Method for evaluating knee bending during forward bending of sitting body based on machine vision
Through the knee evaluation method of forward bend of seat position based on machine vision, the position relationship between the key points of the bone is analyzed by the human posture detection model, which solves the problem of inaccurate knee evaluation in the existing technology, and achieves higher evaluation accuracy and cost-effectiveness.
Patent Information
- Application Number
- CN202311356709.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-10-19
- Publication Date
- 2025-04-22
AI Technical Summary
The existing machine vision-based seat forward flexion evaluation method is difficult to accurately evaluate the knee bent behavior of moving individuals when some human features are blocked in the image.
The knee flexion evaluation method based on machine vision is used. By obtaining real-time images of moving individuals, the human posture key point detection model is input to obtain the location of multiple bone key points, and the position relationship between these key points is analyzed to accurately evaluate whether the moving individual has knees.
It improves the accuracy of knee determination during forward bending of the seat body, reduces hardware costs and layout costs, and at the same time, the environmental requirements and difficulty of use are relatively low.
Smart Images
Figure CN119857246A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of sports evaluation, and in particular to a method for evaluating knee flexion during sit-and-reach flexion based on machine vision. Background Art
[0002] The training and assessment of some sports have been realized intelligently, and the image or video recognition solutions based on machine vision do not need to use wearable devices or sensor devices, which not only reduces the hardware cost, but also reduces the layout cost. The environmental requirements are relatively simple. At the same time, it is also a non-sensing use for users, so it is highly praised. However, in the implementation process of the existing machine vision-based solutions, especially for the sit-and-reach position, although the recognition accuracy of the identification scale of the hardware equipment is sufficient, due to some objective reasons, some human body features in the image are inevitably blocked, which easily makes it difficult to accurately evaluate some behaviors of sports individuals, such as knee bending, and urgently needs to be improved. Summary of the invention
[0003] In order to solve the technical problems existing in the background technology, the present invention proposes a method for evaluating knee flexion during sitting forward bending based on machine vision, comprising:
[0004] Acquire a real-time image of a sports individual in a sitting forward flexion position, wherein the image contains the sports individual's characteristics;
[0005] By inputting the image into the human posture key point detection model to obtain the positions of multiple skeletal key points of the individual motion characteristics;
[0006] By analyzing the positional relationship between multiple bone key points, it is possible to evaluate whether the individual has knee flexion.
[0007] Furthermore, the conditions for evaluating the existence of knee flexion in sports individuals are as follows:
[0008] The angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is less than the preset angle, and / or the angle between the right hip bone key point, the right knee bone key point, and the right ankle bone key point is less than the preset angle.
[0009] Furthermore, the conditions for evaluating the existence of knee flexion in sports individuals are as follows:
[0010] The angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is greater than the preset angle, and / or the angle between the right hip bone key point, the right knee bone key point, and the right ankle bone key point is greater than the preset angle, and the distance between the left knee bone key point and the right knee bone key point is greater than the preset value.
[0011] Furthermore, when the condition that the sports individual is evaluated to have knee flexion is specifically: the angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is greater than the preset angle; it must meet the following conditions: the image is acquired from the left side of the sports individual, or the occluded area of the left leg feature in the image is smaller than the occluded area of the right leg feature;
[0012] When it is assessed that the sport individual has flexed knees, the conditions are as follows: the angles between the right hip bone key points, the right knee bone key points, and the right ankle bone key points are greater than the preset angles; and the following conditions must be met: the image is acquired from the right side of the sport individual, or the occluded area of the right leg features in the image is smaller than the occluded area of the left leg features.
[0013] Furthermore, the training of the human body posture key point detection model specifically includes:
[0014] Batch acquire sitting body flexion images acquired at a unilateral depression angle, wherein the sitting body flexion images all contain human features;
[0015] Label the key points of human skeleton in the sit-and-forth images;
[0016] The associated annotation information and the sitting forward flexion image information are input as samples into the human posture key point detection model to train the human posture key point detection model.
[0017] Furthermore, the sitting forward flexion images at least include a first type of sitting forward flexion image in which a left leg feature among several human features is partially or completely blocked and a second type of sitting forward flexion image in which a right leg feature among several human features is partially or completely blocked.
[0018] Further, at least part of the first type of sitting forward flexion image shows knee flexion, and at least part of the second type of sitting forward flexion image shows knee flexion.
[0019] The motion evaluation method proposed in the present invention has the following technical effects:
[0020] The present invention not only reduces hardware costs, but also reduces layout costs, and has relatively simple environmental requirements. At the same time, it is also non-sensitive for users. Most importantly, the present invention is more accurate than the prior art in determining whether there is knee flexion during sitting forward flexion. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 The present invention provides a flowchart of a method for evaluating knee flexion during sitting forward flexion based on machine vision.
[0022] Figure 2 The present invention provides a flowchart of a method for evaluating knee flexion during sitting forward flexion based on machine vision.
[0023] Figure 3 This is a schematic diagram of a machine vision-based knee flexion evaluation method for sitting forward flexion proposed by the present invention. DETAILED DESCRIPTION
[0024] Example 1
[0025] A method for evaluating knee flexion during sitting forward flexion based on machine vision is proposed, including:
[0026] S1. Acquire a real-time image of a sports individual in a sitting forward flexion position, wherein the image contains characteristics of the sports individual.
[0027] The above real-time images are generally collected from the side of the moving individual at a low angle, which may easily cause partial occlusion of the moving individual's features.
[0028] S2. The image is input into the human posture key point detection model to obtain the positions of multiple skeletal key points of the individual motion characteristics.
[0029] Inputting the image into the human body posture key point detection model to obtain the positions of multiple skeleton key points will ensure that the positions of multiple skeleton key points can be obtained regardless of whether there is occlusion.
[0030] The multiple bone key points of the motion individual features that need to be acquired at the image position include at least: left knee bone key point, right knee bone key point, left ankle bone key point / right ankle bone key point, left hip bone key point / right hip bone key point.
[0031] By inputting the image into a trained human posture key point detection model, the image positions of the above-mentioned multiple skeletal key points can be obtained.
[0032] S20, batch-acquire sitting body flexion images acquired at a unilateral depression angle, wherein the sitting body flexion images all contain human body features.
[0033] Furthermore, the batch of sitting forward flexion images acquired at a unilateral depression angle include at least a first type of sitting forward flexion image in which the left leg features among the human features are partially or completely blocked, and a second type of sitting forward flexion image in which the right leg features among the human features are partially or completely blocked. Furthermore, the first type of sitting forward flexion images at least partially contain knee flexion, and the second type of sitting forward flexion images at least partially contain knee flexion.
[0034] S21. Annotate the key points of the human skeleton in the sit-and-reach image.
[0035] Specifically, the marked human skeleton key points include at least: left knee skeleton key points, right knee skeleton key points, left ankle skeleton key points / right ankle skeleton key points, left hip skeleton key points / right hip skeleton key points, and the marked features also include: human body frame and the correspondence between the above-mentioned human skeleton key points and the human body frame.
[0036] Through manual judgment and labeling, the human body posture key point detection model that is finally trained can also identify some occluded bone key points.
[0037] S22. The associated annotation information and the sit-and-forward flexion image information are input as samples into a human posture key point detection model to train the human posture key point detection model.
[0038] S3. By analyzing the positional relationship between multiple bone key points, it is possible to evaluate whether the individual has knee flexion.
[0039] S3 is described in detail through the following two scenarios:
[0040] Scene 1
[0041] The conditions for evaluating the presence of knee flexion in an athletic individual are specifically as follows: the angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is less than a preset angle, and / or the angle between the right hip bone key point, the right knee bone key point, and the right ankle bone key point is less than a preset angle. When the angle between the three is less than the preset value, it can naturally be determined that the knee is flexed.
[0042] Furthermore, when the condition that the sports individual is evaluated to have knee flexion is specifically: the angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is less than the preset angle. Then it is necessary to meet the following conditions: the image is acquired from the left side of the sports individual, so that the left leg is not easily blocked and the captured area is larger; or, the blocked area of the left leg feature in the image is smaller than the blocked area of the right leg feature.
[0043] The features of the left leg directly photographed are collected to obtain the corresponding bone key points to calculate whether the knee is bent, which is more accurate.
[0044] Furthermore, when the condition that the sports individual is evaluated to have knee flexion is specifically: the angle between the right hip bone key point, the right knee bone key point, and the right ankle bone key point is less than the preset angle. Then it is necessary to meet the following conditions: the image is acquired from the right side of the sports individual, so that the right leg is not easily blocked and the captured area is larger; or, the blocked area of the right leg feature in the image is smaller than the blocked area of the left leg feature.
[0045] The features of the right leg directly photographed are collected to obtain the corresponding bone key points to calculate whether the knee is bent, which is more accurate.
[0046] Scene 2
[0047] The conditions for evaluating the presence of knee flexion in an athlete are as follows:
[0048] The angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is greater than the preset angle, and / or the angle between the right hip bone key point, the right knee bone key point, and the right ankle bone key point is greater than the preset angle.
[0049] It should be pointed out that during the sitting forward bending process, a situation is very likely to occur. For example, the camera collects images from the left side of the moving individual. Figure 3 , the right leg of the sports individual has actually bent its knee, but while bending its knee, the right leg tilts outward, that is, tilts to the right, and the angle formed by the right leg after tilting just overlaps with the camera's top-down angle. Especially when multiple people sit side by side and bend forward at the same time, the probability of one of them having the above situation will be higher. At this time, when analyzing the images collected by the camera, the angles between the key points of the right hip bone, the right knee bone, and the right ankle bone will be greater than the preset angle, and may even appear as a straight line. If this is used to determine that the sports individual does not bend his knee, it is obviously not true. Therefore, in scene 2, In the method, a new analysis method and judgment condition are introduced, that is, the distance between the key points of the right knee bone and the left knee bone must be greater than a preset value to determine that the sports individual does not have knee flexion. The principle of technical implementation is that when the right leg tilts outward, the distance between the key points of the right knee bone and the left knee bone will increase. When the distance between the key points of the right knee bone and the left knee bone is greater than a preset value, it can be determined that knee flexion also exists. If at this time, the distance between the key points of the right knee bone and the left knee bone is less than or equal to the preset value, it can be assessed that there is no knee flexion.
[0050] Furthermore, when the condition that the sports individual has knee flexion is specifically that the angles between the left hip bone key points, the left knee bone key points, and the left ankle bone key points are greater than the preset angle, it is necessary to meet the following conditions: the image is acquired from the left side of the sports individual, so that the left leg is not easily blocked and the captured area is larger; or, the blocked area of the left leg feature in the image is smaller than the blocked area of the right leg feature.
[0051] The features of the left leg directly photographed are collected to obtain the corresponding bone key points to calculate whether the knee is bent, which is more accurate.
[0052] Furthermore, when the condition that the sports individual has knee flexion is specifically that the angles between the right hip bone key points, the right knee bone key points, and the right ankle bone key points are greater than the preset angle, it is necessary to meet the following conditions: the image is acquired from the right side of the sports individual, so that the right leg is not easily blocked and the captured area is larger; or, the blocked area of the right leg feature in the image is smaller than the blocked area of the left leg feature.
[0053] The features of the right leg directly photographed are collected to obtain the corresponding bone key points to calculate whether the knee is bent, which is more accurate.
[0054] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A method for evaluating knee flexion during sitting forward flexion based on machine vision, characterized in that: include: Acquire a real-time image of a sports individual in a sitting forward flexion position, wherein the image contains the sports individual's characteristics; By inputting the image into the human posture key point detection model to obtain the positions of multiple skeletal key points of the individual motion characteristics; By analyzing the positional relationship between multiple bone key points, it is possible to evaluate whether the individual has knee flexion.
2. The motion evaluation method according to claim 1, characterized in that: The conditions for evaluating the presence of knee flexion in an athlete are as follows: The angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is less than the preset angle, and / or the angle between the right hip bone key point, the right knee bone key point, and the right ankle bone key point is less than the preset angle.
3. The motion evaluation method according to claim 1, characterized in that: The conditions for evaluating the presence of knee flexion in an athlete are as follows: The angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is greater than the preset angle, and / or the angle between the right hip bone key point, the right knee bone key point, and the right ankle bone key point is greater than the preset angle, and the distance between the left knee bone key point and the right knee bone key point is greater than the preset value.
4. The motion evaluation method according to claim 3, characterized in that: When the condition that the sports individual is evaluated to have knee flexion is specifically: the angle between the left hip bone key point, the left knee bone key point, and the left ankle bone key point is greater than the preset angle; it must meet the following conditions: the image is acquired from the left side of the sports individual, or the occluded area of the left leg feature in the image is smaller than the occluded area of the right leg feature; When it is assessed that the sport individual has flexed knees, the conditions are as follows: the angles between the right hip bone key points, the right knee bone key points, and the right ankle bone key points are greater than the preset angles; and the following conditions must be met: the image is acquired from the right side of the sport individual, or the occluded area of the right leg features in the image is smaller than the occluded area of the left leg features.
5. The motion evaluation method according to claim 1, characterized in that: The training of the human posture key point detection model specifically includes: Batch acquire sitting body flexion images acquired at a unilateral depression angle, wherein the sitting body flexion images all contain human features; Label the key points of human skeleton in the sit-and-forward flexion images; The associated annotation information and the sitting forward flexion image information are input as samples into the human posture key point detection model to train the human posture key point detection model.
6. The motion evaluation method according to claim 5, characterized in that: The sitting forward flexion images at least include a first type of sitting forward flexion image in which a left leg feature among several human features is partially or completely blocked and a second type of sitting forward flexion image in which a right leg feature among several human features is partially or completely blocked.
7. The motion evaluation method according to claim 6, characterized in that: At least part of the first type of sitting forward flexion image includes knee flexion, and at least part of the second type of sitting forward flexion image includes knee flexion.