Personnel tumble detection method and system based on machine vision

Abstract

The invention provides a person falling detection method and system based on machine vision, and relates to the technical field of machine vision, and the method comprises the steps: carrying out the human skeleton key point recognition processing of a to-be-detected person in a monitoring image, so as to obtain the coordinate information of a plurality of human skeleton key points; calculating a human body inclination angle, a human body aspect ratio and a human body center included angle according to the coordinate information of the plurality of human body skeleton key points; judging whether the to-be-detected person is in a falling posture according to the human body inclination angle, the human body aspect ratio and the human body center included angle; and when judging that the to-be-detected person is in the tumble posture, outputting a tumble alarm message, and comprehensively judging whether the to-be-detected person is in the tumble posture or not based on the human body inclination angle, the human body aspect ratio and the human body center included angle. On one hand, the accuracy of personnel falling detection based on machine vision is effectively improved, and on the other hand, the monitoring precision and the monitoring efficiency of the monitoring end are also improved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision, and in particular to a method and system for detecting a person falling based on machine vision. Background Art

[0002] With the continuous development of machine vision, the research on human behavior analysis has also shifted from long-term camera shooting and wearable sensor equipment research to a more efficient, lower-cost, and more accurate recognition combined with machine learning and deep learning research. Currently, researchers' research on human behavior detection is also mainly concentrated in machine vision.

[0003] In the prior art, there are two main methods for fall detection algorithms in the field of computer vision: threshold method and image processing. The threshold method relies on the extraction of the target's fall behavior features, such as the falling amplitude and speed of the knee joint node, the angle change of the body, etc. In current research, such features are generally used in image segmentation algorithms, which extract such features and then make a fall judgment, but the disadvantage is that the accuracy is low.

[0004] In the field of machine vision, the use of image processing technology to solve the problem of human fall detection is divided into human fall detection based on machine learning and human fall detection based on deep learning. Fall detection based on deep learning is to use a large amount of learning of a deep learning algorithm model in deep learning, input a large number of pictures and videos of human falls, and extract the pixel-level image features of people falling in each frame of the image according to the deep learning network, and then make a fall judgment based on the extracted features. The disadvantage is that the features of people falling in each frame of the image only account for a small part of the full image, which will generate a large amount of redundant content, which is a heavy burden on the processor and requires a high computing power of the computer.

[0005] Image processing technology can be used to extract the behavioral posture features of the human body. A series of machine learning algorithms are used to determine falls. This is the main method of fall detection based on machine learning. The disadvantage is that the accuracy of fall detection basically depends on the extracted feature values, and environmental factors (such as lighting, whether indoors, etc.) will greatly affect the extraction of the feature when extracting the feature, making the detection accuracy worrying. Therefore, how to effectively improve the accuracy of human fall detection based on machine vision under the premise of limited computer computing power has become an urgent problem to be solved.

[0006] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0007] The main purpose of the present invention is to provide a method and system for detecting human falls based on machine vision, aiming to solve the technical problem that the existing technology cannot effectively improve the accuracy of human fall detection based on machine vision under the premise of limited computer computing power.

[0008] The technical solution adopted by the present invention is as follows: a method for detecting a person falling based on machine vision, the method comprising the following steps:

[0009] Perform human skeleton key point recognition processing on the person to be detected in the monitoring image to obtain coordinate information of multiple human skeleton key points; wherein the human skeleton key points include human shoulder key points, human hip key points, human target frame, and human foot key points;

[0010] Calculate the human body inclination angle, human body aspect ratio and human body center angle according to the coordinate information of multiple key points of human skeleton;

[0011] Determine whether the person to be detected is in a falling posture according to the body inclination angle, the body width-to-height ratio and the body center angle;

[0012] When it is determined that the person to be detected is in the falling posture, a fall warning message is output.

[0013] Optionally, the step of calculating the human body inclination angle, the human body aspect ratio and the human body center angle according to the coordinate information of a plurality of human body skeleton key points comprises:

[0014] Calculating the human body inclination angle according to the coordinate information of the human body shoulder key point and the coordinate information of the human body hip key point;

[0015] Calculate the human body aspect ratio according to the coordinate information of the human body target frame;

[0016] The human body center angle is calculated according to the coordinate information of the key points of the human body shoulder, the coordinate information of the key points of the human body hip and the coordinate information of the key points of the human body foot.

[0017] Optionally, the step of calculating the human body inclination angle according to the coordinate information of the human body shoulder key point and the coordinate information of the human body hip key point comprises:

[0018] According to the coordinate information of the key points of the human body shoulder and the coordinate information of the key points of the human body hip, the human body inclination angle is calculated by the following formula;

[0019]

[0020] Where (x5, y5) is the coordinate of the key point of the left shoulder of the human body, (x6, y6) is the coordinate of the key point of the right shoulder of the human body, (x P ,yP ) is the coordinate of the center point P of the left shoulder key point and the right shoulder key point of the human body, (x 11 ,y 11 ) is the coordinate of the left hip key point of the human body, (x 12 ,y 12 ) is the coordinate of the key point of the right hip of the human body, (x Q ,y Q ) is the coordinate of the center point Q of the left hip key point and the right hip key point of the human body, θ PQ is the body inclination angle;

[0021] The step of calculating the human body aspect ratio according to the coordinate information of the human body target frame comprises:

[0022] According to the coordinate information of the human body target frame, the human body aspect ratio is calculated by the following formula;

[0023]

[0024] Wherein, max(x) is the maximum horizontal coordinate of the human body target frame, min(x) is the minimum horizontal coordinate of the human body target frame, dx is the distance between the maximum horizontal coordinate and the minimum horizontal coordinate, max(y) is the maximum vertical coordinate of the human body target frame, min(y) is the minimum vertical coordinate of the human body target frame, dy is the distance between the maximum vertical coordinate and the minimum vertical coordinate, and R is the human body aspect ratio;

[0025] The step of calculating the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point comprises:

[0026] According to the coordinate information of the key points of the human body shoulder, the coordinate information of the key points of the human body hip and the coordinate information of the key points of the human body foot, the human body center angle is calculated by the following formula;

[0027]

[0028] In the formula, θ PQF is the human body center angle, P is the coordinate of the center point of the left shoulder key point and the right shoulder key point, Q is the coordinate of the center point of the left hip key point and the right hip key point, and F is the coordinate of the left ankle key point or the right ankle key point.

[0029] Optionally, the step of judging whether the person to be detected is in a falling posture according to the body inclination angle, the body aspect ratio and the body center angle includes:

[0030] When the body inclination angle is greater than or equal to a preset body inclination angle threshold, the body width-to-height ratio is greater than or equal to a preset body width-to-height ratio threshold, and the body center angle is greater than or equal to a preset body center angle threshold, it is determined that the person to be detected is in a falling posture;

[0031] Otherwise, it is determined that the person to be detected is not in a falling posture.

[0032] Optionally, when it is determined that the person to be detected is in the falling posture, the step of outputting a fall warning message includes:

[0033] When it is determined that the person to be detected is in the falling posture, calculating the falling time of the person to be detected in the falling posture;

[0034] When the fall duration is greater than or equal to the preset fall duration, a fall alarm message is output.

[0035] In addition, to achieve the above-mentioned purpose, the present invention also proposes a personnel fall detection system based on machine vision, and the personnel fall detection system based on machine vision comprises:

[0036] The key point recognition module is used to perform human skeleton key point recognition processing on the person to be detected in the monitoring picture to obtain the coordinate information of multiple human skeleton key points;

[0037] A parameter calculation module is used to calculate the human body inclination angle, human body aspect ratio and human body center angle according to the coordinate information of multiple key points of human skeleton;

[0038] A posture judgment module, used to judge whether the person to be detected is in a falling posture according to the body inclination angle, the body width-to-height ratio and the body center angle;

[0039] The fall alarm module is used to output a fall alarm message when it is determined that the person to be detected is in the fall posture.

[0040] Optionally, the coordinate information of the plurality of key points of the human skeleton includes the coordinate information of the key points of the human shoulder, the coordinate information of the key points of the human hip, the coordinate information of the human target frame, and the coordinate information of the key points of the human foot, and the parameter calculation module is further used to calculate the human body inclination angle according to the coordinate information of the key points of the human shoulder and the coordinate information of the key points of the human hip;

[0041] The parameter calculation module is further used to calculate the human body aspect ratio according to the coordinate information of the human body target frame;

[0042] The parameter calculation module is also used to calculate the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point.

[0043] Optionally, the parameter calculation module is further used to calculate the human body inclination angle according to the coordinate information of the human body shoulder key point and the coordinate information of the human body hip key point by the following formula;

[0044]

[0045] Where (x5, y5) is the coordinate of the key point of the left shoulder of the human body, (x6, y6) is the coordinate of the key point of the right shoulder of the human body, (x P ,y P ) is the coordinate of the center point P of the left shoulder key point and the right shoulder key point of the human body, (x 11 ,y 11 ) is the coordinate of the left hip key point of the human body, (x 12 ,y 12 ) is the coordinate of the key point of the right hip of the human body, (x Q ,y Q ) is the coordinate of the center point Q of the left hip key point and the right hip key point of the human body, θ PQ is the body inclination angle;

[0046] The parameter calculation module is further used to calculate the human body aspect ratio according to the coordinate information of the human body target frame by the following formula;

[0047]

[0048]

[0049] Wherein, max(x) is the maximum horizontal coordinate of the human body target frame, min(x) is the minimum horizontal coordinate of the human body target frame, dx is the distance between the maximum horizontal coordinate and the minimum horizontal coordinate, max(y) is the maximum vertical coordinate of the human body target frame, min(y) is the minimum vertical coordinate of the human body target frame, du is the distance between the maximum vertical coordinate and the minimum vertical coordinate, and R is the human body aspect ratio;

[0050] The parameter calculation module is further used to calculate the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point by the following formula;

[0051]

[0052] In the formula, θ PQF is the human body center angle, P is the coordinate of the center point of the left shoulder key point and the right shoulder key point, Q is the coordinate of the center point of the left hip key point and the right hip key point, and F is the coordinate of the left ankle key point or the right ankle key point.

[0053] Optionally, the posture judgment module is also used to determine that the person to be detected is in a falling posture when the body inclination angle is greater than or equal to a preset body inclination angle threshold, the body aspect ratio is greater than or equal to a preset body aspect ratio threshold, and the body center angle is greater than or equal to a preset body center angle threshold; otherwise, it is determined that the person to be detected is not in a falling posture.

[0054] Optionally, the fall alarm module is further configured to calculate the fall duration of the person to be detected in the fall posture when it is determined that the person to be detected is in the fall posture;

[0055] The fall alarm module is further configured to output a fall alarm message when the fall duration is greater than or equal to a preset fall duration.

[0056] Compared with the prior art, the beneficial effects of the above technical solution are as follows: the present invention performs human skeleton key point recognition processing on the person to be detected in the monitoring screen to obtain the coordinate information of multiple human skeleton key points; calculates the human body inclination angle, human body aspect ratio and human body center angle according to the coordinate information of multiple human body skeleton key points; judges whether the person to be detected is in a falling posture according to the human body inclination angle, human body aspect ratio and human body center angle; when it is judged that the person to be detected is in the falling posture, outputs a fall alarm message. Compared with the processor burden and high computing power requirements of the computer caused by inputting a large number of pictures and videos of human falls in deep learning and performing fall judgment based on the pixel-level image features of people falling in each frame of the image extracted, the present invention performs the above fall detection based on the lightweight YOLOv8-Pose algorithm, realizes the effective improvement of the accuracy of personnel fall detection based on machine vision under the premise of limited computer computing power, and further, comprehensively judges whether the person to be detected is in a falling posture based on the human body inclination angle, human body aspect ratio and human body center angle, effectively improving the accuracy of fall detection. In addition, since comprehensive emergency dispatch and command only need to be carried out in a timely manner when the system alarms, the labor intensity is effectively reduced and the monitoring effectiveness is improved. On the other hand, it also effectively makes up for the weaknesses of human eye monitoring and improves the accuracy of monitoring. By real-time analysis of the posture of the person to be detected in the monitoring screen, it automatically determines whether it has fallen and issues an alarm, which is conducive to the timely prevention and control of hazardous events and improves the monitoring effectiveness and management efficiency of the monitoring end. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of a first embodiment of a method for detecting a person falling down based on machine vision according to the present invention;

[0058] Figure 2A schematic diagram of key points of human skeleton detected by using the YOLOv8-Pose algorithm according to the first embodiment of the method for detecting a person falling based on machine vision of the present invention;

[0059] Figure 3 Schematic diagram of the human body inclination angle involved in the first embodiment of the method for detecting a person falling based on machine vision of the present invention;

[0060] Figure 4 Schematic diagram of the human body width-to-height ratio involved in the first embodiment of the method for detecting a person falling based on machine vision of the present invention;

[0061] Figure 5 Schematic diagram of the human body center angle involved in the first embodiment of the method for detecting a person falling based on machine vision of the present invention;

[0062] Figure 6 This is a human body tilt angle effect diagram involved in the first embodiment of the human fall detection method based on machine vision of the present invention;

[0063] Figure 7 This is a human body aspect ratio effect diagram involved in the first embodiment of the human fall detection method based on machine vision of the present invention;

[0064] Figure 8 The diagram is a before-and-after effect diagram of introducing the human body center angle involved in the first embodiment of the human fall detection method based on machine vision of the present invention;

[0065] Fig. 9 This is a rendering of the human body center angle involved in the first embodiment of the human fall detection method based on machine vision of the present invention;

[0066] Fig.10 This is a fall detection effect diagram involved in the first embodiment of the method for detecting a person falling based on machine vision of the present invention;

[0067] Fig.11 A technical roadmap for the first embodiment of the method for detecting a person falling down based on machine vision according to the present invention;

[0068] Fig.12 It is a structural block diagram of the first embodiment of the personnel fall detection system based on machine vision of the present invention. DETAILED DESCRIPTION

[0069] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar modules or modules with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be construed as limitations on the present application. On the contrary, the embodiments of the present application include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.

[0070] The embodiment of the present invention provides a method for detecting a person falling down based on machine vision. Figure 1 , Figure 1 The figure is a flow chart of the first embodiment of the method for detecting a person falling down based on machine vision according to the present invention.

[0071] Step S10: performing human skeleton key point recognition processing on the person to be detected in the monitoring image to obtain coordinate information of multiple human skeleton key points;

[0072] It should be noted that, in order to improve the detection accuracy, this embodiment adopts the YOLOv8-Pose algorithm for personnel detection. YOLOv8-Pose is an algorithm for posture detection. Its principle is to introduce a posture estimation module on the basis of the YOLOv8 target detection algorithm to achieve accurate posture detection by combining detection and posture estimation. Among them, the posture estimation module is usually composed of a convolution layer, a pooling layer and a fully connected layer, which is used to extract features and predict the position of key points. YOLOv8 is a target detection algorithm with high speed and high accuracy proposed by Glenn-Jocher. By combining YOLOv8 with a posture estimation module, accurate detection of human posture can be achieved while maintaining a high detection speed. The basic idea of ​​the YOLOv8-Pose algorithm is to convert the posture detection task into a detection task of multiple key points, and to achieve detection and positioning of these key points by adding an additional key point detection layer on the basis of YOLOv8.

[0073] In this embodiment, the YOLOv8 model can be used to determine whether there is a person to be detected in the monitoring image. When there is a person to be detected, the human skeleton key points of the person to be detected are identified based on the YOLOv8-Pose model, and then the coordinate information of multiple human skeleton key points is obtained, where the human skeleton key points include the human shoulder key points, the human hip key points, the human target frame, and the human foot key points. Figure 2 , Figure 2 This is a schematic diagram of key points of human skeleton detected using the YOLOv8-Pose algorithm involved in the first embodiment of the method for detecting a person falling based on machine vision of the present invention. Figure 2In the method, the YOLOv8-Pose algorithm is used to detect the human body posture point data. After the target (i.e., the person to be detected) is detected by the YOLOv8-Pose algorithm, the algorithm will output 17 key points of the human skeleton. Referring to Table 1, Table 1 is a table of the numbering of the key points of the human skeleton detected by the YOLOv8-Pose algorithm according to the first embodiment of the method for detecting a person falling based on machine vision of the present invention. Table 1 shows the names of the 17 key points of the human skeleton detected by the YOLOv8-Pose algorithm, and the numbers in Table 1 are the same as those in Table 1. Figure 2 The key points of the human skeleton are matched.

[0074] Table 1. Key point numbering table of human skeleton detected using YOLOv8-Pose algorithm

[0075]

[0076]

[0077] Step S20: Calculate the human body inclination angle, human body aspect ratio and human body center angle according to the coordinate information of multiple human body skeleton key points;

[0078] It is known from prior knowledge of the human body that when the human body is in daily activities or falls, the angle between the center line of the human body and the vertical line of the center point of the human body changes to a certain extent. This embodiment uses the human body tilt angle as one of the characteristics of the fall behavior. Figure 3 , Figure 3 Schematic diagram of the human body inclination angle involved in the first embodiment of the human fall detection method based on machine vision of the present invention. Figure 3 In the figure, 5 is defined as the left shoulder key point, 6 is defined as the right shoulder key point, the center point of the left shoulder key point 5 and the right shoulder key point 6 is P, and the perpendicular line through the center point P is the perpendicular line of the human body center line, denoted as M; 11 is defined as the left hip key point, 12 is defined as the right hip key point, the center point of the left hip key point 11 and the right hip key point 12 is Q, and the line segment through the center points PQ is the human body center line, denoted as N.

[0079] In a specific implementation, the coordinate information of the multiple key points of the human skeleton includes but is not limited to the coordinate information of the key points of the human shoulder, the coordinate information of the key points of the human hip, the coordinate information of the human target frame, and the coordinate information of the key points of the human foot. Thus, the human body inclination angle can be calculated according to the coordinate information of the key points of the human shoulder and the coordinate information of the key points of the human hip. For example, the human body inclination angle can be calculated according to the coordinate information of the key points of the human shoulder and the coordinate information of the key points of the human hip in the coordinate information of the multiple key points of the human skeleton by the following formula;

[0080]

[0081] Where (x5, y5) is the coordinate of the key point of the left shoulder of the human body, (x6, y6) is the coordinate of the key point of the right shoulder of the human body, (x P ,y P ) is the coordinate of the center point P of the left shoulder key point and the right shoulder key point of the human body, (x 11 ,y 11 ) is the coordinate of the left hip key point of the human body, (x 12 ,y 12 ) is the coordinate of the key point of the right hip of the human body, (x Q ,y Q ) is the coordinate of the center point Q of the left hip key point and the right hip key point of the human body, θ PQ is the inclination angle of the human body.

[0082] When a person falls, the person's posture will change dramatically in a short period of time, and the positions of key points of the person will also change accordingly. The aspect ratio of the person's daily activities is different from the aspect ratio of the person's fall, so the aspect ratio of the person's body can be used to describe a fall. Figure 4 , Figure 4 Schematic diagram of the human body width-to-height ratio involved in the first embodiment of the human fall detection method based on machine vision of the present invention. Figure 4 In the method, the human body detection target frame is used to determine the human body width and human body height, so as to further determine the human body width and human body height ratio based on the human body width and human body height. In a specific implementation, the human body width and height ratio can be calculated according to the coordinate information of the human body target frame. For example, the human body width and height ratio can be calculated according to the coordinate information of the human body target frame by the following formula;

[0083]

[0084] Wherein, max(x) is the maximum horizontal coordinate of the human target frame, min(x) is the minimum horizontal coordinate of the human target frame, dx is the distance between the maximum horizontal coordinate and the minimum horizontal coordinate, that is, the human width, max(y) is the maximum vertical coordinate of the human target frame, min(y) is the minimum vertical coordinate of the human target frame, dy is the distance between the maximum vertical coordinate and the minimum vertical coordinate, that is, the human height, and R is the human aspect ratio.

[0085] In daily behavior, when a person is falling or standing, the feet are in contact with the ground in most cases, and extremely special cases are not within the scope of this embodiment. Since the cameras shooting the body are in different positions, there may be misjudgments on whether the person has fallen, so this embodiment proposes the body center angle. The body center angle is defined as the angle between the human torso and the feet, which will help distinguish between a person falling and standing along the camera shooting direction. When a person falls along the camera shooting direction, it is easy to be misjudged as a standing posture based solely on the calculated body inclination angle and the body width-to-height ratio. Therefore, this embodiment uses the body center angle as one of the features for determining the falling posture. Refer to Figure 5 , Figure 5 Schematic diagram of the human body center angle involved in the first embodiment of the human fall detection method based on machine vision of the present invention. Figure 5 In the figure, 5 is defined as the left shoulder key point, 6 is the center point of the right shoulder, the center point of the left shoulder key point 5 and the right shoulder key point 6 is P, 11 is defined as the left hip key point, 12 is defined as the right hip key point, the center point of the left hip key point 11 and the right hip key point 12 is Q, the line segment passing through the center point P and the center point Q is the center line of the human body, denoted by O, 15 is defined as the left ankle key point, 16 is defined as the right ankle key point, the coordinate of the left ankle key point 15 or the right ankle key point 16 is F, that is, the coordinate of F is (x 15 ,y 15 ) or (x 16 ,y 16 ).

[0086] In a specific implementation, the human body center angle may be calculated according to the coordinate information of the human body shoulder key points, the coordinate information of the human body hip key points, and the coordinate information of the human body foot key points. For example, the human body center angle may be calculated according to the coordinate information of the human body shoulder key points, the coordinate information of the human body hip key points, and the coordinate information of the human body foot key points by the following formula;

[0087]

[0088] In the formula, θ PQF is the human body center angle, P is the coordinate of the center point of the right shoulder key point and the left shoulder key point, Q is the coordinate of the center point of the right hip key point and the left hip key point, and F is the coordinate of the left ankle key point or the right ankle key point.

[0089] Step S30: judging whether the person to be detected is in a falling posture according to the body inclination angle, the body width-to-height ratio and the body center angle;

[0090] In this embodiment, whether the body is in a falling posture is comprehensively judged based on three indicators: the body inclination angle, the body width-to-height ratio, and the body center angle.

[0091] In this embodiment, the YOLOv8 Pose algorithm can be used to extract the coordinates of the key points of the human body, and the coordinates of the shoulders and hips can be used to calculate the human body inclination angle. Figure 6 , Figure 6 This is a human body tilt angle effect diagram involved in the first embodiment of the human fall detection method based on machine vision of the present invention, Figure 6 In the figure, P is the center point of the left shoulder key point and the right shoulder key point of the human body, Q is the center point of the left hip key point and the right hip key point of the human body, and the border color represents different human postures. The green border represents the standing posture, the blue border represents the sitting posture, and the red border represents the falling posture. Figure 6 (a) shows the effect of human body tilt angle in standing posture. Figure 6 (b) shows the effect of the human body tilt angle in a sitting position. Figure 6 (c) shows the effect diagram of the human body tilt angle in a falling posture. In the specific implementation, due to the variety of human body movements and postures, it is obviously not enough to use only the human body tilt angle to determine whether the human body is in a falling posture.

[0092] In this embodiment, the human body width-to-height ratio can be calculated based on the frame of the human body target detection, referring to Figure 7 , Figure 7 This is a human body aspect ratio effect diagram involved in the first embodiment of the human fall detection method based on machine vision of the present invention, Figure 7 middle, Figure 7 (a) shows the effect of the width-to-height ratio of the human body in a standing position. Figure 7 (b) shows the effect of the width-to-height ratio of the human body in a sitting position. Figure 7 (c) shows the effect of the human body’s aspect ratio in a falling posture. Since the aspect ratio of the human body varies greatly in different postures, the aspect ratio of the sitting posture is significantly greater than that of the standing posture and the falling posture. The aspect ratio will also increase in the horizontal falling posture in the camera’s field of view.

[0093] Reference Figure 8 , Figure 8 These are the before and after effect diagrams of introducing the human body center angle involved in the first embodiment of the human fall detection method based on machine vision of the present invention. Figure 8 In the case of a human body in a falling posture, Figure 8 As shown in (a), when the central angle is not included, the human body is judged to be in a standing posture, such as Figure 8 As shown in (b), when the center angle is included, the human body is judged to be in a falling posture. This is because the camera shooting angle leads to incorrect calculation of the human body inclination angle. Also affected by the camera inclination angle, the vertical line of the human body center cannot be well positioned. It can be seen that this problem can be significantly improved by adding the human body center angle.

[0094] Therefore, using the coordinates of the torso and feet, the human body center angle is calculated, referring to Fig. 9 , Fig. 9 This is a rendering of the human body center angle involved in the first embodiment of the human fall detection method based on machine vision of the present invention. Fig. 9 In the figure, P is the center point of the left shoulder key point and the right shoulder key point of the human body, Q is the center point of the left hip key point and the right hip key point of the human body, and F is the left ankle key point or the right ankle key point. Fig. 9 (a) shows the effect diagram of the human body center angle in standing posture. Fig. 9 (b) shows the effect diagram of the human body center angle in a sitting posture. Fig. 9 (c) shows the effect diagram of the center angle of the human body in the falling posture. It can be seen that when the human body is in the standing and falling postures, the center angle is larger, up to 180°, and when it is in the sitting posture, the center angle is smaller. It can be seen that the center angle can better distinguish the falling and sitting postures at special angles.

[0095] Further, whether the person to be detected is in a falling posture can be judged according to whether the body inclination angle is greater than or equal to a preset body inclination angle threshold, whether the body width-to-height ratio is greater than or equal to a preset body width-to-height ratio threshold, and whether the body center angle is greater than or equal to a preset body center angle threshold. That is, when the body inclination angle is greater than or equal to the preset body inclination angle threshold, the body width-to-height ratio is greater than or equal to the preset body width-to-height ratio threshold, and the body center angle is greater than or equal to the preset body center angle threshold, it is determined that the person to be detected is in a falling posture; otherwise, it is determined that the person to be detected is not in a falling posture. Among them, the preset body inclination angle threshold, the preset body width-to-height ratio threshold, and the preset body center angle threshold can all be set according to actual needs, and this embodiment does not limit this. For example, in the example shown below, the preset body inclination angle threshold is set to 45°, the preset body width-to-height ratio threshold is set to 0.45, and the preset body center angle threshold is set to 133°.

[0096] In order to determine whether a human body is in a falling posture, this embodiment uses 30 human body posture sample images taken in actual scenes for testing, including 15 sitting postures, 13 falling postures, and 2 standing postures, and designs the following 5 schemes for verification:

[0097] Solution 1: Body tilt angle. Only a single body tilt angle is used for judgment, and its threshold is 45°;

[0098] Solution 2: Human body aspect ratio. Only the human body aspect ratio is used for judgment, and the threshold is 0.45;

[0099] Solution 3: Body center angle. Only the body center angle is used for judgment, and the threshold is 133°;

[0100] Solution 4: Human body tilt angle and human body aspect ratio. The human body tilt angle and human body aspect ratio are used for comprehensive judgment, and the threshold is the same as that of Solution 1 and 2.

[0101] Solution 5: Human body tilt angle, human body aspect ratio, and human body center angle. The three characteristic indicators of human body tilt angle, aspect ratio, and center angle are used for comprehensive judgment, and the thresholds are the same as those of solutions 1, 2, and 3.

[0102] Referring to Table 2, Table 2 is a table of human fall detection results of various schemes involved in the first embodiment of the human fall detection method based on machine vision of the present invention. Table 2 shows the detection results of fall detection by the above five schemes.

[0103] Table 2 Human fall detection results of each scheme

[0104]

[0105]

[0106] It can be seen that solution 5 has the highest accuracy. Combining the above three indicators can better evaluate human posture. Fig.10 , Fig.10 This is a fall detection effect diagram of the first embodiment of the method for detecting a person falling based on machine vision according to the present invention. Fig.10 (a) shows the detected standing posture effect diagram. Fig.10 (b) shows the detected sitting posture effect diagram. Fig.10 (c) shows the effect of the detected fall posture. Fig.10 (c) It can be seen that the human body falling along the camera shooting direction can be well detected, indicating that the model that comprehensively evaluates the human body posture by combining the above three indicators has achieved good results in falling posture detection.

[0107] Step S40: when it is determined that the person to be detected is in the falling posture, a fall alarm message is output.

[0108] In a specific implementation, when it is determined that the person to be detected is in the falling posture, the falling time of the person to be detected in the falling posture can be calculated. When the falling time is greater than or equal to the preset falling time, a falling alarm message is output to remind the monitoring personnel of the monitoring screen to conduct comprehensive emergency dispatch and command in time when the alarm is given, so that the monitoring personnel do not need to stare at the monitoring screen all day for dynamic monitoring of the monitoring control room, but only need to conduct comprehensive emergency dispatch and command in time when the system alarms, which effectively reduces the labor intensity and improves the monitoring effectiveness. On the other hand, it also effectively makes up for the weakness of human eye monitoring and improves the monitoring accuracy. By analyzing the posture of the person to be detected in the monitoring screen in real time, it automatically determines whether it has fallen and alarms, which is conducive to the timely prevention and control of hazardous events. In addition, the monitoring end adopts the "alarm trigger" video acquisition mode, which can only record and collect the target behavior specified by the program (such as the person to be detected in the falling posture), effectively improving the efficiency of post-review video, reducing the storage redundancy of irrelevant video, saving hard disk storage space, and realizing the simplicity of video resource management, so as to further improve the management efficiency of the monitoring end. Among them, the preset fall time can be set according to actual needs, such as 1 minute, and this embodiment does not limit this.

[0109] Reference Fig.11 , Fig.11 The technical roadmap of the first embodiment of the method for detecting a person falling down based on machine vision of the present invention is as follows: Fig.11 In the process, the YOLOv8 model is first used to determine whether there is a person to be detected in the monitoring screen. When the person to be detected is detected, the human skeleton key points of the person to be detected are identified based on the YOLOv8-Pose model, and then the coordinate information of each human skeleton key point is obtained. Then, the human body inclination angle is calculated using the coordinates of the human body shoulder and hip, the human body width-to-height ratio is calculated using the coordinates of the human body target frame, and the human body center angle is calculated using the coordinates of the human body shoulder, hip, and foot. The three characteristic indicators of human body inclination angle, human body width-to-height ratio, and human body center angle are used to comprehensively judge the human body posture to further determine whether it is in a falling posture. Finally, the fall duration is calculated for the person who is currently in a falling posture, and an alarm is issued if the duration exceeds 1 minute.

[0110] In this embodiment, the human skeleton key points of the person to be detected in the monitoring screen are identified and processed to obtain the coordinate information of multiple human skeleton key points; the human body inclination angle, the human body aspect ratio and the human body center angle are calculated according to the coordinate information of the multiple human body skeleton key points; the human body inclination angle, the human body aspect ratio and the human body center angle are used to determine whether the person to be detected is in a falling posture; when it is determined that the person to be detected is in the falling posture, a fall alarm message is output. Compared with the processor burden and high computing power requirements of the computer caused by inputting a large number of pictures and videos of human falls in deep learning and performing fall judgment based on the pixel-level image features of people falling in each frame of the image extracted, this embodiment performs the above-mentioned fall detection based on the lightweight YOLOv8-Pose algorithm, which effectively improves the accuracy of human fall detection based on machine vision under the premise of limited computer computing power. Further, by comprehensively judging whether the person to be detected is in a falling posture based on the human body inclination angle, the human body aspect ratio and the human body center angle, the accuracy of fall detection is effectively improved. In addition, since comprehensive emergency dispatch and command only need to be carried out in a timely manner when the system alarms, the labor intensity is effectively reduced and the monitoring effectiveness is improved. On the other hand, it also effectively makes up for the weaknesses of human eye monitoring and improves the accuracy of monitoring. By real-time analysis of the posture of the person to be detected in the monitoring screen, it automatically determines whether it has fallen and issues an alarm, which is conducive to the timely prevention and control of hazardous events and improves the monitoring effectiveness and management efficiency of the monitoring end.

[0111] Reference Fig.12 , Fig.12 It is a structural block diagram of the first embodiment of the personnel fall detection system based on machine vision of the present invention.

[0112] like Fig.12 As shown, the machine vision-based personnel fall detection system proposed in the embodiment of the present invention includes:

[0113] The key point recognition module 100 is used to perform human skeleton key point recognition processing on the person to be detected in the monitoring picture to obtain coordinate information of multiple human skeleton key points;

[0114] A parameter calculation module 200 is used to calculate the human body inclination angle, the human body aspect ratio and the human body center angle according to the coordinate information of a plurality of human body skeleton key points;

[0115] A posture judgment module 300 is used to judge whether the person to be detected is in a falling posture according to the body inclination angle, the body width-to-height ratio and the body center angle;

[0116] The fall alarm module 400 is used to output a fall alarm message when it is determined that the person to be detected is in the fall posture.

[0117] Optionally, the coordinate information of the plurality of human skeleton key points includes the coordinate information of the human shoulder key points, the coordinate information of the human hip key points, the coordinate information of the human target frame, and the coordinate information of the human foot key points, and the parameter calculation module 200 is further used to calculate the human body inclination angle according to the coordinate information of the human shoulder key points and the coordinate information of the human hip key points;

[0118] The parameter calculation module 200 is further used to calculate the human body aspect ratio according to the coordinate information of the human body target frame;

[0119] The parameter calculation module 200 is further used to calculate the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point.

[0120] Optionally, the parameter calculation module 200 is further used to calculate the human body inclination angle according to the coordinate information of the human body shoulder key point and the coordinate information of the human body hip key point by the following formula:

[0121]

[0122] Where (x5, y5) is the coordinate of the key point of the left shoulder of the human body, (x6, y6) is the coordinate of the key point of the right shoulder of the human body, (x P ,y P ) is the coordinate of the center point P of the left shoulder key point and the right shoulder key point of the human body, (x 11 ,y 11 ) is the coordinate of the left hip key point of the human body, (x 12 ,t 12 ) is the coordinate of the key point of the right hip of the human body, (x Q ,t Q ) is the coordinate of the center point Q of the left hip key point and the right hip key point of the human body, θ PQ is the body inclination angle;

[0123] The parameter calculation module 200 is further used to calculate the human body aspect ratio according to the coordinate information of the human body target frame by the following formula;

[0124]

[0125] Wherein, max(x) is the maximum horizontal coordinate of the human body target frame, min(x) is the minimum horizontal coordinate of the human body target frame, dx is the distance between the maximum horizontal coordinate and the minimum horizontal coordinate, max(y) is the maximum vertical coordinate of the human body target frame, min(y) is the minimum vertical coordinate of the human body target frame, dy is the distance between the maximum vertical coordinate and the minimum vertical coordinate, and R is the human body aspect ratio;

[0126] The parameter calculation module 200 is further used to calculate the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point by the following formula;

[0127]

[0128] In the formula, θ PQF is the human body center angle, P is the coordinate of the center point of the left shoulder key point and the right shoulder key point, Q is the coordinate of the center point of the left hip key point and the right hip key point, and F is the coordinate of the left ankle key point or the right ankle key point.

[0129] Optionally, the posture judgment module 300 is also used to determine that the person to be detected is in a falling posture when the body inclination angle is greater than or equal to a preset body inclination angle threshold, the body aspect ratio is greater than or equal to a preset body aspect ratio threshold, and the body center angle is greater than or equal to a preset body center angle threshold; otherwise, it is determined that the person to be detected is not in a falling posture.

[0130] Optionally, the fall alarm module 400 is further configured to calculate the fall duration of the person to be detected in the fall posture when it is determined that the person to be detected is in the fall posture;

[0131] The fall alarm module 400 is further configured to output a fall alarm message when the fall duration is greater than or equal to a preset fall duration.

[0132] In this embodiment, the human skeleton key points of the person to be detected in the monitoring screen are identified and processed to obtain the coordinate information of multiple human skeleton key points; the human body inclination angle, the human body aspect ratio and the human body center angle are calculated according to the coordinate information of the multiple human body skeleton key points; the human body inclination angle, the human body aspect ratio and the human body center angle are used to determine whether the person to be detected is in a falling posture; when it is determined that the person to be detected is in the falling posture, a fall alarm message is output. Compared with the processor burden and high computing power requirements of the computer caused by inputting a large number of pictures and videos of human falls in deep learning and performing fall judgment based on the pixel-level image features of people falling in each frame of the image extracted, this embodiment performs the above-mentioned fall detection based on the lightweight YOLOv8-Pose algorithm, which effectively improves the accuracy of human fall detection based on machine vision under the premise of limited computer computing power. Further, by comprehensively judging whether the person to be detected is in a falling posture based on the human body inclination angle, the human body aspect ratio and the human body center angle, the accuracy of fall detection is effectively improved. In addition, since comprehensive emergency dispatch and command only need to be carried out in a timely manner when the system alarms, the labor intensity is effectively reduced and the monitoring effectiveness is improved. On the other hand, it also effectively makes up for the weaknesses of human eye monitoring and improves the accuracy of monitoring. By real-time analysis of the posture of the person to be detected in the monitoring screen, it automatically determines whether it has fallen and issues an alarm, which is conducive to the timely prevention and control of hazardous events and improves the monitoring effectiveness and management efficiency of the monitoring end.

[0133] Other embodiments or specific implementations of the machine vision-based personnel fall detection system of the present invention can refer to the above-mentioned method embodiments and will not be described in detail here.

[0134] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, and the computer program includes a program code for executing the method shown in the flowchart.

[0135] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, device or device or used in combination with it. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0136] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0137] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0138] As another aspect, the present application further provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the human fall detection method based on machine vision described in the above embodiment.

[0139] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the human fall detection method based on machine vision described in the above embodiment.

[0140] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0141] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0142] For those skilled in the art, the specific meanings of the above terms in the present invention can be understood in specific situations; the drawings in the embodiments are used to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0143] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for detecting a person falling down based on machine vision, characterized in that: The following steps are involved: Perform human skeleton key point recognition processing on the person to be detected in the monitoring image to obtain coordinate information of multiple human skeleton key points; wherein the human skeleton key points include human shoulder key points, human hip key points, human target frame, and human foot key points; Calculate the human body inclination angle, human body aspect ratio and human body center angle according to the coordinate information of multiple key points of human skeleton; Determine whether the person to be detected is in a falling posture according to the body inclination angle, the body width-to-height ratio and the body center angle; When it is determined that the person to be detected is in the falling posture, a fall warning message is output.

2. The method for detecting a person falling down based on machine vision according to claim 1, characterized in that: The method of calculating the human body inclination angle, the human body aspect ratio and the human body center angle according to the coordinate information of the multiple human body skeleton key points specifically includes: Calculating the human body inclination angle according to the coordinate information of the key points of the human body shoulder and the coordinate information of the key points of the human body hip; Calculate the human body aspect ratio according to the coordinate information of the human body target frame; The human body center angle is calculated according to the coordinate information of the key points of the human body shoulder, the coordinate information of the key points of the human body hip and the coordinate information of the key points of the human body foot.

3. The method for detecting a person falling down based on machine vision according to claim 2, characterized in that: The calculating of the human body inclination angle according to the coordinate information of the human body shoulder key point and the coordinate information of the human body hip key point specifically includes: According to the coordinate information of the key points of the human body shoulder and the coordinate information of the key points of the human body hip, the human body inclination angle is calculated by the following formula; Where (x5, y5) is the coordinate of the key point of the left shoulder of the human body, (x6, y6) is the coordinate of the key point of the right shoulder of the human body, (x P ,y P ) is the coordinate of the center point P of the left shoulder key point and the right shoulder key point of the human body, (x 11 ,y 11 ) is the coordinate of the left hip key point of the human body, (x 12 ,y 12 ) is the coordinate of the key point of the right hip of the human body, (x Q ,y Q ) is the coordinate of the center point Q of the left hip key point and the right hip key point of the human body, θ PQ is the inclination angle of the human body.

4. The method for detecting a person falling down based on machine vision according to claim 2 or 3, characterized in that: The calculating the human body aspect ratio according to the coordinate information of the human body target frame specifically includes: According to the coordinate information of the human body target frame, the human body aspect ratio is calculated by the following formula; Wherein, max(x) is the maximum horizontal coordinate of the human target frame, min(x) is the minimum horizontal coordinate of the human target frame, dx is the distance between the maximum horizontal coordinate and the minimum horizontal coordinate, max(y) is the maximum vertical coordinate of the human target frame, min(y) is the minimum vertical coordinate of the human target frame, dy is the distance between the maximum vertical coordinate and the minimum vertical coordinate, and R is the human body aspect ratio.

5. The method for detecting a person falling down based on machine vision according to claim 2, characterized in that: The calculating of the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point specifically includes: According to the coordinate information of the key points of the human body shoulder, the coordinate information of the key points of the human body hip and the coordinate information of the key points of the human body foot, the human body center angle is calculated by the following formula; In the formula, θ PQF is the human body center angle, P is the coordinate of the center point of the left shoulder key point and the right shoulder key point, Q is the coordinate of the center point of the left hip key point and the right hip key point, and F is the coordinate of the left ankle key point or the right ankle key point.

6. The method for detecting a person falling down based on machine vision according to claim 2, characterized in that: The determining whether the person to be detected is in a falling posture according to the body inclination angle, the body width-to-height ratio, and the body center angle specifically includes: When the body inclination angle is greater than or equal to a preset body inclination angle threshold, the body width-to-height ratio is greater than or equal to a preset body width-to-height ratio threshold, and the body center angle is greater than or equal to a preset body center angle threshold, it is determined that the person to be detected is in a falling posture; Otherwise, it is determined that the person to be detected is not in a falling posture.

7. The method for detecting a person falling down based on machine vision according to claim 6, characterized in that: When it is determined that the person to be detected is in the falling posture, outputting a falling alarm message specifically includes: When it is determined that the person to be detected is in the falling posture, calculating the falling time of the person to be detected in the falling posture; When the fall duration is greater than or equal to the preset fall duration, a fall alarm message is output.

8. A personnel fall detection system based on machine vision, characterized in that: include: The key point recognition module is used to perform human skeleton key point recognition processing on the person to be detected in the monitoring picture to obtain the coordinate information of multiple human skeleton key points; A parameter calculation module is used to calculate the human body inclination angle, human body aspect ratio and human body center angle according to the coordinate information of multiple key points of human skeleton; A posture judgment module, used to judge whether the person to be detected is in a falling posture according to the body inclination angle, the body width-to-height ratio and the body center angle; The fall alarm module is used to output a fall alarm message when it is determined that the person to be detected is in the fall posture.

9. The machine vision-based personnel fall detection system according to claim 8, characterized in that: The coordinate information of the plurality of key points of the human skeleton includes the coordinate information of the key points of the human shoulder, the coordinate information of the key points of the human hip, the coordinate information of the human target frame, and the coordinate information of the key points of the human foot; The parameter calculation module is further used to calculate the human body inclination angle according to the coordinate information of the key points of the human body shoulder and the coordinate information of the key points of the human body hip; The parameter calculation module is further used to calculate the human body aspect ratio according to the coordinate information of the human body target frame; The parameter calculation module is also used to calculate the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point.

10. The machine vision-based personnel fall detection system according to claim 9, characterized in that: The parameter calculation module is further used to calculate the human body inclination angle according to the coordinate information of the human body shoulder key point and the coordinate information of the human body hip key point by the following formula; Where (x5, y5) is the coordinate of the key point of the left shoulder of the human body, (x6, y6) is the coordinate of the key point of the right shoulder of the human body, (x P ,y P ) is the coordinate of the center point P of the left shoulder key point and the right shoulder key point of the human body, (x 11 ,y 11 ) is the coordinate of the left hip key point of the human body, (x 12 ,y 12 ) is the coordinate of the key point of the right hip of the human body, (x Q ,y Q ) is the coordinate of the center point Q of the left hip key point and the right hip key point of the human body, θ PQ is the body inclination angle; The parameter calculation module is further used to calculate the human body aspect ratio according to the coordinate information of the human body target frame by the following formula; Wherein, max(x) is the maximum horizontal coordinate of the human body target frame, min(x) is the minimum horizontal coordinate of the human body target frame, dx is the distance between the maximum horizontal coordinate and the minimum horizontal coordinate, max(y) is the maximum vertical coordinate of the human body target frame, min(y) is the minimum vertical coordinate of the human body target frame, dy is the distance between the maximum vertical coordinate and the minimum vertical coordinate, and R is the human body aspect ratio; The parameter calculation module is further used to calculate the human body center angle according to the coordinate information of the human body shoulder key point, the coordinate information of the human body hip key point and the coordinate information of the human body foot key point by the following formula; In the formula, θ PQF is the human body center angle, P is the coordinate of the center point of the left shoulder key point and the right shoulder key point, Q is the coordinate of the center point of the left hip key point and the right hip key point, and F is the coordinate of the left ankle key point or the right ankle key point.

Images