Sitting posture detection device, control method thereof and sitting posture detection method

Abstract

The invention discloses a sitting posture detection device, a control method thereof and a sitting posture detection method. The sitting posture detection device comprises a base; one end of the fixed supporting bracket is fixedly connected with one end of the base; one end of the height adjusting bracket is rotationally connected with the other end of the base through a first adjusting device; the camera device comprises a shell support, a camera and a second adjusting device, one end of the shell support is rotationally connected with the other end of the height adjusting support through a third adjusting device, the camera and the second adjusting device are both arranged in the shell support, and the second adjusting device is arranged on the back face of the camera and rotationally connected with the camera. The camera is used for detecting a target image; the base, the fixed supporting bracket, the height adjusting bracket and the camera device form a rectangular structure; and the control device is used for adjusting the rotation angles of the first adjusting device, the second adjusting device and the third adjusting device according to the target image and adjusting the shooting parameters of the camera. According to the invention, the accuracy of sitting posture detection can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field related to sitting posture detection, and in particular to a sitting posture detection device and a control method thereof, and a sitting posture detection method. Background Art

[0002] Sitting habits are "muscle memory" formed by maintaining a long-term body posture. A healthy sitting posture can help people improve work efficiency, reduce spinal pressure, and promote bone health; while a bad sitting posture is one of the important causes of spinal diseases such as back pain, scoliosis, and lumbar disc herniation. Therefore, developing a good sitting habit is beneficial to physical health and can also reduce the chances of people suffering from chronic musculoskeletal diseases.

[0003] The traditional sitting posture correction method is to forcibly correct people's incorrect sitting posture through the physical stretching effect of bandages, but it may cause irreversible damage to the body and the posture correction effect is weak. The existing intelligent sitting posture detection equipment can only meet the needs of different desktop heights by adjusting the front and back rotation angles of the detection body on the base. However, various conditions such as the width of the desktop and the dynamic position of the user may cause the camera to collect incomplete portrait data, thereby greatly reducing the accuracy of sitting posture detection. In addition, the accuracy of the existing bad posture detection results is not high. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention provides a sitting posture detection device, which can improve the accuracy of sitting posture detection.

[0005] The present invention also provides a control method for a sitting posture detection device, a sitting posture detection method, a control device for executing the control method for the sitting posture detection device and the sitting posture detection method, and a computer-readable storage medium.

[0006] A sitting posture detection device according to a first aspect of the present invention includes: Base; A fixed support bracket, one end of which is fixedly connected to one end of the base; A height-adjustable bracket, one end of which is rotatably connected to the other end of the base via a first adjusting device; A camera device, comprising a shell bracket, a camera head and a second adjustment device, wherein one end of the shell bracket is rotatably connected to the other end of the height adjustment bracket via a third adjustment device, the camera head and the second adjustment device are both arranged inside the shell bracket, the second adjustment device is arranged on the back of the camera head and rotatably connected to the camera head, and the camera head is used to detect a target image; the base, the fixed support bracket, the height adjustment bracket and the camera device form a rectangular structure; A control device is arranged inside the base, and the control device is electrically connected to the first adjustment device, the second adjustment device, the third adjustment device and the camera respectively. The control device is used to adjust the rotation angles of the first adjustment device, the second adjustment device and the third adjustment device according to the target image, and is used to adjust the camera parameters of the camera.

[0007] The sitting posture detection device according to the embodiment of the present invention has at least the following beneficial effects: By controlling the rotation angle of the first adjustment device, the rotation of the height adjustment bracket can be adjusted, thereby adjusting the height of the camera device; by controlling the rotation angle of the second adjustment device, the lateral position of the camera can be adjusted; by controlling the rotation angle of the third adjustment device, the angle of the camera can be adjusted, thereby realizing dynamic adjustment of the height, angle and lateral position of the camera to adaptively track the target object, so that the camera can always collect complete portrait data, thereby improving the accuracy of sitting posture detection.

[0008] A control method for a sitting posture detection device according to a second aspect of the present invention is applied to the sitting posture detection device described in the first aspect of the present invention, and the method includes: Get the target image; Determine a target key point set of a target object according to the target image; Determining the integrity status of the target key point set; If the completeness state represents an incomplete state, calculating a current three-dimensional bounding box formed by the target key point set, and predicting a target three-dimensional bounding box containing a complete portrait area; Determine a displacement vector according to the current three-dimensional bounding box and the target three-dimensional bounding box; The rotation angles of the first adjusting device, the second adjusting device and the third adjusting device, and the camera parameters of the camera are adjusted based on the displacement vector.

[0009] The control method of the sitting posture detection device according to the embodiment of the present invention has at least the following beneficial effects: By acquiring a target image and determining a target key point set of a target object according to the target image, all human key points of the target object in the captured image can be obtained. By determining the completeness state of the target key point set, it can be determined whether all human key points of the target object appear in the captured image. If it is an incomplete state, a current three-dimensional bounding box formed by the target key point set is calculated, and a target three-dimensional bounding box containing a complete portrait area is predicted. A displacement vector is determined according to the current three-dimensional bounding box and the target three-dimensional bounding box to determine the amount by which the camera needs to move. Then, by controlling the rotation angles of the first adjusting device, the second adjusting device and the third adjusting device, as well as the camera parameters of the camera, the height, angle and lateral position of the camera are dynamically adjusted to adaptively track the target object, so that the camera can always collect complete portrait data, thereby improving the accuracy of sitting posture detection.

[0010] According to some embodiments of the present invention, determining a target key point set of a target object according to the target image includes: Extracting the depth value of each pixel in the target image; Construct a 3D point cloud model based on the depth value of each pixel; The target key point set of the target object is obtained based on the three-dimensional point cloud model calibration.

[0011] According to some embodiments of the present invention, predicting a target three-dimensional bounding box containing a complete portrait area includes: Obtaining current point cloud data of all key points in the target key point set; Inputting the current point cloud data of all key points in the target key point set into a pre-trained NeRF model, and predicting the target point cloud data of all key points in the complete key point set containing the complete portrait area; The target three-dimensional bounding box is constructed according to the target point cloud data of all key points of the complete key point set.

[0012] According to some embodiments of the present invention, determining the integrity state of the target key point set includes: Obtaining depth values ​​corresponding to all key points in the target key point set; If there is a depth value corresponding to at least one key point and the difference between the depth values ​​corresponding to other key points is greater than a preset normal distribution distance, the completeness state is determined to be the incomplete state.

[0013] According to some embodiments of the present invention, determining the integrity state of the target key point set includes: Determine the number of key points of the target key point set; If the number of key points is less than a preset complete key point threshold, the integrity state is determined to be the incomplete state.

[0014] A sitting posture detection method according to a third aspect of the present invention is applied to the sitting posture detection device according to the first aspect of the present invention, and the method includes: Get the target image; Determine the left eye key point coordinates, right eye key point coordinates, left ear key point coordinates, right ear key point coordinates, left shoulder key point coordinates and right shoulder key point coordinates of the target object according to the target image; Determine a left eye-ear vector, a right eye-ear vector, a left shoulder-ear vector, a right shoulder-ear vector and a shoulder height difference according to the left eye key point coordinates, the right eye key point coordinates, the left ear key point coordinates, the right ear key point coordinates, the left shoulder key point coordinates and the right shoulder key point coordinates; Calculating a first angle between the left eye-ear vector and the left shoulder-ear vector, and a second angle between the right eye-ear vector and the right shoulder-ear vector; The sitting posture state of the target object is determined according to the shoulder height difference, the first angle, and the second angle.

[0015] The sitting posture detection method according to the embodiment of the present invention has at least the following beneficial effects: The embodiment of the present invention comprehensively determines the sitting state of the target object through the angle between the eye-ear vector and the shoulder-ear vector, as well as the shoulder height difference, and can more accurately identify the head-down behavior and its accompanying abnormal posture, providing the user with more effective sitting posture correction reminders and data feedback.

[0016] According to some embodiments of the present invention, determining the sitting state of the target object according to the shoulder height difference, the first angle and the second angle includes: If the first angle and the second angle are both smaller than a preset head-down threshold, and the shoulder height difference is smaller than or equal to a preset abnormal height difference, it is determined that the sitting state of the target object is a head-down state; Alternatively, if the first angle or the second angle is smaller than a preset head-down threshold, and the shoulder height difference is smaller than or equal to a preset abnormal height difference, it is determined that the sitting state of the target object is a head-down state.

[0017] According to some embodiments of the present invention, determining the sitting state of the target object according to the shoulder height difference, the first angle and the second angle includes: If the first angle or the second angle is smaller than a preset head-down threshold, and the shoulder height difference is greater than a preset abnormal height difference, it is determined that the sitting posture of the target object is a head-tilted state.

[0018] According to some embodiments of the present invention, determining the sitting state of the target object according to the shoulder height difference, the first angle and the second angle includes: If the first angle and the second angle are both within a preset normal angle range, and the shoulder height difference is less than or equal to a preset abnormal height difference, it is determined that the sitting state of the target object is normal, wherein the minimum value of the normal angle range is a preset head-down threshold.

[0019] According to the fourth aspect of the present invention, the control device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the control method of the sitting posture detection device as described in the second aspect of the present invention and the sitting posture detection method as described in the third aspect of the present invention when executing the computer program. Since the control device adopts all the technical solutions of the control method of the sitting posture detection device and the sitting posture detection method of the above-mentioned embodiments, it has at least all the beneficial effects brought by the technical solutions of the above-mentioned embodiments.

[0020] The computer-readable storage medium according to the fifth aspect of the present invention stores computer-executable instructions, and the computer-executable instructions are used to execute the control method of the sitting posture detection device as described in the second aspect of the present invention and the sitting posture detection method as described in the third aspect of the present invention. Since the computer-readable storage medium adopts all the technical solutions of the control method of the sitting posture detection device and the sitting posture detection method of the above-mentioned embodiments, it at least has all the beneficial effects brought by the technical solutions of the above-mentioned embodiments.

[0021] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which: Figure 1 is a structural diagram of a sitting posture detection device according to an embodiment of the present invention; Figure 2 is a flow chart of a control method of a sitting posture detection device according to an embodiment of the present invention; Figure 3 4 is a flow chart of a sitting posture detection method according to an embodiment of the present invention.

[0023] Reference numerals: Base 100; A fixed support bracket 200; Height adjustment bracket 300; Shell bracket 410 and camera 420 . DETAILED DESCRIPTION

[0024] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0025] In the description of the present invention, if there is a description of first, second, etc., it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0026] In the description of the present invention, it should be understood that descriptions involving orientation, such as orientation or positional relationship indicated as up, down, etc., are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0027] In the description of the present invention, it should be noted that, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific content of the technical solution.

[0028] The following will be combined Figure 1 A clear and complete description is given of the sitting posture detection device according to the embodiment of the present invention. Obviously, the embodiment described below is only a part of the embodiments of the present invention, but not all of the embodiments.

[0029] refer to Figure 1 , Figure 1 It is a structural diagram of a sitting posture detection device according to an embodiment of the present invention.

[0030] The sitting posture detection device according to the first embodiment of the present invention includes a base 100, a fixed support bracket 200, a height adjustment bracket 300, a camera device and a control device.

[0031] Base 100; A fixed support bracket 200, one end of which is fixedly connected to one end of the base 100; A height adjustment bracket 300, one end of which is rotatably connected to the other end of the base 100 via a first adjustment device; The camera device comprises a housing support 410, a camera head 420 and a second adjusting device. One end of the housing support 410 is rotatably connected to the other end of the height adjusting support 300 through a third adjusting device. The camera head 420 and the second adjusting device are both arranged inside the housing support 410. The second adjusting device is arranged on the back of the camera head 420 and is rotatably connected to the camera head 420. The camera head 420 is used to detect a target image. The base 100, the fixed support support 200, the height adjusting support 300 and the camera device form a rectangular structure. The control device is arranged inside the base 100, and the control device is electrically connected to the first adjustment device, the second adjustment device, the third adjustment device and the camera 420 respectively. The control device is used to adjust the rotation angles of the first adjustment device, the second adjustment device and the third adjustment device according to the target image, and is used to adjust the camera parameters of the camera 420.

[0032] It can be understood that the base 100, the fixed support bracket 200, the height adjustment bracket 300 and the camera device form a rectangular structure, which makes the sitting posture detection device of the embodiment of the present invention small in size and easy to carry and place.

[0033] The first adjusting device, the second adjusting device and the third adjusting device may adopt stepping motors, or other devices capable of controlling the rotation angle, which are not limited here.

[0034] The control device and the first adjustment device, the second adjustment device, the third adjustment device and the camera 420 may be connected by wire or by wireless.

[0035] The specific shapes and sizes of the base 100, the fixed support bracket 200, the height adjustment bracket 300 and the camera device can be selected according to actual conditions. Figure 1 The structures shown are not to be construed as limiting the invention.

[0036] The sitting posture detection device of the embodiment of the present invention can be placed directly on the desktop, and can also be equipped with an adjustable support or clamp to adapt to the width and thickness of the desktop. It can be compatible with various widths and narrow tables, baby safety monitoring, and more desktop scenarios.

[0037] According to the sitting posture detection device of the embodiment of the present invention, by controlling the rotation angle of the first adjusting device, the rotation of the height adjustment bracket 300 can be adjusted, thereby adjusting the height of the camera device; by controlling the rotation angle of the second adjusting device, the lateral position of the camera 420 can be adjusted; by controlling the rotation angle of the third adjusting device, the angle of the camera 420 can be adjusted, thereby realizing dynamic adjustment of the height, angle and lateral position of the camera 420 to adaptively track the target object, so that the camera 420 can always collect complete portrait data, thereby improving the accuracy of sitting posture detection.

[0038] The following will be combined Figure 1 and Figure 2 A clear and complete description is given of the control method of the sitting posture detection device according to the embodiment of the present invention. Obviously, the embodiment described below is only a part of the embodiments of the present invention, not all of the embodiments.

[0039] refer to Figure 1 and Figure 2 , Figure 1 is a structural diagram of a sitting posture detection device according to an embodiment of the present invention, Figure 2 4 is a flow chart of a control method of a sitting posture detection device according to an embodiment of the present invention.

[0040] The control method of the sitting posture detection device according to the second aspect of the present invention is applied to the sitting posture detection device of the first aspect, and the method includes: Get the target image; Determine a target key point set of a target object according to a target image; Determining the integrity status of the target keypoint set; If the completeness state represents an incomplete state, calculate the current 3D bounding box formed by the target key point set, and predict the target 3D bounding box containing the complete portrait area; Determine a displacement vector according to the current three-dimensional bounding box and the target three-dimensional bounding box; The rotation angles of the first adjusting device, the second adjusting device and the third adjusting device, and the camera parameters of the camera 420 are adjusted based on the displacement vector.

[0041] The embodiment of the present invention adopts the NeRF model to model the human body. When it is detected that the target object is not completely in the shooting picture, the height, angle and lateral position of the camera 420 are automatically adjusted to adaptively track the target object, ensuring that the face is in the center of the picture, greatly improving convenience and detection accuracy.

[0042] The NeRF model is a technical method that relies on neural networks to achieve accurate reconstruction of three-dimensional scenes. Unlike traditional methods, the NeRF model only needs to predict the color and depth value of each pixel from a single or a few 2D perspectives, without using multiple 2D images or perspectives. It achieves this by learning a neural radiance field function that represents each point in the scene, which makes the model well applicable to the single-perspective sitting posture detection device of the embodiment of the present invention.

[0043] Through training, the spatial position and viewing direction (5D input) can be mapped to color and opacity (4D output), acting as a "volume", so that volume rendering can be used to render new views to achieve three-dimensional scene reconstruction. Then, the depth value is analyzed to predict and infer the displacement vector of the camera 420 that can fully shoot the portrait from the front, so that the camera 420 can automatically adjust the height, angle and lateral position, thereby tracking the face of the target object.

[0044] In some embodiments of the present invention, determining a target key point set of a target object according to a target image includes: Extract the depth value of each pixel in the target image; Construct a 3D point cloud model based on the depth value of each pixel; The target key point set of the target object is obtained based on the three-dimensional point cloud model calibration.

[0045] The depth value is the distance from each pixel captured by the camera 420 to the optical axis of the camera 420, and is used to reconstruct the position of the portrait in the three-dimensional space. The camera 420 of the embodiment of the present invention can be an RGB-D camera 420, which extracts the depth value of each pixel from the target image through the RGB-D camera 420. Assume that the depth value of each pixel in the depth map is D ( u , v ),in( u , v ) is the image plane coordinate of the pixel, and the depth value is the distance from the camera 420 to the point z .set up The core camera 420 is used to capture the target image of the target object's sitting posture. It is the main source of data collection and is used to detect the postures of key points such as the head and shoulders.

[0046] =( x,y,z )= ; in, , is the camera 420 focal length, , is the optical center coordinate, z is the depth value.

[0047] The target key point set includes but is not limited to eye key points, nose key points, mouth key points, ear key points and shoulder key points. First, a 3D point cloud model is constructed based on the depth value of each pixel, and the spatial positions of the eye key points, nose key points, mouth key points, ear key points and shoulder key points are calibrated. Then, the YOLO-Pose algorithm or other similar key point detection algorithms are used to extract the key information of the target object, and the 3D key point coordinates are reconstructed in combination with the depth map.

[0048] In some embodiments of the present invention, determining the integrity status of the target key point set includes: Get the depth values ​​corresponding to all key points in the target key point set; If there is a depth value corresponding to at least one key point whose difference from the depth values ​​corresponding to other key points is greater than a preset normal distribution distance, the integrity state is determined to be an incomplete state.

[0049] It is understandable that if the key point is obviously separated from other points in the scene, it is usually caused by noise, occlusion or beyond the range of the camera 420, and it is also necessary to determine that the target key point set is incomplete.

[0050] In some embodiments of the present invention, determining the integrity status of the target key point set includes: Determine the number of key points of the target key point set; If the number of key points is less than the preset complete key point threshold, the integrity status is determined to be incomplete.

[0051] It is understandable that if some important key points are not detected, it is determined that the target key point set is incomplete.

[0052] It should be noted that whether a key point is missing may also be determined by other means, not limited to the determination of quantity.

[0053] For the entire captured image, it is necessary to preliminarily determine whether the current image contains sufficient valid information for sitting posture analysis. If important points such as ears and shoulders are not detected, or the depth values ​​(i.e., 3D positions) of key points are abnormally distributed, it may indicate that the posture of the target object in the target image is incomplete or there is a problem with the camera 420's viewing angle.

[0054] In some embodiments of the present invention, predicting a target three-dimensional bounding box containing a complete portrait area includes: Get the current point cloud data of all key points in the target key point set; Input the current point cloud data of all key points in the target key point set into the pre-trained NeRF model, and predict the target point cloud data of all key points in the complete key point set containing the complete portrait area; Construct the target 3D bounding box based on the target point cloud data of all key points of the complete key point set.

[0055] It is understandable that, based on the currently detected target key point set, it is necessary to further confirm whether the target area of ​​the target object is within the field of view of the camera 420 to ensure that the bounding box completely covers the target area of ​​the target object, such as whether the shoulders or head exceed the boundary of the screen. If the current three-dimensional bounding box does not match the target area of ​​the target object (such as the head and shoulders) (for example, the shoulders or head partly exceed the boundary of the current three-dimensional bounding box), the camera 420 needs to be adjusted; if the number of key points in the current three-dimensional bounding box is insufficient, or the distribution of the key point depth values ​​is unreasonable (such as not covering key areas such as shoulders), the camera 420 also needs to be adjusted.

[0056] The displacement vector is determined according to the current three-dimensional bounding box and the target three-dimensional bounding box by calculating the difference vector of the three-dimensional coordinates. The specific process is common knowledge and will not be described in detail here.

[0057] Adjusting the rotation angles of the first adjustment device, the second adjustment device and the third adjustment device, and the camera parameters of the camera 420 based on the displacement vector is to convert the displacement vector into the parameters of the first adjustment device, the second adjustment device and the third adjustment device, and the parameters of the camera 420 adjusted by the camera 420.

[0058] The first adjusting device corresponds to height adjustment, so that the height adjustment bracket 300 rotates in a vertical direction relative to the base 100 .

[0059] The second adjusting device corresponds to horizontal adjustment, that is, lateral position adjustment, so that the camera 420 rotates in the horizontal direction.

[0060] The third adjusting device corresponds to depth adjustment, that is, angle adjustment, so that the housing bracket 410 is rotated up and down relative to the height adjustment bracket 300 or the focal length of the camera 420 is adjusted.

[0061] When the height adjustment bracket 300 rotates in the vertical direction relative to the base 100 , the housing bracket 410 can also be rotated up and down relative to the height adjustment bracket 300 or the focal length of the camera 420 can be adjusted to improve the adjustment efficiency.

[0062] In some embodiments, if the integrity state represents a complete state, the sitting posture detection can be performed directly. If the integrity state represents a complete state, the current three-dimensional bounding box can also be calculated. After calculating the current three-dimensional bounding box, if it is found that the current three-dimensional bounding box does not completely cover the target area of ​​the target object, the camera 420 needs to be automatically adjusted. If some key points in the current three-dimensional bounding box are missing or the key point distribution is abnormal, the integrity state of the target key point set can be determined again.

[0063] The embodiment of the present invention optimizes the position and angle of the camera 420 through real-time calculation and feedback, adjusts the position of the camera 420, and recaptures the picture after the adjustment to verify the integrity of the bounding box. If the goal is still not achieved, continue to iterate and adjust until the target area is completely covered. The embodiment of the present invention also establishes an accurate three-dimensional scene model through depth value analysis and NeRF model, analyzes the position of the portrait in the scene, derives the displacement vector of the camera 420 based on the complete portrait prediction, dynamically adjusts the position and angle of the camera 420, realizes automatic tracking and viewing angle optimization, and improves the accuracy and applicability of sitting posture detection.

[0064] According to the control method of the sitting posture detection device of the embodiment of the present invention, by acquiring the target image and determining the target key point set of the target object according to the target image, all the human body key points of the target object in the shooting picture can be obtained, and by determining the completeness state of the target key point set, it can be determined whether all the human body key points of the target object appear in the shooting picture. If it is an incomplete state, the current three-dimensional bounding box composed of the target key point set is calculated, and the target three-dimensional bounding box containing the complete portrait area is predicted. The displacement vector is determined according to the current three-dimensional bounding box and the target three-dimensional bounding box to determine the amount that the camera 420 needs to move. Then, by controlling the rotation angles of the first adjustment device, the second adjustment device and the third adjustment device, and the camera parameters of the camera 420, the height, angle and lateral position of the camera 420 are dynamically adjusted to adaptively track the target object, so that the camera 420 can always collect complete portrait data, thereby improving the accuracy of sitting posture detection.

[0065] The following will be combined Figure 1 and Figure 3 A clear and complete description is given of the sitting posture detection method according to the embodiment of the present invention. Obviously, the embodiment described below is only a part of the embodiments of the present invention, not all of the embodiments.

[0066] refer to Figure 1 and Figure 3 , Figure 1 is a structural diagram of a sitting posture detection device according to an embodiment of the present invention, Figure 3 4 is a flow chart of a sitting posture detection method according to an embodiment of the present invention.

[0067] The sitting posture detection method according to the third aspect of the present invention is applied to the sitting posture detection device according to the first aspect of the present invention, and the method includes: Get the target image; Determine the left eye key point coordinates, the right eye key point coordinates, the left ear key point coordinates, the right ear key point coordinates, the left shoulder key point coordinates and the right shoulder key point coordinates of the target object according to the target image; Determine the left eye-ear vector, the right eye-ear vector, the left shoulder-ear vector, the right shoulder-ear vector and the shoulder height difference according to the left eye key point coordinates, the right eye key point coordinates, the left ear key point coordinates, the right ear key point coordinates, the left shoulder key point coordinates and the right shoulder key point coordinates; Calculate a first angle between the left eye-ear vector and the left shoulder-ear vector, and a second angle between the right eye-ear vector and the right shoulder-ear vector; The sitting posture state of the target object is determined according to the shoulder height difference, the first angle and the second angle.

[0068] The current mainstream human posture estimation algorithms can be divided into two approaches. One is top-down, which first detects human targets, obtains bounding boxes, and then detects human key points in the area box to connect each person's posture; the other is bottom-up, which first detects all human key points in the image, and then defines whether these points belong to the same target, and obtains the boundary of the target to determine the detected target. The embodiment of the present invention adopts the YOLO-Pose algorithm, which belongs to the bottom-up method. The specific principle is the existing technology known to those skilled in the art, and will not be repeated here.

[0069] First, the YOLO-Pose algorithm is used to detect the left eye key point coordinates, right eye key point coordinates, left ear key point coordinates, right ear key point coordinates, left shoulder key point coordinates and right shoulder key point coordinates of the target object in the target image. The above coordinates are all three-dimensional coordinates and are the center points of the eyes, ears, and shoulders respectively. The model is trained with the YOLO-Pose algorithm to detect the target object in the target image. If there are multiple target objects, they are screened according to the key points, and the target objects without facial key points are eliminated. If there are multiple target objects that meet the conditions, the one with the highest confidence is selected as the detection target object, and the feasibility of each key vector is detected. When an unstandard sitting posture occurs and lasts for a period of time, a correction reminder is given.

[0070] The principles of determining the left eye-ear vector, right eye-ear vector, left shoulder-ear vector, right shoulder-ear vector and shoulder height difference according to the left eye key point coordinates, right eye key point coordinates, left ear key point coordinates, right ear key point coordinates, left shoulder key point coordinates and right shoulder key point coordinates, as well as the principle of determining the angle between two vectors are common knowledge and will not be elaborated here.

[0071] In some embodiments of the present invention, determining the sitting posture of the target object according to the shoulder height difference, the first angle, and the second angle includes: If the first angle and the second angle are both smaller than a preset head-down threshold, and the shoulder height difference is smaller than or equal to a preset abnormal height difference, it is determined that the sitting posture state of the target object is a head-down state; Alternatively, if the first angle or the second angle is smaller than a preset head-down threshold, and the shoulder height difference is smaller than or equal to a preset abnormal height difference, it is determined that the sitting state of the target object is a head-down state.

[0072] When the target object is in the head-down state, the eye-ear vector tilts downward, resulting in a decrease in the angle between the eye-ear vector and the shoulder-ear vector. When the angle is less than a preset head-down threshold, it can be determined as a head-down state. Specifically, the head-down threshold can be set to 70°, or to other values ​​according to actual conditions, which cannot be regarded as a limitation of the present invention.

[0073] In some embodiments, a unilateral determination can be made, that is, the first angle or the second angle is less than a preset head-down threshold, and the target object's sitting posture can be considered to be in a head-down state. A bilateral comprehensive determination can also be made, that is, the first angle and the second angle are both less than the preset head-down threshold, and the target object's sitting posture can be considered to be in a head-down state. At this time, the shoulder height difference is less than or equal to the preset abnormal height difference. In a normal sitting posture, the absolute value of the shoulder height difference is usually close to zero. Specifically, the abnormal height difference can be set to 10 mm, or it can be set to other values ​​according to actual conditions, which cannot be regarded as a limitation of the present invention.

[0074] In some embodiments of the present invention, determining the sitting posture of the target object according to the shoulder height difference, the first angle, and the second angle includes: If the first angle or the second angle is smaller than a preset head-down threshold, and the shoulder height difference is greater than a preset abnormal height difference, it is determined that the sitting posture state of the target object is a head-tilted state.

[0075] When the angle between the eye-ear vector and the shoulder-ear vector on only one side of the head is less than 70°, this usually indicates that the head may be tilted to one side rather than completely bowed. In order to further confirm whether the abnormal posture is related to bowing the head, the shoulder height difference can be combined for analysis. When the shoulder height difference is greater than the preset abnormal height difference, it indicates that the target object has abnormal sitting behaviors such as side tilt and one shoulder sinking. If the first angle or the second angle is less than the preset head-down threshold, and the shoulder height difference is greater than the preset abnormal height difference, it is determined that the sitting state of the target object is a head tilt state. Specifically, if the left shoulder is high, it means that the head is tilted to the left, and if the right shoulder is high, it means that the head is tilted to the right.

[0076] In some embodiments, if the first angle and the second angle are both smaller than a preset head-down threshold, and the shoulder height difference is greater than a preset abnormal height difference, it is determined that the target object's sitting posture is leaning forward with one shoulder sinking, or that the head posture is unbalanced.

[0077] In some embodiments of the present invention, determining the sitting posture of the target object according to the shoulder height difference, the first angle, and the second angle includes: If the first angle and the second angle are both within the preset normal angle range, and the shoulder height difference is less than or equal to the preset abnormal height difference, it is determined that the sitting posture of the target object is normal, wherein the minimum value of the normal angle range is the preset head-down threshold.

[0078] In a standard sitting posture, when the angle between the eye-ear vector and the shoulder-ear vector is close to a right angle (usually 80°~100°), the head is basically kept upright, and the minimum value of the normal angle range can be set to the preset head-down threshold, i.e. 70°, to ensure that there is no judgment standard when the angle is 70°~80°. The absolute value of the shoulder height difference is usually close to zero.

[0079] According to the sitting posture detection method of the embodiment of the present invention, the embodiment of the present invention comprehensively determines the sitting posture state of the target object through the angle between the eye-ear vector and the shoulder-ear vector, and the shoulder height difference, and can more accurately identify the head-down behavior and its accompanying posture abnormalities, and provide users with more effective sitting posture correction reminders and data feedback.

[0080] In some embodiments, the overall process of the embodiment of the present invention is: Receive the video stream acquired by the camera 420, and acquire an image of the video stream to obtain a target image; Detect whether there is a portrait feature in the target image. If not, continue to receive the target image for detection. If yes, analyze the vectors of each feature point in the target image. Detect whether the posture of the target object in the target image is optimal, calculate the optimal height and angle of the camera 420 under limited conditions according to the NeRF neural field algorithm, and transmit the data to the first adjustment device, the second adjustment device, the third adjustment device and the camera 420 for posture correction; According to the YOLO-Pose algorithm, the sitting posture of the target object is analyzed to see if it is normal. If it is not normal, the abnormal sitting posture counter value increases; The detection is performed every 2 seconds. If the abnormal sitting posture counter reaches the preset threshold within 5 minutes, the abnormal sitting posture reminder mechanism is triggered and the corresponding data packet is sent to the smart reminder bracelet.

[0081] In some embodiments, considering the scenario where parents cannot remind their children of bad sitting posture in real time, and in order to make children more aware of the harm of their bad sitting posture, the present invention also designs a smart reminder bracelet to cooperate with a sitting posture detection device to remind and correct the child's bad sitting posture. Specifically, it includes the following functions: (1) Posture reminder: When the sitting posture detection device detects that the child maintains a bad posture for a long time, the sitting posture detection device will send the data to the parent end and the bracelet. The smart reminder bracelet will remind the user to adjust the posture through vibration and prompt sound. The child can also view a diagram of his or her bad sitting posture on the bracelet, thereby helping him or her improve his or her sitting posture habits.

[0082] (2) Habit formation: The smart reminder bracelet helps users develop good sitting habits through regular reminders and reward mechanisms. Users can set personalized reminder frequencies and reward goals to gradually improve their sitting habits.

[0083] (3) Health analysis: In addition to monitoring sitting habits, the smart reminder bracelet can also conduct comprehensive analysis based on other health data of the user (such as number of steps, heart rate, etc.) to provide users with comprehensive health management services.

[0084] In addition, the sitting posture detection device of the embodiment of the present invention can also be equipped with a dedicated application software, which has the function of visualizing monitoring data and intelligently generating sitting posture reports, aiming to help parents analyze the changes in their children's daily sitting posture and provide corresponding adjustment suggestions. The following is an introduction to the main functions of the dedicated application software: (1) Data recording and analysis: The app can record and visualize the user's sitting posture data, including duration, frequency, and correctness of posture. This data is displayed to the user's parents to help them gain a deeper understanding of their children's sitting habits and motivate them to actively improve them. At the same time, the data is intelligently converted into a training plan to provide parents with scientific guidance.

[0085] (2) Data cloud synchronization: Users’ sitting posture data can be saved to the cloud through the cloud synchronization function, ensuring that users can access and track their progress on multiple devices. This feature enables users to record and monitor anytime and anywhere, providing them with greater convenience and flexibility.

[0086] (3) Personalized suggestions: Based on the user’s specific situation, the app provides tailored health suggestions and sitting posture improvement plans.

[0087] (4) Interactive Tutorials: Provide a series of sitting posture correction tutorials and exercise videos to help users actively improve their sitting posture in daily life.

[0088] After the sitting posture detection device of the embodiment of the present invention detects the user's bad sitting posture, it will give specific real-time feedback data, including but not limited to the type of abnormal sitting posture, abnormal duration, detection time point, abnormal number statistics, abnormal distribution period, abnormal cumulative duration, sitting posture score, posture trend chart, fatigue index, bone pressure assessment and specific improvement suggestions.

[0089] Types of abnormal sitting posture include but are not limited to slouching (head tilted forward too much), hunchback (back bends at an excessive angle), uneven shoulders (large height difference between left and right shoulders), and head tilt (head tilted to one side).

[0090] Record the duration of abnormal sitting posture (such as "bowing the head for more than 10 seconds"), the specific time when each abnormality is detected, record the frequency of various types of abnormal sitting posture by day, week and month, analyze the high-frequency time periods of abnormal sitting posture in a day, and record the cumulative time of different types of abnormalities. This can facilitate user tracing and provide long-term trend analysis of users' sitting habits.

[0091] The algorithm scores the user's overall sitting posture performance, which can be specifically reflected in the form of a score of 0-100 on the day. A chart is used to show the changing trend of the sitting posture score over a period of time, helping users to intuitively understand the improvement effect.

[0092] Based on the frequency and duration of abnormal sitting posture, it is possible to determine whether the user is in a state of fatigue. Based on long-term abnormal sitting posture (such as hunching), it is possible to indicate the pressure on the spine and evaluate the user's possible physical condition.

[0093] Improvement suggestions include: specific guidance on how to adjust sitting posture, such as: "Please straighten your back, relax your shoulders, and keep your head level." If the continuous working time is too long, remind the user to take a short break, such as: "You have been working for 1 hour, it is recommended to stand up and move around for 5 minutes."

[0094] For different types of incorrect sitting postures, correction suggestions should be combined with ergonomic principles to provide specific and easy-to-understand guidance. The following are detailed correction suggestions for common incorrect sitting postures: (1) Head bowing (head leaning forward too much). Problem manifestation: The head leans forward too much, the line of sight is too close to the desk or book, and the cervical spine is under too much pressure. Correction suggestions: ① Remind users to raise their heads and chests, retract their chins slightly, and align their ears with their shoulders. ② It is recommended to adjust the distance between the line of sight and the desk or screen to keep it between 0 and 40 cm. ③ If using a writing table, it is recommended to add a book stand or adjust the height of the table and chair so that the line of sight and the book are at a downward angle of 5° to 20°.

[0095] (2) Hunchback (back bending). Problem manifestation: The back is bent, the thoracic spine area is compressed, and the chest and waist are hunched. Correction suggestions: ① Remind the user to open the shoulders backwards, imagine the shoulder blades are clamped together, and keep the back straight. ② If the chair does not provide sufficient support, it is recommended to add a cushion or lumbar pillow behind the waist. ③ Remind the user to get up and move around every 30 minutes to avoid maintaining a bent posture for a long time.

[0096] (3) Uneven shoulders. Problem manifestation: One shoulder is higher than the other, which may be caused by leaning against the desk or uneven force on one side. Corrective suggestions: ① Remind users to adjust their sitting posture to make both shoulders level and avoid supporting the desk with one hand or arm for a long time. ② It is recommended to place the center of gravity of the body in the center of the chair to avoid leaning the body. ③ If the force is on one side due to writing habits, it is recommended to adjust the writing posture to avoid pressing down the shoulder for a long time.

[0097] (4) Head tilt (head tilted to one side). Problem manifestation: The head tilts to one side, possibly accompanied by a sinking of one shoulder. Corrective suggestions: ① Remind the user to keep the head neutral and the ears should be aligned with the shoulders. ② Guide the user to adjust the position of the screen or book to ensure that it is directly in the user's line of sight to avoid head tilt due to asymmetrical line of sight. ③ Encourage the user to gently stretch the neck muscles left and right regularly to relieve tension in the neck muscles.

[0098] (5) Crooked sitting posture. Problem manifestation: The pelvis is offset to one side, causing the body to tilt or the center of gravity to shift to one side. Correction suggestions: ① Remind the user to adjust the buttocks to sit on the chair surface, with both feet flat on the ground and the knees parallel to the hips. ② It is recommended to evenly distribute the body weight on both sides when sitting to avoid excessive force on one side. ③ If the chair is unstable or lacks support, it is recommended to replace it with a chair with better stability.

[0099] (6) Leaning forward too much. Problem manifestation: The whole body leans forward, the chest is too close to the table, and the spinal load is increased. Correction suggestions: ① Remind users to keep their backs close to the chair back when sitting, adjust the height of the table or the placement of items to reduce the need to lean forward. ② Encourage the use of adjustable chairs, adjust to the appropriate height, and let the arms rest naturally on the table.

[0100] (7) Maintaining a static posture for a long time. Problem manifestation: Maintaining a single posture for a long time increases muscle fatigue and joint pressure. Corrective suggestions: ① Remind users to get up and move or stretch every 30 minutes to relieve fatigue. ② The reminder function can be used to suggest some simple movements, such as standing stretching, rotating shoulders, etc.

[0101] Specific reminder methods: ① Vibration reminder: a light vibration reminder when the abnormal duration reaches the set threshold. ② Screen prompt: the problem and suggestions are explained in text and animation through the supporting device or application. ③ Voice prompt: play simple voice instructions, such as "please look up" and "sit up straight".

[0102] Through these targeted suggestions, users can clearly understand the causes of the problem and how to improve it, which helps to develop good sitting habits and reduce the negative impact of long-term poor sitting posture on the body.

[0103] In addition, an embodiment of the present invention further provides a control device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor and the memory may be connected via a bus or other means.

[0104] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0105] The non-transient software program and instructions required to implement the control method of the sitting posture detection device and the sitting posture detection method of the above-mentioned embodiment are stored in the memory, and when executed by the processor, the control method of the sitting posture detection device and the sitting posture detection method of the above-mentioned embodiment are executed.

[0106] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0107] In addition, an embodiment of the present invention also provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are executed by a processor or a controller, for example, by the processor of the above-mentioned embodiment, so that the above-mentioned processor can execute the control method of the sitting posture detection device and the sitting posture detection method in the above-mentioned embodiment.

[0108] It will be appreciated by those skilled in the art that all or some of the steps and systems in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0109] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A sitting posture detection device, characterized in that: include: Base; A fixed support bracket, one end of which is fixedly connected to one end of the base; A height-adjustable bracket, one end of which is rotatably connected to the other end of the base via a first adjusting device; A camera device, comprising a shell bracket, a camera head and a second adjustment device, wherein one end of the shell bracket is rotatably connected to the other end of the height adjustment bracket via a third adjustment device, the camera head and the second adjustment device are both arranged inside the shell bracket, the second adjustment device is arranged on the back of the camera head and rotatably connected to the camera head, and the camera head is used to detect a target image; the base, the fixed support bracket, the height adjustment bracket and the camera device form a rectangular structure; A control device is arranged inside the base, and the control device is electrically connected to the first adjustment device, the second adjustment device, the third adjustment device and the camera respectively. The control device is used to adjust the rotation angles of the first adjustment device, the second adjustment device and the third adjustment device according to the target image, and is used to adjust the camera parameters of the camera.

2. A control method for a sitting posture detection device, characterized in that: Applied to the sitting posture detection device as claimed in claim 1, the method comprises: Get the target image; Determine a target key point set of a target object according to the target image; Determining the integrity status of the target key point set; If the completeness state represents an incomplete state, calculating a current three-dimensional bounding box formed by the target key point set, and predicting a target three-dimensional bounding box containing a complete portrait area; Determine a displacement vector according to the current three-dimensional bounding box and the target three-dimensional bounding box; The rotation angles of the first adjusting device, the second adjusting device and the third adjusting device, and the camera parameters of the camera are adjusted based on the displacement vector.

3. The control method of the sitting posture detection device according to claim 2, characterized in that: The step of determining a target key point set of a target object according to the target image comprises: Extracting the depth value of each pixel in the target image; Construct a 3D point cloud model based on the depth value of each pixel; The target key point set of the target object is obtained based on the three-dimensional point cloud model calibration.

4. The control method of the sitting posture detection device according to claim 3, characterized in that: The predicted target 3D bounding box containing the complete portrait area includes: Obtaining current point cloud data of all key points in the target key point set; Inputting the current point cloud data of all key points in the target key point set into a pre-trained NeRF model, and predicting the target point cloud data of all key points in the complete key point set containing the complete portrait area; The target three-dimensional bounding box is constructed according to the target point cloud data of all key points of the complete key point set.

5. The control method of the sitting posture detection device according to claim 3, characterized in that: The determining the integrity status of the target key point set includes: Obtaining depth values ​​corresponding to all key points in the target key point set; If there is a depth value corresponding to at least one key point and the difference between the depth values ​​corresponding to other key points is greater than a preset normal distribution distance, the completeness state is determined to be the incomplete state.

6. The control method of the sitting posture detection device according to claim 2, characterized in that: The determining the integrity status of the target key point set includes: Determine the number of key points of the target key point set; If the number of key points is less than a preset complete key point threshold, the integrity state is determined to be the incomplete state.

7. A sitting posture detection method, characterized in that: Applied to the sitting posture detection device as claimed in claim 1, the method comprises: Get the target image; Determine the left eye key point coordinates, right eye key point coordinates, left ear key point coordinates, right ear key point coordinates, left shoulder key point coordinates and right shoulder key point coordinates of the target object according to the target image; Determine a left eye-ear vector, a right eye-ear vector, a left shoulder-ear vector, a right shoulder-ear vector and a shoulder height difference according to the left eye key point coordinates, the right eye key point coordinates, the left ear key point coordinates, the right ear key point coordinates, the left shoulder key point coordinates and the right shoulder key point coordinates; Calculating a first angle between the left eye-ear vector and the left shoulder-ear vector, and a second angle between the right eye-ear vector and the right shoulder-ear vector; The sitting posture state of the target object is determined according to the shoulder height difference, the first angle, and the second angle.

8. The sitting posture detection method according to claim 7, characterized in that: The determining the sitting state of the target object according to the shoulder height difference, the first angle and the second angle comprises: If the first angle and the second angle are both smaller than a preset head-down threshold, and the shoulder height difference is smaller than or equal to a preset abnormal height difference, it is determined that the sitting state of the target object is a head-down state; Alternatively, if the first angle or the second angle is smaller than a preset head-down threshold, and the shoulder height difference is smaller than or equal to a preset abnormal height difference, it is determined that the sitting state of the target object is a head-down state.

9. The sitting posture detection method according to claim 7, characterized in that: The determining the sitting state of the target object according to the shoulder height difference, the first angle and the second angle comprises: If the first angle or the second angle is smaller than a preset head-down threshold, and the shoulder height difference is greater than a preset abnormal height difference, it is determined that the sitting posture of the target object is a head-tilted state.

10. The sitting posture detection method according to claim 7, characterized in that: The determining the sitting state of the target object according to the shoulder height difference, the first angle and the second angle comprises: If the first angle and the second angle are both within a preset normal angle range, and the shoulder height difference is less than or equal to a preset abnormal height difference, it is determined that the sitting state of the target object is normal, wherein the minimum value of the normal angle range is a preset head-down threshold.

Images