A human joint range of motion analysis system

Abstract

The present invention belongs to the field of joint analysis, and discloses a human joint range of motion analysis system, including a recording module and a video processing module; the recording module is used to shoot the analyzed person during the physical training of the analyzed person to obtain an action video; the video processing module includes a foreground detection unit and a joint recognition unit; the foreground detection unit is used to respectively obtain the foreground image corresponding to each frame in the action video; the joint recognition unit is used to respectively recognize each foreground image to obtain the coordinates of the joint point in each foreground image. The present invention can calculate the variation coefficient using the information of the foreground image corresponding to the first two frames, and obtain the corresponding foreground image based on the variation coefficient, which can avoid obtaining the corresponding foreground image based on all the pixel points in, significantly improve the overall foreground image acquisition efficiency, and thus improve the acquisition efficiency of the range of motion of human joints.

Description

Technical Field

[0001] The invention relates to the field of joint analysis, and in particular to a human joint range of motion analysis system. Background Art

[0002] Human joint range of motion analysis refers to calculating the range of motion of each joint of the human body from a video of the movements made by the human body during physical training, so as to facilitate subsequent analysis processes such as movement improvement based on the obtained range of motion.

[0003] The existing analysis method usually directly applies the same foreground image acquisition method to each frame in the video. This means that all pixels need to participate in the calculation of each foreground image acquisition process. Obviously, this affects the efficiency of acquiring the foreground image in the frame, that is, the area where the human body is located, thereby affecting the efficiency of acquiring the range of motion of the human joints. Summary of the invention

[0004] The purpose of the present invention is to disclose a human joint range of motion analysis system to solve the technical problems raised in the background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] The present invention provides a human joint range of motion analysis system, comprising a recording module and a video processing module;

[0007] The recording module is used to film the person being analyzed during his / her physical training to obtain action videos;

[0008] The video processing module includes a foreground detection unit and a joint recognition unit;

[0009] The foreground detection unit is used to obtain the foreground image corresponding to each frame in the action video;

[0010] The joint recognition unit is used to recognize each foreground image respectively and obtain the coordinates of the joint points in each foreground image;

[0011] The foreground image corresponding to each frame in the action video is obtained respectively, including:

[0012] Number each frame in the action video in the order of imaging time from early to late;

[0013] For frame number 1 and frame number 2 , using the preset foreground detection algorithm and Conduct testing and obtain and The corresponding foreground image;

[0014] For frame numbered q , q is greater than or equal to 2, the following method is used to Identify, obtain The corresponding foreground image:

[0015] based on and calculate The corresponding coefficient of variation;

[0016] Based on the coefficient of variation The area to be identified in ;

[0017] Use the preset foreground detection algorithm to detect the area to be identified, and obtain a set of pixels ForeU belonging to the foreground in the area to be identified;

[0018] Based on the pixel points in ForeU, the foreground completion calculation is performed to obtain The corresponding foreground image.

[0019] Preferably, photographing the person being analyzed includes:

[0020] The person being analyzed is photographed at a preset height.

[0021] Preferably, the preset height is half of the height of the person being analyzed.

[0022] Preferably, the foreground image is detected using the OpenPose algorithm to obtain joint points in the foreground image.

[0023] Preferably, an analysis module is also included, and the analysis module is used to obtain the activity range of each type of joint point based on all foreground images.

[0024] Preferably, the activity range of each type of joint point is obtained based on all foreground images, including:

[0025] For the joint points of type A, respectively obtain the set XA of the X-axis coordinates of the joint points of type A in all frames; respectively obtain the set Y-axis coordinates of the joint points of type A in all frames;

[0026] Get the minimum value xmn and the maximum value xmx in XA;

[0027] Get the minimum value ymn and the maximum value ymx in YA;

[0028] Get the straight line Lxmn which is parallel to the Y axis and passes through xmn;

[0029] Get the straight line Lxmx that is parallel to the Y axis and passes through xmx;

[0030] Get the straight line Lymn which is parallel to the X-axis and passes through ymn;

[0031] Get the straight line Lymx that is parallel to the X-axis and passes through ymx;

[0032] The closed area enclosed by Lxmn, Lxmx, Lymn and Lymx is taken as the range of motion of the joint point of type A.

[0033] Preferably, the types of joint points include left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left knee, right knee, left ankle and right ankle.

[0034] Preferably, each frame in the action video is numbered in the order of imaging time from early to late, including:

[0035] Starting from 1, each frame in the action video is numbered in the order of imaging time from early to late. The difference between the numbers of two frames with adjacent imaging time is 1. The later the imaging time, the larger the number.

[0036] Preferably, a preset foreground detection algorithm is used to and Conduct testing and obtain and The corresponding foreground image includes:

[0037] Respectively and Convert to grayscale image and ;

[0038] Use Otsu's algorithm to calculate and Segmentation thresholds Fr1 and Fr2;

[0039] Based on Fr1 and Get The corresponding foreground image;

[0040] Based on Fr2 and Get The corresponding foreground image.

[0041] Preferably, the calculation formula of the coefficient of variation is:

[0042]

[0043] express The corresponding coefficient of variation is and Respectively and The total number of pixels in the corresponding foreground image, express The center of the foreground image and The distance between the centers of the corresponding foreground images, express The width of Represents the control weight.

[0044] Beneficial effects:

[0045] Compared with the prior art, the present invention can use the information of the foreground image corresponding to the first two frames to calculate the change coefficient, and then obtain The corresponding foreground image, so that we can avoid To obtain all the pixels in The corresponding foreground image effectively reduces the number of participants The total number of pixels in the acquisition process of the corresponding foreground image can significantly improve the acquisition efficiency of the overall foreground image, thereby improving the acquisition efficiency of the range of motion of the human joints. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0047] Figure 1 It is a schematic diagram of a human joint range of motion analysis system of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The present invention provides a human joint range of motion analysis system, comprising a recording module and a video processing module;

[0050] The recording module is used to film the person being analyzed during his / her physical training to obtain action videos;

[0051] The video processing module includes a foreground detection unit and a joint recognition unit;

[0052] The foreground detection unit is used to obtain the foreground image corresponding to each frame in the action video;

[0053] The joint recognition unit is used to recognize each foreground image respectively and obtain the coordinates of the joint points in each foreground image;

[0054] The foreground image corresponding to each frame in the action video is obtained respectively, including:

[0055] Number each frame in the action video in the order of imaging time from early to late;

[0056] For frame number 1 and frame number 2 , using the preset foreground detection algorithm and Conduct testing and obtain and The corresponding foreground image;

[0057] For frame numbered q , q is greater than or equal to 2, the following method is used to Identify, obtain The corresponding foreground image:

[0058] based on and calculate The corresponding coefficient of variation;

[0059] Based on the coefficient of variation The area to be identified in ;

[0060] Use the preset foreground detection algorithm to detect the area to be identified, and obtain a set of pixels ForeU belonging to the foreground in the area to be identified;

[0061] Based on the pixel points in ForeU, the foreground completion calculation is performed to obtain The corresponding foreground image.

[0062] The above process uses the information of the foreground image corresponding to the first two frames to calculate the coefficient of change, and then obtains The corresponding foreground image, so that we can avoid To obtain all the pixels in The corresponding foreground image effectively reduces the number of participants The total number of pixels in the acquisition process of the corresponding foreground image can significantly improve the acquisition efficiency of the overall foreground image, thereby improving the acquisition efficiency of the range of motion of the human joints.

[0063] Preferably, photographing the person being analyzed includes:

[0064] The person being analyzed is photographed at a preset height.

[0065] Specifically, the shooting direction is kept unchanged during the shooting process.

[0066] Preferably, the preset height is half of the height of the person being analyzed.

[0067] Setting it to one-half can make the midpoint of the analyst's body and the midpoint of the sensor at a close height, thereby reducing the degree to which the foreground part of the acquired frame is affected by lens distortion, thereby improving the accuracy of subsequent recognition.

[0068] Preferably, the foreground image is detected using the OpenPose algorithm to obtain joint points in the foreground image.

[0069] OpenPose is an open source algorithm for real-time multi-person pose estimation. It can detect and identify the joints of the human body from pictures or videos, thereby inferring the human body's posture.

[0070] OpenPose mainly uses convolutional neural networks (CNN) to extract features from images. The specific steps include:

[0071] Image preprocessing: resize the input image to a specific size and perform normalization.

[0072] Feature extraction: Extract features from the image through a deep learning model. This usually involves multiple convolutional layers and pooling layers.

[0073] Key point regression: The extracted features are used to predict the joint points, and a heatmap is used to represent the position of each joint point.

[0074] Skeleton connection: Based on the detected joint points, these joint points are connected through defined connection rules (such as bone structure) to form a complete human skeleton.

[0075] Preferably, an analysis module is also included, and the analysis module is used to obtain the activity range of each type of joint point based on all foreground images.

[0076] Specifically, since different joints have different position change ranges during the entire physical training process, the present invention acquires the range of motion for each type of joint respectively.

[0077] Preferably, the activity range of each type of joint point is obtained based on all foreground images, including:

[0078] For the joint points of type A, respectively obtain the set XA of the X-axis coordinates of the joint points of type A in all frames; respectively obtain the set Y-axis coordinates of the joint points of type A in all frames;

[0079] Get the minimum value xmn and the maximum value xmx in XA;

[0080] Get the minimum value ymn and the maximum value ymx in YA;

[0081] Get the straight line Lxmn which is parallel to the Y axis and passes through xmn;

[0082] Get the straight line Lxmx that is parallel to the Y axis and passes through xmx;

[0083] Get the straight line Lymn which is parallel to the X-axis and passes through ymn;

[0084] Get the straight line Lymx that is parallel to the X-axis and passes through ymx;

[0085] The closed area enclosed by Lxmn, Lxmx, Lymn and Lymx is taken as the range of motion of the joint point of type A.

[0086] By obtaining the X-axis and Y-axis coordinates of the joint points in different frames, the range of motion of the joint can be determined by obtaining the minimum and maximum values ​​of the coordinates in all frames.

[0087] In the present invention, the X axis and the Y axis are parallel to the row direction and the column direction of the frame respectively. In some embodiments, the origin of the coordinate axis is the lower left corner of the frame.

[0088] The coordinates of the joint point can be represented by the average value of the coordinates of all pixels in the region to which the joint point belongs. For example, the average value of the X-axis coordinates of all pixels in the region to which the joint point belongs is used as the X-axis coordinate of the joint point.

[0089] Preferably, the types of joint points include left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left knee, right knee, left ankle and right ankle.

[0090] Preferably, each frame in the action video is numbered in the order of imaging time from early to late, including:

[0091] Starting from 1, each frame in the action video is numbered in the order of imaging time from early to late. The difference between the numbers of two frames with adjacent imaging time is 1. The later the imaging time, the larger the number.

[0092] For example, starting from the first frame after sorting, the numbers are 1, 2, 3, ... n.

[0093] Preferably, a preset foreground detection algorithm is used to and Conduct testing and obtain and The corresponding foreground image includes:

[0094] Respectively and Convert to grayscale image and ;

[0095] Use Otsu's algorithm to calculate and Segmentation thresholds Fr1 and Fr2;

[0096] Based on Fr1 and Get The corresponding foreground image;

[0097] Based on Fr2 and Get The corresponding foreground image.

[0098] Specifically, based on Fr1 and Get The corresponding foreground image includes:

[0099] Get The connected domain composed of pixels whose gray value is greater than Fr1;

[0100] The connected domain with the largest area is taken as The corresponding foreground image.

[0101] Specifically, based on Fr2 and Get The corresponding foreground image includes:

[0102] Get The connected domain composed of pixels whose gray value is greater than Fr2;

[0103] The connected domain with the largest area is taken as The corresponding foreground image.

[0104] Preferably, the calculation formula of the coefficient of variation is:

[0105]

[0106] express The corresponding coefficient of variation is and Respectively and The total number of pixels in the corresponding foreground image, express The center of the foreground image and The distance between the centers of the corresponding foreground images, express The width of Represents the control weight.

[0107] Specifically, the variation coefficient of the present invention is mainly used to reflect the variation of the foreground image between two adjacent frames. If the difference between the centers of the foreground images of two adjacent frames is greater and the difference between the number of pixels of the foreground images is greater, the variation coefficient is greater; if the difference between the centers of the foreground images of two adjacent frames is smaller and the difference between the number of pixels of the foreground images is smaller, the variation coefficient is smaller. The variation coefficient calculation method of the present invention can, in the subsequent process of acquiring the area to be identified, obtain an area to be identified with as many pixels of the foreground image as possible while making the area of ​​the area to be identified as small as possible, thereby achieving a balance between the area of ​​the area to be identified and the number of pixels belonging to the foreground image contained in the area to be identified. It can effectively reduce the subsequent use of a preset foreground detection algorithm to detect the area to be identified, and when obtaining the set ForeU of pixels belonging to the foreground in the area to be identified, avoid calculations based on all pixels, effectively control the number of pixels involved in the calculation, and effectively improve The acquisition efficiency of the corresponding foreground image process.

[0108] Preferably, the control weight is .

[0109] Preferably, based on the coefficient of variation, The areas to be identified in include:

[0110] Get The minimum value x1 and the maximum value x2 of the X-axis coordinates of the corresponding pixel points of the foreground image;

[0111] Get The minimum value y1 and the maximum value y2 of the Y-axis coordinates of the corresponding pixel points of the foreground image;

[0112] Calculate the minimum value x3 of the X-axis coordinate of the area to be identified based on the coefficient of variation:

[0113]

[0114] Calculate the maximum value x4 of the X-axis coordinate of the area to be identified based on the coefficient of variation:

[0115]

[0116] Calculate the minimum Y-axis coordinate y3 of the area to be identified based on the coefficient of variation:

[0117]

[0118] Calculate the maximum value y4 of the Y-axis coordinate of the area to be identified based on the coefficient of variation:

[0119]

[0120] but The X-axis coordinate x and Y-axis coordinate y of the pixel point in the area to be identified meet the following conditions:

[0121] .

[0122] Specifically, the present invention expands the range of the foreground image of the frame numbered q-1 through the variation coefficient. Since the shooting interval between two adjacent frames is small, the variation range of the foreground image is also relatively small. Therefore, the variation coefficient can be used to determine the approximate area where the pixel points in the foreground image of the frame numbered q are located. Then, the foreground pixel points are detected based on the area, which can greatly reduce the number of background pixel points involved in the detection process and improve the efficiency of obtaining a complete foreground image.

[0123] Preferably, a preset foreground detection algorithm is used to detect the area to be identified, and a set of pixel points belonging to the foreground in the area to be identified, ForeU, is obtained, including:

[0124] Grayscale the pixels in the area to be identified to obtain a grayscale image ;

[0125] Calculated using the Otsu algorithm The segmentation threshold Fr3;

[0126] Get The connected domain composed of pixels whose gray value is greater than Fr3;

[0127] Store the pixels in the connected domain with the largest area into the set ForeU.

[0128] Specifically, screening the connected domains by area can effectively reduce the probability that pixels in the background that are similar to the foreground are mistakenly identified as pixels belonging to the foreground.

[0129] Preferably, the foreground completion calculation is performed based on the pixel points in ForeU to obtain The corresponding foreground image includes:

[0130] Calculate the foreground completion coefficient of the pixel points in ForeU;

[0131] Based on the foreground completion coefficient, the pixel points in ForeU are calculated respectively to obtain the completion set corresponding to each pixel point in ForeU;

[0132] Get the union AU of the completed set of all pixels in ForeU;

[0133] It is composed of pixels in AU and pixels in ForeU. The corresponding foreground image.

[0134] Specifically, the reason for performing foreground completion calculation is that the pixels in ForeU may not be all the pixels in the foreground image. When the movement of the person being analyzed near the frame numbered q changes greatly, some pixels that should belong to the foreground may not be correctly included in ForeU. Therefore, the integrity of the foreground can be effectively improved through foreground completion calculation.

[0135] Preferably, the calculation formula of the foreground completion coefficient is:

[0136]

[0137] represents the foreground completion coefficient of pixel s, Represents the total number of pixels in the 8-neighborhood of pixel s that belong to the set ForeU, Represents the average grayscale value of the pixels in the 8-neighborhood of pixel s. Indicates that among the pixels in the 8-neighborhood of pixel s, the grayscale value is greater than The average gray value of all pixels. Indicates that among the pixels in the 8-neighborhood of pixel s, the gray value is less than The average gray value of all pixels; represents the first weight, represents the second weight, represents the third weight, Indicates that the gray value of the pixel in the 8-neighborhood of pixel s is greater than The point corresponding to the average coordinates of all pixels in the 8-neighborhood of pixel s has a gray value less than The distance between the points corresponding to the average coordinates of all pixels.

[0138] The foreground completion coefficient of the present invention is calculated based on the total number of pixels belonging to ForeU in the 8 neighborhoods of s, the difference between the average grayscale values ​​of the two kinds of pixels whose grayscale is greater than the average grayscale and less than the average grayscale, and the distance between the average coordinates of the two kinds of pixels whose grayscale is greater than the average grayscale and less than the average grayscale. When the total number of pixels belonging to ForeU in the 8 neighborhoods of s is small, the difference between the average grayscale values ​​of the two kinds of pixels whose grayscale is greater than the average grayscale and less than the average grayscale is larger, and the distance between the average coordinates of the two kinds of pixels whose grayscale is greater than the average grayscale and less than the average grayscale is larger, the value of the foreground completion coefficient is larger; when the total number of pixels belonging to ForeU in the 8 neighborhoods of s is large, the difference between the average grayscale values ​​of the two kinds of pixels whose grayscale is greater than the average grayscale and less than the average grayscale is smaller, and the distance between the average coordinates of the two kinds of pixels whose grayscale is greater than the average grayscale and less than the average grayscale is smaller, the value of the foreground completion coefficient is smaller. In this way, in the subsequent completion calculation process, the greater the probability that the pixel s is at the boundary of the foreground image, the more priority is given to the completion calculation of s. This can avoid the completion calculation of all pixels, greatly reducing the number of pixels involved in the subsequent completion calculation, and effectively improving the efficiency of the subsequent completion calculation.

[0139] Preferably, the first weight is , the second and third weights are .

[0140] Preferably, based on the foreground completion coefficient, the pixel points in ForeU are calculated respectively to obtain the completion set corresponding to each pixel point in ForeU, including:

[0141] Sort all pixels in ForeU according to the foreground coefficient from large to small;

[0142] The front with the largest prospect coefficient The pixel points are stored in the set FTU, where T represents the preset value;

[0143] Set the completion set corresponding to the remaining pixels in ForeU except the pixels in FTU to an empty set;

[0144] Get the completion set corresponding to each pixel point in the FTU respectively.

[0145] In this step, the present invention screens the pixels that need to be completed by setting a ratio, which can avoid completing the calculation of the pixels in ForeU that belong to the inside of the connected domain, thereby effectively reducing the possibility of invalid calculations and improving the efficiency of calculations. Because if the pixel belongs to the inside of the connected domain, then the surrounding pixels belong to the same connected domain. Completing such a pixel will repeatedly calculate more pixels in ForeU, resulting in the inability to effectively complete the foreground based on the pixels in ForeU to obtain a more complete foreground.

[0146] Preferably, the preset value is 10.

[0147] Preferably, the completion set corresponding to each pixel point in the FTU is obtained respectively, including:

[0148] Each pixel in the FTU is processed as follows:

[0149] For the pixel point ft in FTU, ft is processed as follows:

[0150] The first step is to use ft as the seed pixel;

[0151] The second step is to store the seed pixel point into the set CSU, and obtain the set SEDU of the pixel points in the 8-domain of the seed pixel point that do not belong to the set ForeU;

[0152] The third step is to calculate the absolute value of the difference between the grayscale value of each pixel in SEDU and the seed pixel;

[0153] Step 4: Get the minimum absolute value;

[0154] Step 5: determine whether the minimum absolute value is less than the set absolute value threshold. If so, proceed to step 6; if not, proceed to step 7.

[0155] Step 6: Take the pixel corresponding to the smallest absolute value in SEDU as the next seed pixel and proceed to step 2;

[0156] Step 7: Use the pixels in CSU as the completion set corresponding to ft.

[0157] The present invention utilizes the idea of ​​region growing when obtaining the completion set. However, each time the present invention selects the next seed pixel point, it avoids selecting from ForeU. In this way, except for ft, the pixel points in the completion set obtained by performing region growing calculation on ft are all pixel points that do not belong to ForeU. Therefore, when ForeU does not contain all the pixels belonging to the foreground, the foreground is effectively extended, which is conducive to obtaining a more complete foreground image.

[0158] Specifically, the absolute value threshold is set as One fifth of the average grayscale value of the pixels in .

[0159] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A human joint range of motion analysis system, characterized in that: Including recording module and video processing module; The recording module is used to film the person being analyzed during his / her physical training to obtain action videos; The video processing module includes a foreground detection unit and a joint recognition unit; The foreground detection unit is used to obtain the foreground image corresponding to each frame in the action video; The joint recognition unit is used to recognize each foreground image respectively and obtain the coordinates of the joint points in each foreground image; The foreground image corresponding to each frame in the action video is obtained respectively, including: Number each frame in the action video in the order of imaging time from early to late; For frame number 1 and frame number 2 , using the preset foreground detection algorithm and Conduct testing and obtain and The corresponding foreground image; For frame numbered q , q is greater than or equal to 2, the following method is used to Identify, obtain The corresponding foreground image: based on and calculate The corresponding coefficient of variation; Based on the coefficient of variation The area to be identified in ; Use the preset foreground detection algorithm to detect the area to be identified, and obtain a set of pixels ForeU belonging to the foreground in the area to be identified; Based on the pixel points in ForeU, the foreground completion calculation is performed to obtain The corresponding foreground image; The coefficient of variation is calculated as: express The corresponding coefficient of variation is and Respectively and The total number of pixels in the corresponding foreground image, express The center of the foreground image and The distance between the centers of the corresponding foreground images, express The width of represents the control weight; Based on the coefficient of variation The areas to be identified in include: Get The minimum value x1 and the maximum value x2 of the X-axis coordinates of the corresponding pixel points of the foreground image; Get The minimum value y1 and the maximum value y2 of the Y-axis coordinates of the corresponding pixel points of the foreground image; Calculate the minimum value x3 of the X-axis coordinate of the area to be identified based on the coefficient of variation: Calculate the maximum value x4 of the X-axis coordinate of the area to be identified based on the coefficient of variation: Calculate the minimum Y-axis coordinate y3 of the area to be identified based on the coefficient of variation: Calculate the maximum value y4 of the Y-axis coordinate of the area to be identified based on the coefficient of variation: but The X-axis coordinate x and Y-axis coordinate y of the pixel point in the area to be identified meet the following conditions: 。 2. A human joint range of motion analysis system according to claim 1, characterized in that: Filming of the person being analyzed, including: The person being analyzed is photographed at a preset height.

3. A human joint range of motion analysis system according to claim 2, characterized in that: The preset height is half of the height of the person being analyzed.

4. A human joint range of motion analysis system according to claim 1, characterized in that: The OpenPose algorithm is used to detect the foreground image and obtain the joint points in the foreground image.

5. A human joint range of motion analysis system according to claim 1, characterized in that: The method also includes an analysis module, which is used to obtain the activity range of each type of joint point based on all foreground images.

6. A human joint range of motion analysis system according to claim 4, characterized in that: Based on all foreground images, the range of motion of each type of joint point is obtained separately, including: For the joint points of type A, respectively obtain the set XA of the X-axis coordinates of the joint points of type A in all frames; respectively obtain the set Y-axis coordinates of the joint points of type A in all frames; Get the minimum value xmn and the maximum value xmx in XA; Get the minimum value ymn and the maximum value ymx in YA; Get the straight line Lxmn which is parallel to the Y axis and passes through xmn; Get the straight line Lxmx that is parallel to the Y axis and passes through xmx; Get the straight line Lymn which is parallel to the X-axis and passes through ymn; Get the straight line Lymx that is parallel to the X-axis and passes through ymx; The closed area enclosed by Lxmn, Lxmx, Lymn and Lymx is taken as the range of motion of the joint point of type A.

7. A human joint range of motion analysis system according to claim 1, characterized in that: The types of joint points include left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left knee, right knee, left ankle, and right ankle.

8. A human joint range of motion analysis system according to claim 1, characterized in that: According to the imaging time from early to late, each frame in the action video is numbered, including: Starting from 1, each frame in the action video is numbered in the order of imaging time from early to late. The difference between the numbers of two adjacent frames in imaging time is 1. The later the imaging time, the larger the number.

9. A human joint range of motion analysis system according to claim 1, characterized in that: Use the preset foreground detection algorithm to and Conduct testing and obtain and The corresponding foreground image includes: Respectively and Convert to grayscale image and ; Use Otsu's algorithm to calculate and Segmentation thresholds Fr1 and Fr2; Based on Fr1 and Get The corresponding foreground image; Based on Fr2 and Get The corresponding foreground image.

Images