Method for calculating shoulder joint mobility based on visual recognition

Abstract

The embodiment of the application discloses a shoulder joint activity calculation method based on visual recognition, and relates to the technical field of body measurement and evaluation. The method comprises the following steps: acquiring a front body image of a tester performing a first preset test action, i.e., arm forward-up, by a depth camera; extracting first spatial coordinate data of joint mark points of the tester from the front body image; calculating a first length value and a second length value according to the first spatial coordinate data; dividing the second length value by the first length value, and taking the inverse sine to obtain arm forward-up joint angle data; and calculating a shoulder joint activity score according to the arm forward-up joint angle data. The embodiment of the application calculates the arm forward-up joint angle data and the arm abduction-up joint angle data at one time by using the front body image of the subject, has high accuracy, does not need to collect the body image in multiple directions, is simple to operate, and saves time and effort.

Description

Technical Field

[0001] This invention relates to the field of body measurement and assessment technology, and in particular to a method for calculating shoulder joint range of motion based on visual recognition. Background Technology

[0002] Decreased shoulder joint range of motion is commonly seen in non-manual laborers who engage in frequent but not strenuous shoulder and arm movements, such as chefs, writers, drivers, and certain office workers. Although these individuals are not manual laborers, their jobs require frequent shoulder and arm movements or prolonged periods with the shoulder and arm fixed in one position. The incidence of conditions similar to frozen shoulder increases with age. Decreased shoulder joint range of motion can affect a person's work and daily living abilities. Therefore, measuring shoulder joint range of motion can help predict the risk of developing frozen shoulder at an early stage.

[0003] Current shoulder joint range of motion testing is mostly instrument-based, involving fixing multiple measuring rods on one side and measuring the angle by rotation. For example, Chinese patent application CN113080943A discloses a shoulder joint range of motion measuring device, and Chinese patent CN210749218U discloses a device for evaluating shoulder and neck range of motion after thyroid cancer surgery. While camera-based visual recognition based on big data can now identify human joint angles, current calculations of joint angles based on visual recognition mostly rely on coordinate points identified by the camera. However, because cameras cannot yet accurately identify human body markers, current joint angle calculations still lack high accuracy. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a highly accurate method for calculating shoulder joint range of motion based on visual recognition.

[0005] A method for calculating shoulder joint range of motion based on visual recognition, comprising:

[0006] Acquire a frontal human image of the test subject as they complete a first preset test action, captured by a depth camera. The first preset test action is to extend and raise the arm forward.

[0007] Extract the first spatial coordinate data of the test subject's joint markers from the frontal human image;

[0008] Based on the first spatial coordinate data, calculate the average absolute value of the vertical length of the test subject's arm from its natural downward position to its maximum forward extension and upward movement, and multiply the average value by a preset coefficient to obtain the first length value; and record the vertical length of the test subject's arm when it is fully extended forward and raised to its maximum as the second length value;

[0009] Divide the second length value by the first length value and take the arcsine to obtain the arm extension and upward joint angle data;

[0010] The shoulder joint range of motion score is calculated based on the arm extension and elevation joint angle data.

[0011] The shoulder joint range of motion calculation method based on visual recognition provided in this invention has two aspects. First, it uses visual recognition to collect the joint points of the subject. Using the subject's frontal human image, it calculates a first length value (which can be understood as the length of the hypotenuse in a right triangle) and a second length value (which can be understood as the length of one leg in a right triangle). The second length value is divided by the first length value, and the arcsine is taken to obtain the arm extension and elevation joint angle data, thus indirectly calculating the arm extension and elevation joint angle with high accuracy. Second, using the subject's frontal human image, it calculates the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data at the same time. Based on the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data, it can more accurately calculate the shoulder joint range of motion score. It does not require collecting human images from multiple directions, making the operation simple, time-saving, and labor-saving. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating the shoulder joint range of motion calculation method based on visual recognition of the present invention.

[0014] Figure 2 This is a schematic diagram of the location of joint markers in this invention, where (a) is a side view of the human body and (b) is a front view of the human body;

[0015] Figure 3 This is a schematic diagram of the projection of the joint space angle in the coronal, sagittal, and horizontal planes in this invention;

[0016] Figure 4 This is a schematic diagram illustrating the change in the joint angle of the arm during the testing process of this invention. Detailed Implementation

[0017] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0018] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0019] This invention provides a method for calculating shoulder joint range of motion based on visual recognition, such as... Figure 1 As shown, it includes:

[0020] Step 101: Acquire a frontal human image of the test subject completing the first preset test action, captured by a depth camera. The first preset test action is to extend and raise the arm.

[0021] In this step, a frontal human image of the test subject performing a preset test action is captured by a camera, preferably a depth camera with 3D functionality. The first preset test action can be extending and raising the arm forward.

[0022] In practice, the subject can extend and raise their arm forward three times, and the camera will capture the X, Y, and Z directions during the arm extension and raising process (see reference). Figure 3 The joint marker time series is obtained to obtain the first spatial coordinate data, which is helpful for subsequent steps to calculate joint angle data and shoulder joint range of motion score.

[0023] For example, during the test, the subject follows the instructions of the testing instrument to extend both arms forward and upward: the subject stands upright, arms hang naturally with palms facing inward, and extends both arms forward and upward to the maximum extent, or stops when the arms are perpendicular to the ground, and follows the instructions of the testing instrument to perform the action 3 times.

[0024] Step 102: Extract the first spatial coordinate data of the joint markers of the test subject from the frontal human image;

[0025] In this step, a visual recognition system can be used to extract the first spatial coordinate data corresponding to the time series of the joint markers of the test subject from the frontal human image. The specific extraction method can adopt conventional techniques in this field, which will not be elaborated here.

[0026] As an optional embodiment, the joint markers may include at least 20 markers, including: head, upper cervical vertebrae, middle cervical vertebrae, lower cervical vertebrae, middle thoracic vertebrae, upper lumbar vertebrae, middle lumbar vertebrae, sacrum, left / right shoulder, left / right elbow, left / right wrist, left / right hip, left / right knee, and left / right ankle.

[0027] Understandably, in this step, only the spatial coordinate data of the joint markers involved in the arm can be extracted. The joint markers involved in the arm are the left / right shoulder, left / right elbow, and left / right wrist.

[0028] The extracted spatial coordinate data can be, for example, as follows:

[0029] ι i =(x i y i , z i Let R(x, y, z) represent the coordinates of the i-th joint marker. The set of joint markers for the test subject is represented as R(x, y, z). Spatial coordinate data of 20 joint markers were collected from various time series of frontal human images during the test. Point 1 is the first cervical vertebra (C1); point 2 is the fourth cervical vertebra (C4); point 3 is the seventh cervical vertebra (C7); point 10 is the sixth thoracic vertebra (T6); point 11 is the twelfth thoracic vertebra (T12); point 12 is the third lumbar vertebra (L3); and point 13 is the upper edge of the sacrum.

[0030] The joint markers and numbers of the test subjects can be shown in Table 1 and... Figure 2 As shown.

[0031] Table 1 Joint Markers and Numbers

[0032] serial number Marker point serial number Marker point 0 head 10 Mid-thoracic vertebrae 1 upper cervical spine 11 upper lumbar spine 2 Middle cervical spine 12 Mid-lumbar spine 3 lower cervical spine 13 Sacrum 4 left shoulder 14 Left hip 5 left elbow 15 left knee 6 left wrist 16 left ankle 7 right shoulder 17 Right hip 8 right elbow 18 right knee 9 right wrist 19 Right ankle

[0033] The joint markers defined in this invention typically correspond to joints or bony landmarks on the human body with a certain degree of freedom. The current state of the human body is estimated by calculating the relative positions of the joints in three-dimensional space. Due to the special nature of the human body structure, the lines connecting the markers are only connected by the human body structure.

[0034] To improve the accuracy of subsequent evaluations, the extracted first spatial coordinate data can be preprocessed as follows:

[0035] Preprocessing of spatial coordinate data:

[0036] When training the system to automatically identify joint markers, in addition to using the source dataset, the system can also be trained based on the company's own exercise library. Unlike open-source libraries, this library includes a large number of test and exercise movements, which can improve the system's accuracy in identifying markers.

[0037] After marker point identification, data denoising is performed first. Using wavelet thresholding, the marker point acceleration is obtained based on the marker point coordinates and time series. Then, based on the upper limit of human acceleration and expert discussions, a new data limit is set. If the limit is exceeded, the data is deleted. These two processes result in a marker point sequence that better reflects the human body.

[0038] Step 103: Based on the first spatial coordinate data, calculate the average absolute value of the vertical length of the test subject's arm from its natural downward position to its maximum forward extension, and multiply the average value by a preset coefficient to obtain the first length value; and record the vertical length of the test subject's arm when it is fully extended forward as the second length value.

[0039] In this step, the vertical length of the arm can be any one of the vertical length of the entire arm, the vertical length of the upper arm, or the vertical length of the forearm.

[0040] In practice, spatial coordinate data / sequences of the shoulder, elbow, and wrist can be obtained based on joint marker point identification. Here, the Y-axis coordinates of the shoulder, elbow, and wrist (denoted as Y) can be used as the basis for identification. s Y e Y w Calculate the vertical length of the arm. When the vertical length of the arm is equal to the vertical length of the upper arm, the vertical length S of the upper arm is... u =Y e -Y s When the vertical length of the arm is equal to the vertical length of the forearm, the vertical length S of the forearm is... l =Y w -Y e When the vertical length of the arm is equal to the total vertical length of the arm, the vertical length S of the entire arm is... u+l =Y w -Y s .

[0041] Based on the above, calculate the average value L0 of the absolute value of the vertical length of the test subject's arm from its natural downward position to its maximum forward extension, and multiply this average value L0 by a preset coefficient k to obtain the first length value kL0; and record the vertical length L of the test subject's arm when it is fully extended forward as the second length value.

[0042] Step 104: Divide the second length value by the first length value and take the arcsine to obtain the arm extension and upward joint angle data;

[0043] In this step, based on the previous example, the formula for calculating the joint angle D of the arm extension and upward movement can be as follows:

[0044]

[0045] Here, ASIN() is the arcsine function, and DEGREES() converts the arcsine value into an angle.

[0046] In this invention, the joint angle of the arm extended forward and raised is the shoulder angle. As can be seen from the above calculation formula, it is actually the shoulder angle in the sagittal plane.

[0047] In existing technologies, because the captured image is a frontal human image of the test subject, and the test subject's arm extends along the 3D depth direction of the camera (not in the vertical plane in front of the camera), it is difficult for the camera to accurately identify joint markers. Even with highly accurate image recognition algorithms, the calculated joint angles are difficult to achieve high accuracy. To solve this problem, this invention takes a different approach. It calculates a first length value (which can be understood as the length of the hypotenuse in a right triangle) and a second length value (which can be understood as the length of one leg in a right triangle). The second length value is divided by the first length value, and the arcsine is taken to obtain the arm extension and elevation joint angle data, thus indirectly calculating the arm extension and elevation joint angle with high accuracy.

[0048] In this invention, the preset coefficient k can be determined in the following way:

[0049] The deviation r of the arm extension and overhead joint angle data was obtained by least squares fitting. i =(D i -y i If (i = 1, 2, 3, ..., m), then the least squares fitting criteria are:

[0050] (i = 1, 2, 3, ..., m)

[0051] Where i is the number of image frames during the test, and D i For the true value of the angle, y i The fitted value calculated using the aforementioned method, and satisfying y i =f(D i ), r i This represents the error of the i-th sample.

[0052] make Right now The goal is to find the value of k that minimizes F(k). Through formula calculation and practical testing, it was found that a k value of approximately 1.2 yields the best fitting result. In other words, the preset coefficient k can be between 1.1 and 1.3, with 1.2 being the preferred value.

[0053] As an optional embodiment, to reduce measurement errors, the step of calculating the average absolute value of the vertical length of the test subject's arm from its natural downward position to its maximum forward extension based on the first spatial coordinate data, and multiplying this average value by a preset coefficient to obtain a first length value; and recording the vertical length of the test subject's arm when it is fully extended forward as a second length value (step 103), may include:

[0054] Step 1031: Based on the first spatial coordinate data, calculate the average absolute value of the vertical length of the upper arm during the process of the test subject's arm from naturally hanging down to being stretched straight forward and raised to the maximum extent, and multiply the average value by a preset coefficient to obtain the first length value of the upper arm; and record the vertical length of the upper arm when the test subject's arm is stretched straight forward and raised to the maximum extent as the second length value of the upper arm.

[0055] Step 1032: Based on the first spatial coordinate data, calculate the average absolute value of the vertical length of the forearm during the process of the test subject's arm from hanging down naturally to being raised straight forward to the maximum extent, and multiply the average value by a preset coefficient to obtain the first length value of the forearm; and record the vertical length of the forearm when the test subject's arm is raised straight forward to the maximum extent as the second length value of the forearm.

[0056] At this point, dividing the second length value by the first length value and taking the arcsine to obtain the arm extension and upward joint angle data (step 104) may include:

[0057] Step 1041: Divide the second length value of the upper arm by the first length value of the upper arm, and take the arcsine to obtain the upper arm joint angle data;

[0058] In this step, the upper arm joint angle data can be temporarily understood as the shoulder angle D. s The calculation method is the same as the aforementioned arm extension and upward raising joint angle D, and will not be repeated here.

[0059] Step 1042: Divide the second length value of the forearm by the first length value of the forearm, and take the arcsine to obtain the forearm joint angle data;

[0060] In this step, the forearm joint angle data can be temporarily understood as the elbow angle D. e The calculation method is the same as the aforementioned arm extension and upward raising joint angle D, and will not be repeated here.

[0061] Step 1043: Take the average value of the upper arm joint angle data and the forearm joint angle data as the upper arm forward extension joint angle data.

[0062] In this step, the final arm extension and upward raising joint angle D can be calculated using the formula D = (D... s +D e ) / 2.

[0063] Thus, by taking the average of the shoulder angle calculated using the upper arm length value and the elbow angle calculated using the forearm length value (the subject's elbow is basically not flexed during the forward extension and raising of the arm, so the shoulder angle and elbow angle are basically the same) through the above steps 1031-1032 and 1041-1043, the final joint angle data of the forward extension and raising of the arm can be used to reduce measurement errors and further improve accuracy.

[0064] Step 105: Calculate the shoulder joint range of motion score based on the arm extension and elevation joint angle data.

[0065] As an optional embodiment, calculating the shoulder joint range of motion score based on the arm extension and elevation joint angle data (step 105) may include:

[0066] Step 1051: Convert the arm extension and elevation joint angle data to the range of 0-180 degrees;

[0067] As can be seen from the aforementioned calculation principle, the arm extension and elevation joint angle calculated in this invention is not the actual angle (the angle between the shoulder and the body), but rather an indirect / intermediate angle that reflects the size of the arm extension and elevation joint angle. Calculations show that this intermediate angle ranges from -60 degrees to 60 degrees. The changes in the calculated arm extension and elevation joint angle during the testing process are as follows: Figure 4 As shown.

[0068] Therefore, in this step, the joint angle data of the arm extension and elevation is proportionally converted to the range of 0 to 180 degrees to conform to the actual angle range and make it easier for subsequent evaluation.

[0069] Step 1052: Calculate the shoulder joint range of motion score based on the converted arm extension and elevation joint angle data.

[0070] In this step, based on literature and expert discussions, segmented scoring for each indicator can be established, and individual scoring for each indicator can be performed. The scoring rules for arm extension and overhead raising are shown in Table 2 below.

[0071] Table 2 Scoring Rules for Arm Extension and Overhead Raise

[0072]

[0073] Thus, by following steps 1051-1052 above, the shoulder joint range of motion score can be calculated relatively easily based on the joint angle data of the arm extension and elevation.

[0074] As another optional embodiment, the step of calculating the shoulder joint range of motion score based on the arm extension and elevation joint angle data (step 105) may include the following prior steps:

[0075] Step 101': Acquire a frontal human image of the test subject completing the second preset test action, captured by a depth camera. The second preset test action is arm abduction and raising.

[0076] In practice, the subject can perform three arm abductions and raises. The camera collects the time series of joint markers in the X, Y, and Z directions during the arm abduction and raise process, thereby obtaining subsequent second spatial coordinate data, which is helpful for calculating joint angle data and shoulder joint range of motion scores in subsequent steps.

[0077] For example, during the test, the subject follows the instructions of the testing instrument to raise both arms outward: the subject stands upright with arms extended to the sides, abdomen tucked in and chest out, maintaining a stable standing posture, and uses the shoulder muscles to pull both arms to the sides and upward until both arms are horizontal, and then continues to raise them to the maximum extent, or stops when both arms are raised to the direction of perpendicularity to the ground, and follows the instructions of the testing instrument to perform the movement 3 times.

[0078] Step 102': Extract the second spatial coordinate data of the test subject's joint markers from the frontal human image;

[0079] This step can be referred to in step 102 above, and will not be repeated here.

[0080] Step 103': Calculate the arm abduction and elevation joint angle data based on the second spatial coordinate data;

[0081] In this step, the subject's arm abduction and elevation (shoulder) joint angle can be obtained by calculating the spatial angle after identifying the joint markers. In this invention, the arm abduction and elevation joint angle is also the shoulder angle, which is actually the shoulder angle in the coronal plane, and can be calculated using conventional techniques in the art.

[0082] It is understandable that, since the image captured is a frontal human image of the test subject, the arm abduction and elevation joint angle is the shoulder angle in the coronal plane. The range of arm movement of the test subject is exactly in the vertical plane in front of the camera (not extending in the 3D depth direction of the camera). At this time, the camera can relatively accurately identify the arm length and accurately calculate the arm abduction and elevation joint angle. Therefore, the calculation of the arm abduction and elevation joint angle can use conventional algorithms in this field, without needing to refer to the calculation principle of the arm extension and elevation joint angle mentioned above.

[0083] In practice, step 103' may include:

[0084] Step 1031': Based on the second spatial coordinate data, calculate the first position data of the test subject's arm when it is naturally hanging down, and the second position data of the arm when it is abducted and raised to its maximum extent in the coronal plane;

[0085] In this step, the first arm position data may include the shoulder and elbow coordinates when the subject is standing upright; the second arm position data may include the elbow coordinates when the subject's arm is abducted and raised to its highest position.

[0086] Step 1032': Calculate the abduction and elevation joint angle data of the arm based on the first position data and the second position data of the arm.

[0087] In this embodiment of the invention, the shoulder abduction and elevation angle is the shoulder joint abduction angle in the coronal plane. The shoulder abduction angle θ can be represented by markers at the shoulder, the elbow in a standing position, and the elbow after abduction and elevation. The specific calculation method is as follows:

[0088] ι ik0 =(x ik0 y ik0 , z ik0 ), where the coordinates of the ik0th marker point represent the elbow marker point when the person is upright;

[0089] ι ik =(x ik y ik , z ik ), where ik is the coordinate of the i-th marker, representing the shoulder marker;

[0090] ι ik1 =(x ik1 y ik1 , z ik1 ), where the coordinates of the ik-th marker point represent the elbow marker point when the arm is raised to its highest point;

[0091] In the calculation, vectors m = (x1, y1) and n = (x2, y2)

[0092] (1) x1, y1, z1 = (x ik0 -x ik ), (y ik0 -y ik )

[0093] (2) x², y², z² = (x ik1 -x ik ), (y ik1 -y ik )

[0094]

[0095] The obtained cosθ k By taking the inverse cosine and converting the inverse cosine value into an angle, the abduction angle θ of the shoulder joint can be obtained.

[0096] At this point, the step of calculating the shoulder joint range of motion score based on the arm extension and elevation joint angle data (step 105) can be further defined as follows:

[0097] The shoulder joint range of motion score is calculated based on the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data.

[0098] In this step, based on literature and expert discussions, segmented scoring for each indicator can be established, and individual scoring for each indicator can be performed. The scoring rules for arm abduction and elevation are shown in Table 3 below.

[0099] Table 3 Scoring Rules for Arm Abduction and Overhead Raise

[0100]

[0101] As another optional embodiment, the step of calculating the shoulder joint range of motion score based on the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data (step 105) can be further described as follows:

[0102] Based on the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data, the shoulder joint range of motion score is calculated and the average value is taken as the final shoulder joint range of motion score.

[0103] In this way, by simultaneously using the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data, the shoulder joint range of motion score can be calculated more accurately.

[0104] In this invention, after obtaining the shoulder joint range of motion score, an evaluation (total score) can be performed based on the obtained shoulder joint range of motion score. Specifically, according to discussions among experts and researchers, the total score can be graded into four levels: normal, mildly restricted, moderately restricted, and severely restricted. The total score evaluation rules are shown in Table 4 below:

[0105] Table 4. Grading Rules for Total Score Evaluation

[0106]

[0107] In this way, the total score can be evaluated by averaging the scores of the two shoulder joint angles: the arm extended forward and the arm abducted and raised.

[0108] In summary, the shoulder joint range of motion calculation method based on visual recognition of this invention first acquires a frontal human image of the test subject completing a first preset test action captured by a depth camera. The first preset test action is to extend and raise the arm forward. Then, the first spatial coordinate data of the test subject's joint marker points are extracted from the frontal human image. Based on the first spatial coordinate data, the average value of the absolute value of the vertical length of the test subject's arm during the process from naturally hanging down to extending and raising it forward to the maximum extent is calculated. This average value is multiplied by a preset coefficient to obtain a first length value. The vertical length of the test subject's arm when it is extended and raised forward to the maximum extent is recorded as a second length value. Then, the second length value is divided by the first length value, and the arcsine is taken to obtain the arm extension and raising joint angle data. Finally, the shoulder joint range of motion score is calculated based on the arm extension and raising joint angle data. Thus, on the one hand, this embodiment of the invention uses visual recognition to collect the joint points of the subject. Using the subject's frontal human image, it calculates a first length value (which can be understood as the length of the hypotenuse in a right triangle) and a second length value (which can be understood as the length of one leg in a right triangle). The second length value is divided by the first length value, and the arcsine is taken to obtain the arm extension and elevation joint angle data, thereby indirectly calculating the arm extension and elevation joint angle with high accuracy. On the other hand, this embodiment of the invention uses the subject's frontal human image to calculate the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data at the same time. Based on the arm extension and elevation joint angle data and the arm abduction and elevation joint angle data, the shoulder joint range of motion score can be calculated more accurately. It does not require collecting human images from multiple directions, making the operation simple, time-saving, and labor-saving.

[0109] The present invention primarily improves the calculation of the shoulder joint angle during the forward extension and upward movement of the arm. Existing technologies use angle recognition methods that extract the spatial coordinates of marker points on the test subject and directly calculate the angle based on the coordinate points. Compared to existing technologies that rely solely on visual recognition, the method of the present invention (employing all the aforementioned features) increases the accuracy from 82.59% to 91.92%, demonstrating that the visual recognition-based shoulder joint range of motion calculation method of the present invention has higher accuracy.

[0110] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A visual recognition-based shoulder range of motion calculation method, characterized in that, The method comprises the following steps: acquiring a front human body image of a tester performing a first preset test action collected by a depth camera, the first preset test action being arm forward stretching and upward lifting; extracting first spatial coordinate data of joint marker points of the tester from the front human body image; calculating an average value of absolute values of vertical lengths of the arm of the tester from natural drooping to forward stretching and upward lifting to the maximum limit according to the first spatial coordinate data, multiplying the average value by a preset coefficient to obtain a first length value; and recording a vertical length of the arm of the tester when the arm is forward stretched and upward lifted to the maximum limit as a second length value; dividing the second length value by the first length value and taking an inverse sine to obtain arm forward stretching and upward lifting joint angle data; calculating a shoulder joint range of motion score according to the arm forward stretching and upward lifting joint angle data.

2. The visual recognition-based shoulder range of motion calculation method according to claim 1, characterized in that, The joint marker points include at least 20 marker points, including: head, upper end of cervical vertebra, middle part of cervical vertebra, lower end of cervical vertebra, middle part of thoracic vertebra, upper end of lumbar vertebra, middle part of lumbar vertebra, sacrum, left / right shoulder, left / right elbow, left / right wrist, left / right hip, left / right knee, and left / right ankle.

3. The visual recognition-based shoulder range of motion calculation method according to claim 1, characterized in that, The preset coefficient is 1.1-1.

3.

4. The visual recognition-based shoulder range of motion calculation method according to claim 1, characterized in that, The vertical length of the arm is any one of a vertical length of the whole arm, a vertical length of the upper arm, and a vertical length of the lower arm.

5. The visual recognition-based shoulder range of motion calculation method according to claim 1, characterized in that, The method further comprises the following steps: calculating an average value of absolute values of vertical lengths of the upper arm of the tester from natural drooping to forward stretching and upward lifting to the maximum limit according to the first spatial coordinate data, multiplying the average value by a preset coefficient to obtain an upper arm first length value; and recording a vertical length of the upper arm of the tester when the arm is forward stretched and upward lifted to the maximum limit as an upper arm second length value; calculating an average value of absolute values of vertical lengths of the lower arm of the tester from natural drooping to forward stretching and upward lifting to the maximum limit according to the first spatial coordinate data, multiplying the average value by a preset coefficient to obtain a lower arm first length value; and recording a vertical length of the lower arm of the tester when the arm is forward stretched and upward lifted to the maximum limit as a lower arm second length value; The method further comprises the following steps: dividing the upper arm second length value by the upper arm first length value and taking an inverse sine to obtain upper arm joint angle data; dividing the lower arm second length value by the lower arm first length value and taking an inverse sine to obtain lower arm joint angle data; averaging the upper arm joint angle data and the lower arm joint angle data to obtain the arm forward stretching and upward lifting joint angle data. The method further comprises the following steps:

6. The visual recognition-based shoulder range of motion calculation method according to any one of claims 1-5, characterized in that, converting the arm forward stretching and upward lifting joint angle data to a range of 0-180 degrees; calculating a shoulder joint range of motion score according to the converted arm forward stretching and upward lifting joint angle data. ​ 7. The visual recognition-based shoulder range of motion calculation method according to claim 6, characterized in that, The calculating the shoulder joint range of motion score according to the arm forward raising joint angle data further comprises: obtaining a front body image of the tester performing a second preset test action of arm abduction raising collected by the depth camera; extracting second spatial coordinate data of joint marker points of the tester from the front body image; calculating arm abduction raising joint angle data according to the second spatial coordinate data; The calculating the shoulder joint range of motion score according to the arm forward raising joint angle data further comprises: calculating the shoulder joint range of motion score according to the arm forward raising joint angle data and the arm abduction raising joint angle data.

8. The visual recognition-based shoulder range of motion calculation method according to claim 7, characterized in that, The calculating the arm abduction raising joint angle data according to the second spatial coordinate data comprises: calculating, according to the second spatial coordinate data, arm first position data of the tester when the arm is naturally drooping and arm second position data when the arm is raised to the maximum limit in the coronal plane; calculating the arm abduction raising joint angle data according to the arm first position data and the arm second position data.

9. The visual recognition-based shoulder range of motion calculation method according to claim 8, characterized in that, The calculating the shoulder joint range of motion score according to the arm forward raising joint angle data and the arm abduction raising joint angle data further comprises: calculating the shoulder joint range of motion score according to the arm forward raising joint angle data and the arm abduction raising joint angle data, respectively, and taking the average value as the final shoulder joint range of motion score.

Images