represents the average diastolic blood pressure of the elderly after a fall; represents the weight parameter for the average diastolic blood pressure deviation score; in the present invention, ; is 0.73;

[0231] S454. Calculate the comprehensive risk degree value ; specifically, in step S454, the calculation formula of the comprehensive risk degree value is:

[0232]

[0233] In the formula, , , , and respectively represent the first, second, third, fourth, and fifth weight parameters for the comprehensive risk degree value; in the present invention, , , , and are 0.4, 0.3, 0.1, 0.1, and 0.1 respectively.

[0234] In the present invention, when the elderly are about to fall, first calculate the limb angle change score and the speed change score , and then calculate the fall risk value according to them, and when calculating the fall risk value , considering that when the elderly fall, not only will the angle of the limbs change significantly, but their moving speed will also suddenly change, so the fall risk value has higher accuracy compared to other calculation methods. Then, when calculating the comprehensive risk degree value , considering that the fall of the elderly may cause changes in heart rate and blood pressure, so the calculation accuracy of the comprehensive risk degree value is further improved.

[0235] S5. Feed back the comprehensive risk degree value to the staff in real time; specifically, when the staff obtains the comprehensive risk degree value of the elderly, they can formulate a response plan according to the comprehensive risk degree value , for example, if the comprehensive risk degree value When the value is relatively high, directly call the emergency number, and have specialized personnel go to help the elderly with emergency rescue equipment.

[0236] The above embodiments have introduced the present invention in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An elderly state analysis method based on image recognition, characterized in that, Including the following steps: S1. Obtain the first image data of the elderly in real time from multiple angles, as well as the physiological data of the elderly including body size data, heart rate records, and blood pressure records; S2. Identify the identity of the elderly, extract the initial key points in the first image data, and determine whether there are errors; If so, proceed to step S3; If not, mark the first image data as the second image data and proceed to step S4; In step S2, it specifically includes the following steps: S21. Construct the initial key edges for representing the bones and torso of the elderly based on the initial key points; S22. Obtain the calibration parameters of the monitoring device including the focal length, principal point coordinates, rotation matrix, and translation vector of the monitoring device; S23. Construct a world coordinate system with one of the monitoring devices as the origin, select any two pieces of first image data, and calculate the three-dimensional coordinates of each initial key point according to the calibration parameters ; In step S23, it specifically includes the following steps: S231. Optionally select an initial key point and obtain its positions in the two first image data and ; S232. According to and construct a first linear equation system, and their expressions are respectively: ， ， Simplify it to: ， ， Among them, ; S233. Define the normalized coordinates , and its expression is: ， ； S234. According to the normalized coordinates Construct a second linear equation system, and its expression is: ， ， Expand it to: ， ； S235, Merge and , to obtain a third linear equation system, whose expression is: ， Wherein, ； ； S236. Solve the third linear equation system by the least square method to obtain the three-dimensional coordinates of the initial key points ; S24. Calculate the length of each initial critical edge , and its calculation formula is: ， In the formula, represents the length of the i-th initial critical edge; represents the three-dimensional coordinates of one of the initial key points of the i-th initial critical edge; represents the three-dimensional coordinates of the other initial key point of the i-th initial critical edge; S25. Preset tolerance range , and its expression is: ， In the formula, represents the tolerance range of the i-th initial critical edge; represents the actual length of the limb corresponding to the i-th initial critical edge; represents the tolerance percentage; S26. Determine the length Is it within ; If so, it means that there are no errors in the initial key points; If not, it means that there are errors in the initial key points; S3. Correct the initial key points in the first image data to generate the second image data; S4. Analyze the current state of the elderly based on the continuous second image data and the physiological data of the elderly; S5. Feed the comprehensive risk degree value back to the staff in real time.

2. The method for analyzing the state of the elderly based on image recognition according to claim 1, wherein In step S3, it specifically includes the following steps: S31. Divide the initial key points into the first initial key points and the second initial key points; S32. Extract the actual limb direction vector of the elderly using an edge detection algorithm , ; S33. Calculate the estimated direction vector of the initial critical edge , and its calculation formula is as follows: ， wherein, represents the estimated direction vector of the initial key edge between the i-th and j-th initial key points; and respectively represent the three-dimensional coordinates of the i-th and j-th initial key points; S34. Calculate the actual limb direction vector and the estimated direction vector to obtain the included angle , and its calculation formula is as follows: ， ； S35. Calculate the offset influence degree of the key points in each direction, and its calculation formula is: ， ， ， Wherein, , and respectively represent the offset influence degrees of the i-th key point on the x, y, and z axes; represents the length of the initial key edge between the i-th and j-th initial key points; represents the actual length of the old person's limb corresponding to the initial key edge between the i-th and j-th initial key points; S36. Correct the three-dimensional coordinates of the second initial key point according to the offset influence degree ( ), and its calculation formula is: ， ， ； S37. Repeat the above steps until all the second initial key points are corrected, mark all the key points as determined key points, mark the corrected initial key edges as determined key edges, and construct the second image data based on them.

3. The method for analyzing the state of the elderly based on image recognition according to claim 2, characterized in that, In step S4, it specifically includes the following steps: S41. Calculate the limb angle change score of the elderly , and its calculation formula is: ， In the formula, represents the limb angle change score indicating the degree of change in the angles of all the determined key edges of the elderly; represents the total number of determined key edges; represents with respect to the weight coefficient; represents the angle between the i-th determined key edge and the vertical direction at time t; represents the i-th determined key edge at time and the angle with the vertical direction; represents the time interval between two adjacent frames of the first image data; S42. Calculate the speed change score of the elderly , and its calculation formula is: ， In the formula, represents the speed change score indicating the degree of speed change of all determined key points of the elderly; represents the speed of the i-th determined key point at time t; represents the i-th determined key point at time when the speed; represents with respect to the weight coefficient; S43. Calculate the fall risk value based on the limb angle change score and the speed change score ; ; S44. Evaluate the current physical health of the elderly based on their heart rate records and blood pressure records ; S45. Calculate the comprehensive risk degree value based on the fall risk value and the physical health level .

4. The method for analyzing the state of the elderly based on image recognition according to claim 3, wherein In step S43, the fall risk value is calculated by the formula: ， In the formula, represents the falling probability of the elderly; , and respectively represent the first, second, and third weight parameters regarding the falling risk value.

5. The method for analyzing the state of the elderly based on image recognition according to claim 3, wherein In step S44, it specifically includes the following steps: S441. Obtain the blood pressure records and heart rate records of several test elderly people within the time period T as training samples; S442. Calculate the physical condition score of each tested elderly person one day after time T , and use this as the sample label; S443. Use the training samples and sample labels to train the physical condition assessment model to obtain the target model; S444. Input the heart rate record and blood pressure record and output the physical health level of the elderly .

6. The method for analyzing the status of the elderly based on image recognition according to claim 5, characterized in that, In step S442, it specifically includes the following steps: S4421. Calculate the heart rate deviation score , and its calculation formula is: ， In the formula, represents the weight coefficient regarding the heart rate; represents the heart rate of the tested elderly person one day after T; represents the normal heart rate level value; is the standard deviation of the heart rate; S4422. Calculate the systolic blood pressure deviation score and the diastolic blood pressure deviation score , and their calculation formulas are as follows: ， ， Wherein, and respectively represent the weight coefficients for systolic blood pressure and diastolic blood pressure; and respectively represent the systolic blood pressure and diastolic blood pressure of the tested elderly on the day after T; and respectively represent the normal systolic blood pressure and diastolic blood pressure level values; and The standard deviations of systolic blood pressure and diastolic blood pressure; S4423. Calculate the physical condition score based on the heart rate deviation score , the systolic blood pressure deviation score and the diastolic blood pressure deviation score , and the calculation formula is: , and its calculation formula is: ， In the formula, represents the physical condition score of the tested elderly person on the day after T.

7. The method for analyzing the state of the elderly based on image recognition according to claim 3, wherein In step S45, it specifically includes the following steps: S451. Calculate the average heart rate deviation score of the elderly after a fall , and its calculation formula is: ， wherein, represents the average heart rate of the elderly after falling; represents the weight parameter for the deviation score of the average heart rate; S452. Calculate the average systolic blood pressure deviation score of the elderly after a fall , and its calculation formula is: ， In the formula, represents the average systolic blood pressure of the elderly after falling; represents the weight parameter for the deviation score of the average systolic blood pressure; S453. Calculate the average diastolic blood pressure deviation score of the elderly after a fall , and its calculation formula is: ， In the formula, represents the average diastolic blood pressure of the elderly after falling; represents the weight parameter for the deviation score of the average diastolic blood pressure; S454. Calculate the comprehensive risk value .

8. The method for analyzing the state of the elderly based on image recognition according to claim 7, wherein In step S454, the comprehensive risk degree value is calculated according to the formula: ， In the formula, and respectively represent the first, second, third, fourth, and fifth weight parameters for the comprehensive risk degree value.

Images