WO2021000401A1 - 体态评估方法、电子装置、计算机设备及存储介质 - Google Patents

体态评估方法、电子装置、计算机设备及存储介质 Download PDF

Info

Publication number
WO2021000401A1
WO2021000401A1 PCT/CN2019/102793 CN2019102793W WO2021000401A1 WO 2021000401 A1 WO2021000401 A1 WO 2021000401A1 CN 2019102793 W CN2019102793 W CN 2019102793W WO 2021000401 A1 WO2021000401 A1 WO 2021000401A1
Authority
WO
WIPO (PCT)
Prior art keywords
vector
shoulder
calculate
coordinates
angle
Prior art date
Application number
PCT/CN2019/102793
Other languages
English (en)
French (fr)
Inventor
王义文
王健宗
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2021000401A1 publication Critical patent/WO2021000401A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/107—Measuring physical dimensions, e.g. size of the entire body or parts thereof
    • A61B5/1071—Measuring physical dimensions, e.g. size of the entire body or parts thereof measuring angles, e.g. using goniometers
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/107—Measuring physical dimensions, e.g. size of the entire body or parts thereof
    • A61B5/1079—Measuring physical dimensions, e.g. size of the entire body or parts thereof using optical or photographic means

Definitions

  • This application relates to the field of computer vision technology, and in particular to a posture assessment method, electronic device, computer equipment and storage medium.
  • this method is difficult to obtain comprehensive and accurate results, and it is often time-consuming and laborious to find a way for professionals to perform the assessment.
  • this application proposes a posture assessment method, electronic device, computer equipment, and storage medium, which can improve the comprehensiveness and accuracy of posture assessment and is simple to operate.
  • this application proposes a posture evaluation method, which includes the steps of: acquiring an image to be tested, the image to be tested includes a full body image of the front and a full body side of the tester standing upright; Extract bone key points from the image; calculate the tester's posture vector according to the bone key points; and obtain the bending angle of the tester's posture vector.
  • the present application also provides an electronic device, which includes: an acquisition module adapted to acquire an image to be tested, the image to be tested includes a frontal full-body image and a lateral full-body image of the tester standing upright; an extraction module , Suitable for extracting bone key points from the image to be tested; a calculation module, suitable for calculating the tester's pose vector according to the bone key points; and an evaluation module, suitable for obtaining the bending angle of the tester's pose vector.
  • an acquisition module adapted to acquire an image to be tested, the image to be tested includes a frontal full-body image and a lateral full-body image of the tester standing upright
  • an extraction module Suitable for extracting bone key points from the image to be tested
  • a calculation module suitable for calculating the tester's pose vector according to the bone key points
  • an evaluation module suitable for obtaining the bending angle of the tester's pose vector.
  • this application also provides a computer device, including a memory, a processor, and computer-readable instructions stored in the memory and running on the processor, and the processor executes the computer-readable instructions When implementing the steps of the above method.
  • the present application also provides a non-volatile computer-readable storage medium on which computer-readable instructions are stored, and the computer-readable instructions implement the steps of the foregoing method when executed by a processor.
  • the posture evaluation method, electronic device, computer equipment and storage medium proposed in this application can extract key bone points from the tester’s image to be tested, calculate the tester’s posture vector according to the bone key points, and obtain the tester’s posture
  • the bending angle of the vector, and then performing posture evaluation based on the bending angle can improve the comprehensiveness and accuracy of posture evaluation, and the operation is simple.
  • Fig. 1 is a schematic diagram of a posture assessment process according to an exemplary embodiment of the present application
  • Fig. 2 is a schematic flow chart of posture assessment shown in an exemplary embodiment of the present application
  • Fig. 3 is a schematic diagram of bone key points shown in an exemplary embodiment of the present application.
  • Fig. 4 is a schematic diagram of a posture assessment process according to an exemplary embodiment of the present application.
  • Fig. 5 is a schematic diagram of a posture assessment process according to an exemplary embodiment of the present application.
  • Fig. 6 is a schematic diagram of a posture assessment process according to an exemplary embodiment of the present application.
  • FIG. 7 is a schematic diagram showing the height of shoulders according to an exemplary embodiment of the present application.
  • Fig. 8 is a schematic diagram of a humpback shown in an exemplary embodiment of the present application.
  • Fig. 9 is a schematic diagram of scoliosis according to an exemplary embodiment of the present application.
  • Fig. 10 is a schematic diagram of an O-shaped leg and an X-shaped leg shown in an exemplary embodiment of the present application;
  • Fig. 11 is a schematic diagram of a posture assessment process according to an exemplary embodiment of the present application.
  • FIG. 12 is a schematic diagram of program modules of an electronic device shown in an exemplary embodiment of the present application.
  • FIG. 13 is a schematic diagram of the hardware architecture of an electronic device according to an exemplary embodiment of the present application.
  • Fig. 1 is a schematic flowchart of a posture assessment method according to an embodiment of the present application. The method includes the following steps:
  • Step S110 Obtain an image to be tested, the image to be tested includes a frontal full-body image and a lateral full-body image of the tester standing upright;
  • Step S120 extract bone key points from the image to be tested
  • Step S130 calculating the pose vector of the tester according to the bone key points.
  • Step S140 Obtain the bending angle of the tester's posture vector.
  • the shoulders on both sides are not on the same level, it may be high and low shoulders. However, it is often difficult to obtain comprehensive and accurate results from the self-test. Because height and low shoulders generally do not have a serious impact on life, most people usually do not feel the need to seek professional evaluation.
  • the posture assessment can be performed based on several full-body images provided by the tester standing upright, with simple operation and high accuracy.
  • step S110 an image to be tested is acquired, and the image to be tested includes a frontal full-body image and a lateral full-body image of the tester standing upright.
  • posture assessment is a set of methodology, a connection established between the function and form of our body structure. For example, when some of our muscles are dysfunctional, then, from the morphological point of view, these muscles may be lengthened or shortened, and the morphology of the bones and joints connected will also change accordingly. Posture assessment is to capture these morphological changes in order to infer the specific conditions of the dysfunction.
  • posture assessment does not limit in which posture the assessment is performed, nor does it limit whether it is static or dynamic, we generally think that posture assessment refers to static stance assessment.
  • the standing posture reveals a lot of information about the maintenance of body posture, and it is overall information.
  • the image to be tested for posture assessment is a full-body image of the tester standing upright, and may include a front full-body image, a side full-body image, a back full-body image, and the like.
  • Simple and reasonable environment settings can simplify the evaluation process and improve the accuracy of the evaluation.
  • the image to be tested can be taken under the guidance of a professional; it can also be taken by the tester using a device with a shooting function such as a mobile phone, after making corresponding actions under voice prompts.
  • This application does not determine the source of the image to be tested. limited.
  • step S120 bone key points are extracted from the image to be tested.
  • an embodiment of the application adopts The posture assessment is based on the key points of the bones.
  • the step of extracting bone key points from the image to be tested may include the following steps:
  • Step S201 input the image to be measured into a neural network, and estimate the key point heat map of the image to be measured based on a human body pose estimation algorithm;
  • Step S202 Calculate the Gaussian value of the hot spot at each position in the key point heat map, and select the hot spot where the peak of the Gaussian value is located as the bone key point of the position.
  • the image to be tested is input to a neural network (for example, a convolutional neural network), the image to be tested is processed by the convolutional neural network to generate a feature atlas F, and then the network is trained before the visual geometry group (Visual Geometry Group pre-train network, VGG pre-train network) As a skeleton neural network, it respectively regresses the key point position and the direction of the key point to output the key point heat map, as shown in Figure 3, which can output 25 The location of the key points of the bones.
  • the bone key points may include: head, neck, torso center, left shoulder, left elbow, left wrist, left hip, left knee, left ankle, right shoulder, right elbow, right wrist, right hip , Right knee, right ankle, etc.
  • the neural network estimates the key point heat map of the image to be measured according to the human body posture estimation algorithm, and uses the hot spot of the Gaussian peak in the heat map as the bone key point of the position. For example, predicting the position of the right shoulder of the human body on the image to be tested, the detection result is obtained by predicting the heat map of the key points of the human body, calculating the Gaussian value of each hot spot in the heat map of the right shoulder position, and selecting the hot spot where the peak of the Gaussian value is located As the key point of the right shoulder bone.
  • Each key point of the human body is the Gaussian peak at that position, which means that the neural network believes that there is a key point of the human bone. Do similar processing to other positions, such as the right elbow, to get the bone key points of the corresponding position.
  • connection the bone key points After obtaining the bone key points, connect the bone key points to determine the connection relationship between each bone key point. In particular, when there is more than one person in the image to be tested, it can be determined that each bone key point belongs to the picture. Which one of them.
  • the step of calculating the pose vector of the tester according to the bone key points may include the following steps:
  • Step S401 connecting the key bone points based on the posture of the human body
  • Step S402 obtaining the coordinates of the key points of the skeleton.
  • Step S403 Calculate the limb vectors of the two connected bone key points based on the coordinates, and generate the pose vector of the tester according to the limb vectors.
  • the human body key point affinity field (Part Affinity Fields, referred to as PAFs) is used to infer the connection with other bone key points, and this step is repeated until the human body is obtained.
  • PAFs Part Affinity Fields
  • Key points of all bones Obtain the coordinates of each bone key point, calculate the limb vectors of the two connected bone key points based on the coordinates, and generate the pose vector of the tester according to the limb vectors.
  • a coordinate system In an embodiment of this application, horizontal to the right is the positive direction of the X axis, vertical upward is the positive direction of the Y axis, and the vertical frame (frontal full-body image) inward is recorded as the positive direction of the Z axis.
  • Figure 3 the coordinates of the key points of each bone of the tester can be obtained according to the pixel points.
  • the coordinates of the key points of each bone are marked as Where n represents the serial number of the bone key point, so that the limb vector of the two connected bone key points can be calculated based on the coordinates.
  • the step of calculating the limb vectors of two connected bone key points based on the coordinates may include the following steps:
  • Step S501 calculating a left shoulder vector based on the coordinates of the neck and left shoulder, and calculating a right shoulder vector based on the coordinates of the neck, the neck and the left shoulder;
  • Step S502 calculating a waist vector based on the coordinates of the neck and the torso center;
  • Step S503 calculating a neck vector based on the coordinates of the neck and the head;
  • Step S504 calculating the left thigh vector based on the coordinates of the left hip and left knee, and calculating the right thigh vector based on the coordinates of the right hip and right knee;
  • a left calf vector is calculated based on the coordinates of the left knee and left ankle
  • a right calf vector is calculated based on the coordinates of the right knee and right ankle.
  • bone key point 0 is the head
  • bone key point 1 is the neck
  • bone key point 2 is the right shoulder
  • bone key point 5 is the left shoulder
  • bone key point 8 is the torso center
  • bone key point 9 is the right hip.
  • the bone key point 10 is the right knee
  • the bone key point 11 is the right ankle
  • the bone key point 12 is the left hip
  • the bone key point 13 is the left knee
  • the bone key point 14 is the left ankle
  • the sitting mark of each bone key point is
  • n represents the serial number of the bone key point.
  • the step of obtaining the bending angle of the tester's posture vector may include the following steps:
  • Step S601 Calculate the shoulder angle to be measured based on the left shoulder vector and the right shoulder vector, and obtain the shoulder bending angle based on the angular relationship between the measured shoulder angle and the standard shoulder angle;
  • Step S602 calculating the to-be-tested spine angle based on the waist vector and/or neck vector, and obtaining the spine bending angle based on the angular relationship between the to-be-tested spine angle and the standard spine angle;
  • Step S603 Calculate the leg angle to be measured based on the left thigh vector, left calf vector, right thigh vector, and right calf vector, based on the measured leg angle and the standard leg The angle relationship of the included angle obtains the bending angle of the leg;
  • Step S604 Obtain the bending angle of the tester's posture vector based on the shoulder bending angle, spine bending angle, and leg bending angle.
  • L 12 is the tester's right shoulder vector
  • L 15 is the tester's left shoulder vector
  • B 12 is the standard model's right shoulder vector
  • B 15 is the standard model's left shoulder vector.
  • S j exceeds a preset threshold (for example, 5 degrees), it means that the tester's shoulders have a larger tilt compared with the standard posture, and the tester's shoulders are assessed as high and low shoulders.
  • a preset threshold for example, 5 degrees
  • kyphosis refers to a person’s spine arching backward, and the spine bending angle can be obtained according to the kyphosis evaluation model, which is as follows:
  • Is the tester’s neck vector The vectors in the Y-axis and Z-axis directions
  • Is the waist vector of the tester The vectors in the Y-axis and Z-axis directions
  • Neck vector for standard model The vectors in the Y-axis and Z-axis directions
  • Waist vector for standard model The vector in the Y-axis and Z-axis directions.
  • S b If the value of S b exceeds a preset threshold value (for example, 5 degrees), it means that compared with the standard posture, the tester's back is arched backward and is evaluated as a hunchback.
  • a preset threshold value for example, 5 degrees
  • scoliosis refers to the curvature of the spine caused by the deviation of the spine to the side.
  • the angle of spine curvature can be obtained according to the scoliosis evaluation model, which is as follows:
  • Is the waist vector of the tester The vectors in the X-axis and Y-axis directions
  • Waist vector for standard model The vector in the X-axis and Y-axis directions.
  • the value of S z exceeds the preset threshold value (for example, 8 degrees), it means that the tester has a larger scoliosis compared with the standard posture, and it is evaluated as scoliosis.
  • the preset threshold value for example, 8 degrees
  • leg angle is calculated according to the leg evaluation model
  • the tester’s leg angle to be measured is calculated according to the leg evaluation model as follows:
  • Is the tester’s right thigh vector The vectors in the X-axis and Y-axis directions
  • Is the tester’s right calf vector The vectors in the X-axis and Y-axis directions
  • Is the tester’s left thigh vector The vectors in the X-axis and Y-axis directions
  • Is the tester’s left calf vector The vector in the X-axis and Y-axis directions.
  • the standard leg angle of the standard model is calculated as follows:
  • Is the right thigh vector of the standard model The vectors in the X-axis and Y-axis directions
  • Is the right calf vector of the standard model The vectors in the X-axis and Y-axis directions
  • Is the left thigh vector of the standard model The vectors in the X-axis and Y-axis directions
  • Is the left calf vector of the standard model The vector in the X-axis and Y-axis directions.
  • S t -S' t > ⁇ ( ⁇ is a preset threshold, for example, 8 degrees), it means that compared with the standard posture, the tester's legs are bent outward, and it is evaluated as an O-shaped leg; if S t -S' t ⁇ ( ⁇ is a preset threshold, for example, 10 degrees), it means that compared with the standard posture, the tester’s legs are bent to the inside and are evaluated as X-shaped legs.
  • the bending angle of the tester's posture vector is obtained based on the shoulder bending angle, the spine bending angle, and the leg bending angle.
  • the corresponding standard posture can also be selected as the evaluation standard according to the sex and age of the tester.
  • the step of performing posture evaluation according to the tester's posture vector may include the following steps:
  • Step S111 when the image to be tested does not meet the test requirements, a prompt message is sent to prompt the tester to adjust the posture.
  • the image to be tested can be taken by the tester through a mobile phone or other device with shooting function, after making corresponding actions under voice prompts. Therefore, after the image to be tested is taken, the image to be tested can be analyzed to determine whether the tester is standing upright, whether the image to be tested is a full-body image, etc. If the image to be tested does not conform to the test If required, the tester can also be prompted to adjust the posture by sending a prompt message to improve the accuracy of the test.
  • a prompt message can be sent intermittently to remind the user not to lift his legs, get up after sitting for a long time, etc. , To remind the tester to improve bad posture.
  • the posture assessment method proposed in this application can extract key bone points from the tester’s image to be tested, calculate the tester’s posture vector according to the bone key points, obtain the bending angle of the tester’s posture vector, and then according to the The posture evaluation of the tester's posture vector can improve the comprehensiveness and accuracy of the posture evaluation, and the operation is simple.
  • the application further provides an electronic device.
  • FIG. 12 is a schematic diagram of program modules of an electronic device 20 according to an exemplary embodiment of the present application.
  • the electronic device 20 includes:
  • the acquiring module 201 is adapted to acquire an image to be tested, and the image to be tested includes a frontal full-body image and a lateral full-body image of the tester standing upright;
  • the extraction module 202 is adapted to extract key bone points from the image to be tested
  • the calculation module 203 is adapted to calculate the pose vector of the tester according to the bone key points.
  • the evaluation module 204 is adapted to obtain the bending angle of the tester's posture vector.
  • the extraction module 202 includes: an estimation unit, adapted to input the image to be measured into a neural network, and estimate the key point heat map of the image to be measured based on a human posture estimation algorithm; and a first calculation unit, It is suitable for calculating the Gaussian value of the hot spot of each position in the key point heat map, and selecting the hot spot where the peak value in the Gaussian value is located as the bone key point of the position.
  • the key bone points include: head, neck, torso center, left shoulder, left elbow, left wrist, left hip, left knee, left ankle, right shoulder, right elbow, right wrist, right hip, right knee, right ankle.
  • the calculation module 203 includes: a connecting unit adapted to connect the bone key points based on the posture of the human body; an acquisition unit adapted to obtain the coordinates of the bone key points; and a second calculation unit adapted to connect the bone key points based on the The coordinates are calculated for the limb vectors of the two key points of the bones connected, and the pose vector of the tester is generated according to the limb vectors.
  • the second calculation unit is further adapted to calculate the left shoulder vector based on the coordinates of the neck and left shoulder, calculate the right shoulder vector based on the coordinates of the neck and right shoulder; calculate the waist vector based on the coordinates of the center of the neck and the torso Calculate the neck vector based on the coordinates of the neck and head; calculate the left thigh vector based on the coordinates of the left hip and left knee; calculate the right thigh vector based on the coordinates of the right hip and right knee; and based on the left knee and left
  • the ankle coordinates calculate the left calf vector, and the right calf vector based on the coordinates of the right knee and right ankle.
  • the evaluation module 204 includes: a third calculation unit adapted to calculate the included angle of the to-be-measured shoulder based on the left shoulder vector and the right-shoulder vector, based on the included angle of the to-be-measured shoulder and the standard shoulder The angle relationship of the angle to obtain the shoulder bending angle; calculate the measured spine angle based on the waist vector and/or the neck vector, and obtain the spine bending angle based on the angular relationship between the measured spine angle and the standard spine angle; The left thigh vector, left calf vector, right thigh vector and right calf vector calculate the angle of the leg to be measured, based on the angle between the angle of the leg to be measured and the angle of the standard leg Relationship acquiring the bending angle of the leg; and an evaluation unit adapted to acquire the bending angle of the tester's posture vector based on the shoulder bending angle, the spine bending angle, and the leg bending angle.
  • the electronic device 20 further includes a prompt module, which is adapted to send prompt information to prompt the tester to adjust the posture when the image to be tested does not meet the test requirements.
  • the electronic device proposed in this application can extract key bone points from the tester’s image to be tested, calculate the tester’s posture vector based on the bone key points, obtain the bending angle of the tester’s posture vector, and then according to the bending Performing posture evaluation from an angle can improve the comprehensiveness and accuracy of posture evaluation, and the operation is simple.
  • the present application also provides a computer device 20, including a memory 21, a processor 22, and computer-readable instructions stored in the memory 21 and running on the processor 22, When the processor 22 executes the computer-readable instructions, the steps of the foregoing method are implemented.
  • the computer-readable instructions may be stored in the memory 24.
  • the present application also provides a non-volatile computer-readable storage medium on which computer-readable instructions are stored, and the computer-readable instructions implement the steps of the foregoing method when executed by a processor.
  • This application also provides a computer device, such as a smart phone, a tablet computer, a notebook computer, a desktop computer, a rack server, a blade server, a tower server or a cabinet server (including independent servers, or more A server cluster composed of two servers), etc.
  • the computer device in this embodiment at least includes, but is not limited to: a memory, a processor, etc. that can be communicatively connected to each other through a system bus.
  • This embodiment also provides a non-volatile computer-readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory ( SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, server, App application mall, etc., on which storage There are computer-readable instructions, which implement corresponding functions when executed by the processor.
  • the non-volatile computer-readable storage medium of this embodiment is used to store the electronic device 20, and when executed by the processor 22, realizes the posture assessment method of the present application.
  • the method of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better. ⁇
  • the technical solution of this application essentially or the part that contributes to the existing technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk, The optical disc) includes several instructions to enable a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the method described in each embodiment of the present application.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Engineering & Computer Science (AREA)
  • Dentistry (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Physics & Mathematics (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Image Analysis (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

本申请公开了一种体态评估方法、电子装置、计算机设备及存储介质，通过获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；从所述待测图像中提取骨骼关键点；根据所述骨骼关键点计算测试者姿态向量；获取所述测试者姿态向量的弯曲角度。本申请所提出的体态评估方法、电子装置、计算机设备及存储介质，能够从测试者的待测图像中提取骨骼关键点，根据所述骨骼关键点计算测试者姿态向量，获取所述测试者姿态向量的弯曲角度，进而根据所述弯曲角度进行体态评估，能够提高体态评估的全面性及准确性，且操作简单。

Description

体态评估方法、电子装置、计算机设备及存储介质
本申请要求于2019年7月4日提交中国专利局，专利名称为“体态评估方法、电子装置、计算机设备及存储介质”，申请号为201910599978.7的发明专利的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机视觉技术领域，尤其涉及一种体态评估方法、电子装置、计算机设备及存储介质。
背景技术
现代人在生活中多少都会有一些不良体态的行为，驼背、高低肩、脊柱侧弯、O型腿、X型腿等都属于不良体态。而不良体态不仅会影响人的体型与气质，长期的不良体态还会导致伤痛与骨骼变形，影响人的身体健康，例如，身体僵硬、慢性疼痛、肌肉劳损、骨刺、椎间盘突出等。
发明人发现，目前的体态评估，通常是用户根据资料的描述进行自测并评估，但是，该方法难以得出全面且准确的结果，而寻找专业人士进行评估的方式通常费时费力。
发明内容
有鉴于此，本申请提出一种体态评估方法、电子装置、计算机设备及存储介质，能够提高体态评估的全面性及准确性，且操作简单。
首先，为实现上述目的，本申请提出一种体态评估方法，该方法包括步骤：获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；从所述待测图像中提取骨骼关键点；根据所述骨骼关键点计算 测试者姿态向量；及获取所述测试者姿态向量的弯曲角度。
此外，为实现上述目的，本申请还提供一种电子装置，其包括：获取模块，适于获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；提取模块，适于从所述待测图像中提取骨骼关键点；计算模块，适于根据所述骨骼关键点计算测试者姿态向量；及评估模块，适于获取所述测试者姿态向量的弯曲角度。
为实现上述目的，本申请还提供一种计算机设备，包括存储器、处理器以及存储在存储器上并可在所述处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现上述方法的步骤。
为实现上述目的，本申请还提供非易失性计算机可读存储介质，其上存储有计算机可读指令，所述计算机可读指令被处理器执行时实现上述方法的步骤。
本申请所提出的体态评估方法、电子装置、计算机设备及存储介质，能够从测试者的待测图像中提取骨骼关键点，根据所述骨骼关键点计算测试者姿态向量，获取所述测试者姿态向量的弯曲角度，进而根据所述弯曲角度进行体态评估，能够提高体态评估的全面性及准确性，且操作简单。
附图说明
图1是本申请一示例性实施例示出的体态评估的流程示意图；
图2是本申请一示例性实施例示出的体态评估的流程示意图；
图3是本申请一示例性实施例示出的骨骼关键点示意图；
图4是本申请一示例性实施例示出的体态评估的流程示意图；
图5是本申请一示例性实施例示出的体态评估的流程示意图；
图6是本申请一示例性实施例示出的体态评估的流程示意图；
图7是本申请一示例性实施例示出的高低肩示意图；
图8是本申请一示例性实施例示出的驼背示意图；
图9是本申请一示例性实施例示出的脊柱侧弯示意图；
图10是本申请一示例性实施例示出的O型腿和X型腿示意图；
图11是本申请一示例性实施例示出的体态评估的流程示意图；
图12是本申请一示例性实施例示出的电子装置的程序模块示意图；
图13是本申请一示例性实施例示出的电子装置的硬件架构示意图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
需要说明的是，在本申请中涉及“第一”、“第二”等的描述仅用于描述目的，而不能理解为指示或暗示其相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括至少一个该特征。另外，各个实施例之间的技术方案可以相互结合，但是必须是以本领域普通技术人员能够实现为基础，当技术方案的结合出现相互矛盾或无法实现时应当认为这种技术方案的结合不存在，也不在本申请要求的保护范围之内。
参阅图1所示，是本申请一实施例之体态评估方法的流程示意图，所述方法包括以下步骤：
步骤S110，获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；
步骤S120，从所述待测图像中提取骨骼关键点；
步骤S130，根据所述骨骼关键点计算测试者姿态向量；及
步骤S140，获取所述测试者姿态向量的弯曲角度。
现代人在生活中多少都会有一些不良体态的行为，驼背、高低肩、脊柱侧弯、O型腿、X型腿等都属于不良体态。由于不良体态通常不会对人的生活造成严重影响，因此，重视程度较低，即便想知道自己是否存在不良体态，大部分人也只是根据资料介绍的方法自己来判断是否存在不良体态的情况，仅极少数人会寻找专业人士进行评估。例如，当想知道自己是否有高低肩时，可以采用自照镜子检查的方式：把上衣脱去，放松双肩，自然站立，然后观察镜子中的自己，肩膀是否存在一高一低的现象。如果两边的肩膀不在同一水平线上，就有可能是高低肩了。但自测的方式通常难以得出全面且准确的结果，由于高低肩一般不会对生活造成严重影响，因此，大部分人通常也不会觉得需要寻找专业人士进行评估。
本申请一实施例中，可以根据测试者提供的若干张立正站立的全身图像进行体态评估，操作简单，准确度高。
在步骤S110中，获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像。
从本质上讲，体态评估是一套方法论，是在我们身体结构的功能和形态之间建立的联系。比如，当我们的某些肌肉出现功能紊乱，那么，从形态上看，这些肌肉可能被拉长或缩短，其连接的骨骼和关节的形态也会发生相应的变化。体态评估正是要捕捉到这些形态上的变化，以此推断功能紊乱的具体情况。
虽然体态评估并没有限定是在哪种体态下进行评估，也没有限定是静态还是动态，但我们一般认为体态评估指的是静态站姿评估。站姿中透露了很多关于身体姿势维持的信息，而且是整体性的信息。
因此，本申请一实施例中，用于进行体态评估的待测图像，是测试者立正站立的全身图像，可以包括正面全身图像、侧面全身图像、背面全身图像 等。
体态评估有四个原则：
1)、要拍照
由于体态维持是动态的，即便是在静止站立的状态下，仍然有微小的动作。通过拍照的方法，记录瞬间的体态表现，能够更好的评估和分析。
2)、要在自然状态下评估
在日常生活中，各种体态的维持都是下意识的，比如自然站立时，不会刻意收缩哪块肌肉来保持平衡。为了更准确地反映真实状态，在下意识的状态下进行评估是很必要的。
3)、要整体评估
整体评估以便发现不良体态的根源。
4)、要关注评估环境
简单合理的环境设置能简化评估过程，还能提高评估的准确性。
因此，所述待测图像可以是在专业人员指导下拍摄；也可以是测试者通过手机等具有拍摄功能的设备，在语音提示下做出相应动作后拍摄的，本申请对待测图像的来源不作限定。
在步骤S120中，从所述待测图像中提取骨骼关键点。
由于测试者的胖瘦会影响待测图像的视觉效果，而实际上，不良体态是人的骨骼形态异常，为了尽可能降低肌肉和脂肪等对评估结果产生的影响，本申请一实施例采用的是通过骨骼关键点进行体态评估。
如图2所示，本申请一实施例中，所述从所述待测图像中提取骨骼关键点的步骤可以包括以下步骤：
步骤S201，将所述待测图像输入神经网络，基于人体姿态估计算法预估所述待测图像的关键点热点图；及
步骤S202，计算所述关键点热点图中各位置的热点的高斯值，选取高斯值中的峰值所在的热点作为该位置的骨骼关键点。
本申请一实施例中，将所述待测图像输入神经网络(例如，卷积神经网络)，待测图像被卷积神经网络处理后生成特征图集F，然后进入以视觉几何组前训练网络(Visual Geometry Group pre-train network，VGG pre-train network)作为骨架的神经网络，分别对关键点位置和关键点的走向进行回归，从而输出关键点热点图，如图3所示，可以输出25个骨骼关键点的位置。本申请一实施例中，所述骨骼关键点可以包括：头部、颈部、躯干中心、左肩、左肘、左腕、左臀、左膝、左踝、右肩、右肘、右腕、右臀、右膝、右踝等。
神经网络根据人体姿态估计算法预估所述待测图像的关键点热点图，以热点图中的高斯峰值所在的热点作为该位置的骨骼关键点。比如预估待测图像上人体右肩膀的位置，得到的检测结果是通过预测人体关键点的热点图，计算右肩膀位置的热点图中各热点的高斯值，选取高斯值中的峰值所在的热点作为右肩膀的骨骼关键点。每个人体关键点都是该位置处的高斯峰值，代表神经网络相信这里有一个人体的骨骼关键点。对其他的位置，比如说右肘作类似的处理，可以得到相应位置的骨骼关键点。
在得到骨骼关键点之后，对骨骼关键点进行连接，以确定每个骨骼关键点间的连接关系，特别地，当待测图像中不止一个人时，可以确定每个骨骼关键点具体是属于图片中哪个人的。
如图4所示，本申请一实施例中，所述根据所述骨骼关键点计算测试者姿态向量的步骤可以包括以下步骤：
步骤S401，基于人体姿态连接所述骨骼关键点；
步骤S402，获取所述骨骼关键点的坐标；及
步骤S403，基于所述坐标计算相连接的两个骨骼关键点的肢体向量，根据所述肢体向量生成测试者姿态向量。
本申请一实施例中采用的是，根据一个骨骼关键点，通过人体关键点亲和场(Part Affinity Fields，简称PAFs)推测与其他的骨骼关键点之间的连接，重复这个步骤，直到得到人体的全部骨骼关键点。获取各骨骼关键点的坐标， 基于所述坐标计算相连接的两个骨骼关键点的肢体向量，根据所述肢体向量生成测试者姿态向量。
首先建立坐标系，本申请一实施例中，以水平向右为X轴正方向，以竖直向上为Y轴正方向，以垂直画面(正面全身图像)向内记为Z轴正方向，参考图3，可以根据像素点获取测试者各骨骼关键点的坐标，各骨骼关键点的坐标记为
Figure PCTCN2019102793-appb-000001
其中n表示骨骼关键点的序号，从而可以基于所述坐标计算相连接的两个骨骼关键点的肢体向量。
如图5所示，本申请一实施例中，所述基于所述坐标计算相连接的两个骨骼关键点的肢体向量的步骤可以包括以下步骤：
步骤S501，基于颈部与左肩的坐标计算左肩部向量，基于颈部与颈部与左肩的坐标计算右肩部向量；
步骤S502，基于颈部与躯干中心的坐标计算腰部向量；
步骤S503，基于颈部与头部的坐标计算颈部向量；
步骤S504，基于左臀与左膝的坐标计算左大腿部向量，基于右臀与右膝的坐标计算右大腿部向量；及
步骤S504，基于左膝与左踝的坐标计算左小腿部向量，基于右膝与右踝的坐标计算右小腿部向量。
参考图3，骨骼关键点0为头部，骨骼关键点1为颈部，骨骼关键点2为右肩，骨骼关键点5左肩，骨骼关键点8为躯干中心，骨骼关键点9为右臀，骨骼关键点10为右膝，骨骼关键点11为右踝，骨骼关键点12为左臀，骨骼关键点13为左膝，骨骼关键点14为左踝，各骨骼关键点的坐标记为
Figure PCTCN2019102793-appb-000002
其中n表示骨骼关键点的序号。本申请一实施例中，可以通过以下公式计算测试者的肢体向量：颈部向量为L 10＝P 1-P 0，右肩部向量为L 12＝P 1-P 2，左肩部向量为L 15＝P 1-P 5，腰部向量为L 81＝P 8-P 1，右大腿部向量为L 910＝P 9-P 10，右小腿部向量为L 1011＝P 10-P 11，左大腿部向量为L 1213＝P 12-P 13，左小腿部向量为L 1314＝P 13-P 14。
如图6所示，本申请一实施例中，所述获取所述测试者姿态向量的弯曲角度的步骤可以包括以下步骤：
步骤S601，基于所述左肩部向量与右肩部向量计算待测肩部夹角，基于所述待测肩部夹角与标准肩部夹角的角度关系获取肩部弯曲角度；
步骤S602，基于所述腰部向量和/或颈部向量计算待测脊柱夹角，基于所述待测脊柱夹角与标准脊柱夹角的角度关系获取脊柱弯曲角度；
步骤S603，基于所述左大腿部向量、左小腿部向量、右大腿部向量与右小腿部向量计算待测腿部夹角，基于所述待测腿部夹角与标准腿部夹角的角度关系获取腿部弯曲角度；及
步骤S604，基于所述肩部弯曲角度、脊柱弯曲角度、腿部弯曲角度获取所述测试者姿态向量的弯曲角度。
常见的不良体态包括驼背、高低肩、脊柱侧弯、O型腿、X型腿等。参见图7，高低肩是指两边的肩膀不在同一水平线上，可以根据肩部评估模型获取肩部弯曲角度，所述肩部评估模型如下：
Figure PCTCN2019102793-appb-000003
其中，L 12为测试者的右肩部向量，L 15为测试者的左肩部向量，B 12为标准模特的右肩部向量，B 15为标准模特的左肩部向量。
若S j的值超过预设阈值(例如，5度)，则表示与标准体态相比，测试者的肩部有较大的倾斜，评估为高低肩。
参见图8，驼背是指人的脊柱向后拱起，可以根据驼背评估模型获取脊柱弯曲角度，所述驼背评估模型如下：
Figure PCTCN2019102793-appb-000004
其中，
Figure PCTCN2019102793-appb-000005
为测试者的颈部向量
Figure PCTCN2019102793-appb-000006
中Y轴、Z轴方向的向量，
Figure PCTCN2019102793-appb-000007
为测试者的腰部向量
Figure PCTCN2019102793-appb-000008
中Y轴、Z轴方向的向量，
Figure PCTCN2019102793-appb-000009
为标准模特的颈部向量
Figure PCTCN2019102793-appb-000010
中Y轴、Z轴方向的向量，
Figure PCTCN2019102793-appb-000011
为标准模 特的腰部向量
Figure PCTCN2019102793-appb-000012
中Y轴、Z轴方向的向量。
若S b的值超过预设阈值(例如，5度)，则表示与标准体态相比，测试者的背部向后有较大拱起，评估为驼背。
参见图9，脊柱侧弯是指脊柱向侧方偏移引起的脊柱弯曲，可以根据脊柱侧弯评估模型获取脊柱弯曲角度，所述脊柱侧弯评估模型如下：
Figure PCTCN2019102793-appb-000013
其中，
Figure PCTCN2019102793-appb-000014
为测试者的腰部向量
Figure PCTCN2019102793-appb-000015
中X轴、Y轴方向的向量，
Figure PCTCN2019102793-appb-000016
为标准模特的腰部向量
Figure PCTCN2019102793-appb-000017
中X轴、Y轴方向的向量。
若S z的值超过预设阈值(例如，8度)，则表示与标准体态相比，测试者脊柱有较大侧弯，评估为脊柱侧弯。
参见图10，图中展示了O型腿和X型腿，显然，可以根据两腿间的腿部夹角获取腿部弯曲角度。本申请一实施例中，根据腿部评估模型计算腿部夹角，根据腿部评估模型计算测试者的待测腿部夹角如下：
Figure PCTCN2019102793-appb-000018
其中，
Figure PCTCN2019102793-appb-000019
为测试者的右大腿部向量
Figure PCTCN2019102793-appb-000020
中X轴、Y轴方向的向量，
Figure PCTCN2019102793-appb-000021
为测试者的右小腿部向量
Figure PCTCN2019102793-appb-000022
中X轴、Y轴方向的向量，
Figure PCTCN2019102793-appb-000023
为测试者的左大腿部向量
Figure PCTCN2019102793-appb-000024
中X轴、Y轴方向的向量，
Figure PCTCN2019102793-appb-000025
为测试者的左小腿部向量
Figure PCTCN2019102793-appb-000026
中X轴、Y轴方向的向量。
根据腿部评估模型计算标准模特的标准腿部夹角如下：
Figure PCTCN2019102793-appb-000027
其中，
Figure PCTCN2019102793-appb-000028
为标准模特的右大腿部向量
Figure PCTCN2019102793-appb-000029
中X轴、Y轴方向的向量，
Figure PCTCN2019102793-appb-000030
为标准模特的右小腿部向量
Figure PCTCN2019102793-appb-000031
中X轴、Y轴方向的向量，
Figure PCTCN2019102793-appb-000032
为标准模特的左大腿部向量
Figure PCTCN2019102793-appb-000033
中X轴、Y轴方向的向量，
Figure PCTCN2019102793-appb-000034
为标准模特的左小腿部向量
Figure PCTCN2019102793-appb-000035
中X轴、Y轴方向的向量。
若S t-S′ t＞α(α为预设阈值，例如，8度)，则表示与标准体态相比，测试者腿部向外侧弯曲，评估为O型腿；若S t-S′ t<β(β为预设阈值，例如，10度)，则表示与标准体态相比，测试者腿部向内侧弯曲，评估为X型腿。
基于所述肩部弯曲角度、脊柱弯曲角度、腿部弯曲角度获取所述测试者姿态向量的弯曲角度。当然，还可以根据测试者的性别、年龄等选取对应的标准体态作为评估标准。
如图11所示，本申请一实施例中，所述根据所述测试者姿态向量进行体态评估的步骤可以包括以下步骤：
步骤S111，在所述待测图像不符合测试要求时，发送提示信息提示所述测试者调整姿态。
如前所述，待测图像可以是测试者通过手机等具有拍摄功能的设备，在语音提示下做出相应动作后拍摄的。因此，在拍摄了待测图像后，可以对所述待测图像进行分析，以判断所述测试者是否立正站立，所述待测图像是否为全身图像等，若所述待测图像不符合测试要求，还可以通过发送提示信息提示所述测试者调整姿态，以提高测试的准确性。
进一步地，若所述测试者存在不良体态，例如脊柱侧弯，可能与所述测试者长期翘二郎腿有关，则可以间歇性的发送提示信息，提醒用户不要翘二郎腿、久坐后起身活动一下等，以提醒测试者改善不良体态。
本申请所提出的体态评估方法，能够从测试者的待测图像中提取骨骼关键点，根据所述骨骼关键点计算测试者姿态向量，获取所述测试者姿态向量的弯曲角度，进而根据所述测试者姿态向量进行体态评估，能够提高体态评估的全面性及准确性，且操作简单。
本申请进一步提供一种电子装置。参阅图12，是本申请一示例性实施例示出的电子装置20的程序模块示意图。
所述电子装置20包括：
获取模块201，适于获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；
提取模块202，适于从所述待测图像中提取骨骼关键点；
计算模块203，适于根据所述骨骼关键点计算测试者姿态向量；及
评估模块204，适于获取所述测试者姿态向量的弯曲角度。
进一步地，所述提取模块202包括：预估单元，适于将所述待测图像输入神经网络，基于人体姿态估计算法预估所述待测图像的关键点热点图；及第一计算单元，适于计算所述关键点热点图中各位置的热点的高斯值，选取高斯值中的峰值所在的热点作为该位置的骨骼关键点。
进一步地，所述骨骼关键点包括：头部、颈部、躯干中心、左肩、左肘、左腕、左臀、左膝、左踝、右肩、右肘、右腕、右臀、右膝、右踝。
进一步地，所述计算模块203包括：连接单元，适于基于人体姿态连接所述骨骼关键点；获取单元，适于获取所述骨骼关键点的坐标；及第二计算单元，适于基于所述坐标计算相连接的两个骨骼关键点的肢体向量，根据所述肢体向量生成测试者姿态向量。
进一步地，所述第二计算单元，还适于基于颈部与左肩的坐标计算左肩部向量，基于颈部与右肩的坐标计算右肩部向量；基于颈部与躯干中心的坐标计算腰部向量；基于颈部与头部的坐标计算颈部向量；基于左臀与左膝的坐标计算左大腿部向量，基于右臀与右膝的坐标计算右大腿部向量；及基于左膝与左踝的坐标计算左小腿部向量，基于右膝与右踝的坐标计算右小腿部向量。
进一步地，所述评估模块204包括：第三计算单元，适于基于所述左肩部向量与右肩部向量计算待测肩部夹角，基于所述待测肩部夹角与标准肩部夹角的角度关系获取肩部弯曲角度；基于所述腰部向量和/或颈部向量计算待测脊柱夹角，基于所述待测脊柱夹角与标准脊柱夹角的角度关系获取脊柱弯曲角度；基于所述左大腿部向量、左小腿部向量、右大腿部向量与右小腿部 向量计算待测腿部夹角，基于所述待测腿部夹角与标准腿部夹角的角度关系获取腿部弯曲角度；及评估单元，适于基于所述肩部弯曲角度、脊柱弯曲角度、腿部弯曲角度获取所述测试者姿态向量的弯曲角度。
进一步地，所述电子装置20还包括：提示模块，适于在所述待测图像不符合测试要求时，发送提示信息提示所述测试者调整姿态。
本申请所提出的电子装置，能够从测试者的待测图像中提取骨骼关键点，根据所述骨骼关键点计算测试者姿态向量，获取所述测试者姿态向量的弯曲角度，进而根据所述弯曲角度进行体态评估，能够提高体态评估的全面性及准确性，且操作简单。
为实现上述目的，如图13所示，本申请还提供一种计算机设备20，包括存储器21、处理器22以及存储在存储器21上并可在所述处理器22上运行的计算机可读指令，所述处理器22执行所述计算机可读指令时实现上述方法的步骤。可以将所述计算机可读指令存储于内存24中。
为实现上述目的，本申请还提供非易失性计算机可读存储介质，其上存储有计算机可读指令，所述计算机可读指令被处理器执行时实现上述方法的步骤。
本申请还提供一种计算机设备，如可以执行程序的智能手机、平板电脑、笔记本电脑、台式计算机、机架式服务器、刀片式服务器、塔式服务器或机柜式服务器(包括独立的服务器，或者多个服务器所组成的服务器集群)等。本实施例的计算机设备至少包括但不限于：可通过系统总线相互通信连接的存储器、处理器等。
本实施例还提供一种非易失性计算机可读存储介质，如闪存、硬盘、多媒体卡、卡型存储器(例如，SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘、服务器、App应用商城等等，其上存储有计算机可读指令，指令被处理器执行 时实现相应功能。本实施例的非易失性计算机可读存储介质用于存储电子装置20，被处理器22执行时实现本申请的体态评估方法。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。

Claims (20)

  1. 一种体态评估方法，所述方法包括步骤：
    获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；
    从所述待测图像中提取骨骼关键点；
    根据所述骨骼关键点计算测试者姿态向量；及
    获取所述测试者姿态向量的弯曲角度。
  2. 如权利要求1所述的体态评估方法，所述从所述待测图像中提取骨骼关键点的步骤还包括：
    将所述待测图像输入神经网络，基于人体姿态估计算法预估所述待测图像的关键点热点图；及计算所述关键点热点图中各位置的热点的高斯值，选取高斯值中的峰值所在的热点作为该位置的骨骼关键点。
  3. 如权利要求2所述的体态评估方法，所述骨骼关键点包括：头部、颈部、躯干中心、左肩、左肘、左腕、左臀、左膝、左踝、右肩、右肘、右腕、右臀、右膝、右踝。
  4. 如权利要求3所述的体态评估方法，所述根据所述骨骼关键点计算测试者姿态向量的步骤还包括：
    基于人体姿态连接所述骨骼关键点；
    获取所述骨骼关键点的坐标；及基于所述坐标计算相连接的两个骨骼关键点的肢体向量，根据所述肢体向量生成测试者姿态向量。
  5. 如权利要求4所述的体态评估方法，所述基于所述坐标计算相连接的两个骨骼关键点的肢体向量的步骤还包括：
    基于颈部与左肩的坐标计算左肩部向量，基于颈部与右肩的坐标计算右肩部向量；
    基于颈部与躯干中心的坐标计算腰部向量；
    基于颈部与头部的坐标计算颈部向量；
    基于左臀与左膝的坐标计算左大腿部向量，基于右臀与右膝的坐标计算右大腿部向量；及基于左膝与左踝的坐标计算左小腿部向量，基于右膝与右踝的坐标计算右小腿部向量。
  6. 如权利要求5所述的体态评估方法，所述获取所述测试者姿态向量的弯曲角度的步骤还包括：
    基于所述左肩部向量与右肩部向量计算待测肩部夹角，基于所述待测肩部夹角与标准肩部夹角的角度关系获取肩部弯曲角度；
    基于所述腰部向量和/或颈部向量计算待测脊柱夹角，基于所述待测脊柱夹角与标准脊柱夹角的角度关系获取脊柱弯曲角度；
    基于所述左大腿部向量、左小腿部向量、右大腿部向量与右小腿部向量计算待测腿部夹角，基于所述待测腿部夹角与标准腿部夹角的角度关系获取腿部弯曲角度；及基于所述肩部弯曲角度、脊柱弯曲角度、腿部弯曲角度获取所述测试者姿态向量的弯曲角度。
  7. 如权利要求1所述的体态评估方法，所述获取所述测试者姿态向量的弯曲角度的步骤之前，还包括：
    在所述待测图像不符合测试要求时，发送提示信息提示所述测试者调整姿态。
  8. 一种电子装置，其包括：
    获取模块，适于获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；
    提取模块，适于从所述待测图像中提取骨骼关键点；
    计算模块，适于根据所述骨骼关键点计算测试者姿态向量；及
    评估模块，适于获取所述测试者姿态向量的弯曲角度。
  9. 一种计算机设备，包括存储器、处理器以及存储在存储器上并可在处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现所述体态评估方法的步骤包括：
    获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；
    从所述待测图像中提取骨骼关键点；
    根据所述骨骼关键点计算测试者姿态向量；及
    获取所述测试者姿态向量的弯曲角度。
  10. 如权利要求9所述的设备，所述从所述待测图像中提取骨骼关键点的步骤还包括：
    将所述待测图像输入神经网络，基于人体姿态估计算法预估所述待测图像的关键点热点图；及计算所述关键点热点图中各位置的热点的高斯值，选取高斯值中的峰值所在的热点作为该位置的骨骼关键点。
  11. 如权利要求10所述的设备，所述骨骼关键点包括：头部、颈部、躯干中心、左肩、左肘、左腕、左臀、左膝、左踝、右肩、右肘、右腕、右臀、右膝、右踝。
  12. 如权利要求11所述的设备，所述根据所述骨骼关键点计算测试者姿态向量的步骤还包括：
    基于人体姿态连接所述骨骼关键点；
    获取所述骨骼关键点的坐标；及基于所述坐标计算相连接的两个骨骼关键点的肢体向量，根据所述肢体向量生成测试者姿态向量。
  13. 如权利要求12所述的设备，所述基于所述坐标计算相连接的两个骨骼关键点的肢体向量的步骤还包括：
    基于颈部与左肩的坐标计算左肩部向量，基于颈部与右肩的坐标计算右肩部向量；
    基于颈部与躯干中心的坐标计算腰部向量；
    基于颈部与头部的坐标计算颈部向量；
    基于左臀与左膝的坐标计算左大腿部向量，基于右臀与右膝的坐标计算右大腿部向量；及基于左膝与左踝的坐标计算左小腿部向量，基于右膝与右 踝的坐标计算右小腿部向量。
  14. 如权利要求13所述的设备，所述获取所述测试者姿态向量的弯曲角度的步骤还包括：
    基于所述左肩部向量与右肩部向量计算待测肩部夹角，基于所述待测肩部夹角与标准肩部夹角的角度关系获取肩部弯曲角度；
    基于所述腰部向量和/或颈部向量计算待测脊柱夹角，基于所述待测脊柱夹角与标准脊柱夹角的角度关系获取脊柱弯曲角度；
    基于所述左大腿部向量、左小腿部向量、右大腿部向量与右小腿部向量计算待测腿部夹角，基于所述待测腿部夹角与标准腿部夹角的角度关系获取腿部弯曲角度；及基于所述肩部弯曲角度、脊柱弯曲角度、腿部弯曲角度获取所述测试者姿态向量的弯曲角度。
  15. 一种非易失性计算机可读存储介质，其上存储有计算机可读指令，所述计算机可读指令被处理器执行时实现所述体态评估方法的步骤包括：
    获取待测图像，所述待测图像包括测试者立正站立的正面全身图像及侧面全身图像；
    从所述待测图像中提取骨骼关键点；
    根据所述骨骼关键点计算测试者姿态向量；及
    获取所述测试者姿态向量的弯曲角度。
  16. 如权利要求15所述的存储介质，所述从所述待测图像中提取骨骼关键点的步骤还包括：
    将所述待测图像输入神经网络，基于人体姿态估计算法预估所述待测图像的关键点热点图；及计算所述关键点热点图中各位置的热点的高斯值，选取高斯值中的峰值所在的热点作为该位置的骨骼关键点。
  17. 如权利要求16所述的存储介质，所述骨骼关键点包括：头部、颈部、躯干中心、左肩、左肘、左腕、左臀、左膝、左踝、右肩、右肘、右腕、右臀、右膝、右踝。
  18. 如权利要求17所述的存储介质，所述根据所述骨骼关键点计算测试者姿态向量的步骤还包括：
    基于人体姿态连接所述骨骼关键点；
    获取所述骨骼关键点的坐标；及基于所述坐标计算相连接的两个骨骼关键点的肢体向量，根据所述肢体向量生成测试者姿态向量。
  19. 如权利要求18所述的存储介质，所述基于所述坐标计算相连接的两个骨骼关键点的肢体向量的步骤还包括：
    基于颈部与左肩的坐标计算左肩部向量，基于颈部与右肩的坐标计算右肩部向量；
    基于颈部与躯干中心的坐标计算腰部向量；
    基于颈部与头部的坐标计算颈部向量；
    基于左臀与左膝的坐标计算左大腿部向量，基于右臀与右膝的坐标计算右大腿部向量；及基于左膝与左踝的坐标计算左小腿部向量，基于右膝与右踝的坐标计算右小腿部向量。
  20. 如权利要求19所述的存储介质，所述获取所述测试者姿态向量的弯曲角度的步骤还包括：
    基于所述左肩部向量与右肩部向量计算待测肩部夹角，基于所述待测肩部夹角与标准肩部夹角的角度关系获取肩部弯曲角度；
    基于所述腰部向量和/或颈部向量计算待测脊柱夹角，基于所述待测脊柱夹角与标准脊柱夹角的角度关系获取脊柱弯曲角度；
    基于所述左大腿部向量、左小腿部向量、右大腿部向量与右小腿部向量计算待测腿部夹角，基于所述待测腿部夹角与标准腿部夹角的角度关系获取腿部弯曲角度；及基于所述肩部弯曲角度、脊柱弯曲角度、腿部弯曲角度获取所述测试者姿态向量的弯曲角度。
PCT/CN2019/102793 2019-07-04 2019-08-27 体态评估方法、电子装置、计算机设备及存储介质 WO2021000401A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910599978.7A CN110495889B (zh) 2019-07-04 2019-07-04 体态评估方法、电子装置、计算机设备及存储介质
CN201910599978.7 2019-07-04

Publications (1)

Publication Number Publication Date
WO2021000401A1 true WO2021000401A1 (zh) 2021-01-07

Family

ID=68585849

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/102793 WO2021000401A1 (zh) 2019-07-04 2019-08-27 体态评估方法、电子装置、计算机设备及存储介质

Country Status (2)

Country Link
CN (1) CN110495889B (zh)
WO (1) WO2021000401A1 (zh)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112932470A (zh) * 2021-01-27 2021-06-11 上海萱闱医疗科技有限公司 俯卧撑训练的评估方法及装置、设备、存储介质
CN113128337A (zh) * 2021-03-10 2021-07-16 恒鸿达科技有限公司 一种斜体引身测试计数方法、装置、设备和介质
CN113221815A (zh) * 2021-05-25 2021-08-06 北京无垠创新科技有限责任公司 一种基于骨骼关键点自动检测技术的步态鉴定方法
CN113486771A (zh) * 2021-06-30 2021-10-08 福州大学 基于关键点检测的视频动作整齐度评估方法及系统
CN113850150A (zh) * 2021-09-02 2021-12-28 苏州爱可尔智能科技有限公司 一种基于深度学习3d姿态分析的运动评分方法及装置
CN115861381A (zh) * 2023-02-28 2023-03-28 中国民用航空飞行学院 球体协同运动中鱼跃救球动作标准性的检测方法及系统
CN115890671A (zh) * 2022-11-17 2023-04-04 山东大学 一种基于smpl参数的多几何体人体碰撞模型生成方法及系统
CN116052273A (zh) * 2023-01-06 2023-05-02 北京体提科技有限公司 一种基于体态鱼骨线的动作比对方法及装置
WO2024124927A1 (zh) * 2022-12-12 2024-06-20 海信视像科技股份有限公司 一种虚拟数字人的生成方法、装置和电子设备
CN118230427A (zh) * 2024-05-24 2024-06-21 浪潮软件科技有限公司 一种适用线上体育活动的仰卧起坐计数方法及系统

Families Citing this family (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111274990B (zh) * 2020-02-11 2024-01-12 广东同天科技产业发展有限公司 对脊柱形态进行分类的计算机设备和存储介质
CN111311660A (zh) * 2020-02-21 2020-06-19 联想(北京)有限公司 一种镜像显示人体姿态的方法、智能镜及存储介质
WO2021179230A1 (zh) * 2020-03-12 2021-09-16 南方科技大学 一种脊柱侧弯检测模型的生成方法和计算机设备
CN112102947B (zh) * 2020-04-13 2024-02-13 国家体育总局体育科学研究所 用于身体姿态评估的装置和方法
CN111358471B (zh) * 2020-04-15 2023-04-28 青岛一小步科技有限公司 一种身体姿态检测装置及检测方法
CN111666844A (zh) * 2020-05-26 2020-09-15 电子科技大学 一种羽毛球运动员运动姿态评估方法
CN111914790B (zh) * 2020-08-14 2022-08-02 电子科技大学 基于双摄像头的不同场景下实时人体转动角度识别方法
CN111814772B (zh) * 2020-09-07 2020-12-29 科大讯飞(苏州)科技有限公司 人体姿态评估方法、装置、电子设备和存储介质
CN112070031A (zh) * 2020-09-09 2020-12-11 中金育能教育科技集团有限公司 体态检测方法、装置及设备
CN112107318B (zh) * 2020-09-24 2024-02-27 自达康(北京)科技有限公司 一种身体活动能力评估系统
CN114627546A (zh) * 2020-11-26 2022-06-14 广州源动智慧体育科技有限公司 跑步姿态的识别方法、装置和计算机设备
CN112842779A (zh) * 2021-01-15 2021-05-28 上海市第十人民医院 脊柱退行性疾病的自然位影像检查智能辅助系统
CN112801979B (zh) * 2021-01-28 2023-09-05 广东工业大学 一种基于不确定性推理的脊柱侧弯检测方法及系统
CN113128336A (zh) * 2021-03-10 2021-07-16 恒鸿达科技有限公司 一种引体向上测试计数方法、装置、设备和介质
CN113139962B (zh) * 2021-05-26 2021-11-30 北京欧应信息技术有限公司 用于脊柱侧凸概率评估的系统和方法
CN113842622B (zh) * 2021-09-23 2023-05-30 京东方科技集团股份有限公司 一种运动教学方法、装置、系统、电子设备及存储介质
CN113925497B (zh) * 2021-10-22 2023-09-15 吉林大学 一种基于双目视觉测量系统的汽车乘员乘坐姿态提取方法
CN114224322B (zh) * 2021-10-25 2023-06-02 上海工程技术大学 一种基于人体骨骼关键点的脊柱侧弯评估方法
CN113951874B (zh) * 2021-10-25 2023-12-26 中国科学院长春光学精密机械与物理研究所 一种脊柱侧弯风险评估系统
CN116503892A (zh) * 2022-01-21 2023-07-28 华为技术有限公司 基于图像的姿态判断方法及电子设备
CN114519866A (zh) * 2022-01-29 2022-05-20 阿里巴巴（中国）有限公司 人体测量数据的获取方法、处理方法及设备
CN115316941A (zh) * 2022-06-15 2022-11-11 芯海科技（深圳）股份有限公司 基于背包的人体姿态检测方法、装置、背包及存储介质
CN114952854B (zh) * 2022-06-16 2024-07-09 山东大学 基于人机协作的人体碰撞物对接方法及系统
CN114821006B (zh) 2022-06-23 2022-09-20 盾钰（上海）互联网科技有限公司 交互式间接推理的孪生体态检测方法及系统
CN115462761B (zh) * 2022-09-29 2024-02-23 中国医学科学院北京协和医院 一种身体平衡监测方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109330602A (zh) * 2018-11-01 2019-02-15 中山市人民医院 一种女性形体智能评估检测装置及方法、存储介质
CN109674477A (zh) * 2018-08-06 2019-04-26 深圳创感科技有限公司 基于深度学习的计算机视觉体态分析方法
US20190142328A1 (en) * 2017-10-17 2019-05-16 Beneufit, Inc. Measuring dynamic body movement
CN109948590A (zh) * 2019-04-01 2019-06-28 启霖世纪（北京）教育科技有限公司 姿态问题检测方法及装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107886069A (zh) * 2017-11-10 2018-04-06 东北大学 一种多目标人体2d姿态实时检测系统及检测方法
CN108830150B (zh) * 2018-05-07 2019-05-28 山东师范大学 一种基于三维人体姿态估计方法及装置
CN109344803A (zh) * 2018-10-29 2019-02-15 广东工业大学 一种姿态特征提取方法及相关装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190142328A1 (en) * 2017-10-17 2019-05-16 Beneufit, Inc. Measuring dynamic body movement
CN109674477A (zh) * 2018-08-06 2019-04-26 深圳创感科技有限公司 基于深度学习的计算机视觉体态分析方法
CN109330602A (zh) * 2018-11-01 2019-02-15 中山市人民医院 一种女性形体智能评估检测装置及方法、存储介质
CN109948590A (zh) * 2019-04-01 2019-06-28 启霖世纪（北京）教育科技有限公司 姿态问题检测方法及装置

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112932470B (zh) * 2021-01-27 2023-12-29 上海萱闱医疗科技有限公司 俯卧撑训练的评估方法及装置、设备、存储介质
CN112932470A (zh) * 2021-01-27 2021-06-11 上海萱闱医疗科技有限公司 俯卧撑训练的评估方法及装置、设备、存储介质
CN113128337A (zh) * 2021-03-10 2021-07-16 恒鸿达科技有限公司 一种斜体引身测试计数方法、装置、设备和介质
CN113221815A (zh) * 2021-05-25 2021-08-06 北京无垠创新科技有限责任公司 一种基于骨骼关键点自动检测技术的步态鉴定方法
CN113486771A (zh) * 2021-06-30 2021-10-08 福州大学 基于关键点检测的视频动作整齐度评估方法及系统
CN113486771B (zh) * 2021-06-30 2023-07-07 福州大学 基于关键点检测的视频动作整齐度评估方法及系统
CN113850150A (zh) * 2021-09-02 2021-12-28 苏州爱可尔智能科技有限公司 一种基于深度学习3d姿态分析的运动评分方法及装置
CN115890671A (zh) * 2022-11-17 2023-04-04 山东大学 一种基于smpl参数的多几何体人体碰撞模型生成方法及系统
WO2024124927A1 (zh) * 2022-12-12 2024-06-20 海信视像科技股份有限公司 一种虚拟数字人的生成方法、装置和电子设备
CN116052273A (zh) * 2023-01-06 2023-05-02 北京体提科技有限公司 一种基于体态鱼骨线的动作比对方法及装置
CN116052273B (zh) * 2023-01-06 2024-03-08 北京体提科技有限公司 一种基于体态鱼骨线的动作比对方法及装置
CN115861381A (zh) * 2023-02-28 2023-03-28 中国民用航空飞行学院 球体协同运动中鱼跃救球动作标准性的检测方法及系统
CN118230427A (zh) * 2024-05-24 2024-06-21 浪潮软件科技有限公司 一种适用线上体育活动的仰卧起坐计数方法及系统

Also Published As

Publication number Publication date
CN110495889B (zh) 2022-05-27
CN110495889A (zh) 2019-11-26

Similar Documents

Publication Publication Date Title
WO2021000401A1 (zh) 体态评估方法、电子装置、计算机设备及存储介质
US20190066327A1 (en) Non-transitory computer-readable recording medium for storing skeleton estimation program, skeleton estimation device, and skeleton estimation method
CN109637625B (zh) 自学习式健身计划生成系统
AU2017202152B2 (en) Method and system for determining postural balance of a person
Brown et al. The X-Factor: An evaluation of common methods used to analyse major inter-segment kinematics during the golf swing
CN112464918A (zh) 健身动作纠正方法、装置、计算机设备和存储介质
US20230000409A1 (en) Fatigue estimation system, estimation device, and fatigue estimation method
WO2020211255A1 (zh) 人体形体及体质数据采集方法、装置及存储介质
US20160203358A1 (en) Systems and methods for using curvatures to analyze facial and body features
CN113139962B (zh) 用于脊柱侧凸概率评估的系统和方法
CN110020628B (zh) 基于面部检测的坐姿检测方法、系统、设备及存储介质
WO2023143215A1 (zh) 人体测量数据的获取方法、处理方法及设备
CN112070031A (zh) 体态检测方法、装置及设备
CN115035542A (zh) 用于将测试帧序列与参考帧序列进行匹配的系统和方法
JP2017047105A (ja) エクササイズメニュー提供システムおよびエクササイズメニュー提供方法
KR101118372B1 (ko) 가상 마커 방식을 이용한 비만/체형을 평가하는 시스템 및 방법
Wheat et al. The measurement of upper body alignment during the golf drive
CN116994339B (zh) 一种基于图像处理的用于坐位体前屈测试的方法和系统
KR100930594B1 (ko) 안면 영상 촬영장치 및 그의 안면 특징점 검출 방법
CN117503099A (zh) 人体成分分析仪、测量方法、装置、设备及存储介质
CN112818800A (zh) 一种基于人体骨骼点深度图像的体育动作评估方法及系统
CN116978127A (zh) 基于姿态估计的健身姿态矫正方法、装置、设备及介质
US20220078339A1 (en) Method for obtaining picture for measuring body size and body size measurement method, server, and program using same
WO2023135625A1 (ja) 評価装置、評価方法、および記録媒体
CN116228727A (zh) 一种基于图像的人体背部表皮脊柱线的检测方法和装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19936038

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19936038

Country of ref document: EP

Kind code of ref document: A1