Control console sedentariness reminding method based on machine vision

Abstract

The invention discloses a console sedentariness reminding method based on machine vision, and relates to the technical field of sedentariness reminding, the console sedentariness reminding method fuses a biometric authentication main channel and a visual auxiliary feature slave channel in a breakthrough manner, and can still stably bind the identity of an operator in complex scenes such as face shielding, sideways, multi-person overlapping and the like; the dynamic weight of the work color gamut compensates for the illumination change, the Frechet distance of the equipment contour accurately matches the geometric features, and the timing interruption problem caused by identity confusion is eliminated from the source; micro-actions are quantized through displacement standard deviation, sitting posture types are distinguished through spinal curvature angles, non-working duration is filtered through a service window state, and a pseudo sedentariness state is removed through cooperation of three mechanisms; only when an operator keeps a work concentration posture, timing is accumulated, and ineffective reminding is prevented from interfering with an emergency scheduling task.

Description

Technical Field

[0001] The present invention relates to the technical field of sedentary reminder technology, and in particular to a console sedentary reminder method based on machine vision. Background Art

[0002] Modern high-end dispatch centers widely adopt multi-person collaborative workstation designs to support multi-position shifts and emergency collaboration. In such scenarios, operators are often in a dynamic working state: frequently turning sideways to discuss work orders, briefly overlapping positions during shift changes, or multiple people monitoring large-screen data simultaneously. The control console needs to continuously track individual sedentary behavior to trigger health interventions.

[0003] The current mainstream solution relies on a fusion algorithm of facial recognition and skeleton tracking. When an operator leans sideways to communicate with a neighbor, if their face deviates from the camera by more than 60 degrees, the system mistakenly identifies them as having left their seat and resets the timer, resulting in no reminder of the actual sedentary timeout. Furthermore, during shift handovers, the new operator's body may block the incumbent's position, such as by reaching out to pass a document. This can cause skeleton detection points to overlap and trigger identity binding confusion. Therefore, a machine vision-based console sedentary reminder solution is urgently needed to address this issue. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] The present invention provides a console sedentary reminder method based on machine vision to solve the problem that in multi-person collaborative scenarios, existing face recognition solutions lead to identity misjudgment due to occlusion and perspective offset, causing the sedentary reminder to fail.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: The embodiment of the present invention provides a console sedentary reminder method based on machine vision, which includes: Step S1, establishing the primary identity index: generating a unique operator identity by logging into the terminal's biometric module or electronic nameplate system through the console; Step S2, capturing video data: using the console's built-in camera to collect the workstation area video stream at a set frequency; Step S3: Extract visual auxiliary features: Based on the video frames associated with the identity primary index, identify the operator's unique static feature set, including the color range of the work clothes, the outline of the device worn, and the geometric features of the hairstyle; Step S4, decoupling multi-target poses: for overlapping human targets in the video stream, the skeleton nodes are segmented using partial convolution masks and assigned to corresponding identity primary indexes based on visual auxiliary features; Step S5, calculating the sitting time: when the displacement change of the skeleton node associated with the identity primary index is continuously lower than the activity threshold, the effective sitting time is accumulated; Step S6, triggering health intervention: when the effective sedentary time exceeds a preset threshold, a prompt instruction is sent to the corresponding console terminal.

[0007] As a preferred solution of the console sedentary reminder method based on machine vision described in the present invention, the establishment of the identity primary index in step S1 includes: The biometric module is a fingerprint or iris sensor, which generates an identity token containing the work number, shift, and position code after successful authentication; The electronic nameplate system transmits employee ID and device binding code via near-field communication protocol, which is mapped to the identity master index.

[0008] As a preferred solution of the console sedentary reminder method based on machine vision described in the present invention, the extraction of visual auxiliary features in step S3 includes: Color gamut range extraction for work clothes: Take the first frame image after identity binding, convert the operator's upper body ROI area into HSV space, and cluster to extract the average value and variance range of the main hues; Wearable device contour locking: Generates a region of interest based on the head and shoulders key points, and fits the polygon vertex sequence of the headphones / glasses after Canny edge detection; Hairstyle geometric feature construction: Principal component analysis is performed on the circumscribed ellipse of the head, and the ratio of the length of the first principal component to the minor axis is used as the feature quantity.

[0009] As a preferred solution of the console sedentary reminder method based on machine vision described in the present invention, the decoupling of multiple target postures in step S4 includes: When skeleton detection outputs overlapping objects, a binary mask of each object is generated and expanded; Perform matching calculation on nodes within the mask area; Associate the local node set with the highest matching degree to the corresponding identity primary index.

[0010] As a preferred solution of the console sedentary reminder method based on machine vision described in the present invention, wherein: in step S4, during the decoupled multi-target posture matching calculation process, when the skeleton detection outputs overlapping human targets, a binary mask of each target is generated and a morphological dilation process is performed to separate the adhesion area. For the candidate skeleton node set within each mask area, a multi-feature fusion matching calculation is performed, including: Color gamut distribution of work clothes based on identity master index storage , HSV hue histogram, extract the hue histogram of the same dimension for the ROI area where the candidate node set is located , use Bhattacharyya coefficient to calculate similarity: , in, The histogram bin index ranges from 1 to 16. Indicates the reference color gamut in The normalized frequency value of each bin, Indicates the candidate region in The normalized frequency value of each bin; Introducing lighting robustness weights : , in, is the variance of the reference color gamut distribution, the constant term is the maximum allowable variance threshold, an empirical calibration value; Wearable device polygon vertex sequence stored based on identity primary index , for the candidate region edge point sequence , sample 50 points and calculate the Fréchet distance : , in, and is a continuous parameterized function of the two curves, is a parameter variable in the interval [0,1], represents the Euclidean distance operator; Normalize and convert to similarity score: , Among them, the constant term is the distance attenuation coefficient, in mm⁻¹, Define the silhouette significance weight as: , in, is the total length of the reference profile, in mm, and the constant is the maximum length threshold, in mm; Calculate the similarity between the candidate region and the reference hairstyle geometric features: , in, is the ratio of the length and short axis of the reference hairstyle, is the ratio of the major and minor axes of the hairstyle in the candidate area, and the constant is the Gaussian kernel standard deviation; View adaptation weight: ,in, is the angle between the head normal and the camera optical axis, in radians; Fusion of three types of feature scores: , in, is the comprehensive matching degree; after traversing all candidate node sets: if the highest matching degree satisfies , associate the node set to the corresponding identity primary index, otherwise it is marked as an unidentified target, where the constant term is the matching threshold.

[0011] As a preferred solution of the console sedentary reminder method based on machine vision of the present invention, the calculation of the sitting time in step S5 includes: The three-dimensional coordinates of the sacral nodes were obtained in a 10-second cycle; Calculate the standard deviation of the node displacement in the XY plane over 6 consecutive cycles; When the standard deviation is lower than the set threshold and the spinal curvature angle is in the range of 20°-40°, it is judged as effective sedentary.

[0012] As a preferred embodiment of the machine vision-based console sedentary reminder method of the present invention, in step S5, during the calculation of the sitting time, the effective sedentary time is determined as follows: when the skeleton node associated with the identity primary index persists, the following determination process is executed, and only the sedentary behavior in the working state is accumulated as the effective sedentary time: The three-dimensional coordinates of the sacral node are collected at a cycle of 10 seconds , calculate the displacement fluctuations in the intrinsic XY plane for 6 consecutive cycles, 60 seconds: , in, is the displacement vector of adjacent periods, is its module length in mm, is the average displacement, is the index period, 5 and 6 are the consecutive period numbers; Judgment conditions: When When the body is considered to have no significant displacement, is the empirical threshold; Based on the spatial coordinates of the sacrum S, cervical spine C, and head T, construct a vector and calculate the real-time bending angle: , , , in, is the vector from sacrum to cervical vertebra, unit is mm, is the vector from the cervical spine to the top of the head, in mm, represents the vector dot product, is the vector modulus, is the spinal curvature angle, in degrees, is the conversion factor from radians to degrees; When the following conditions are met at the same time, the cumulative effective sedentary time is: , in, The xy in the calculation indicates that only the XY plane component is calculated. The business software status is verified in real time through the operating system API. If any condition fails, the timing is paused until it is met again for 60 consecutive seconds. The business software list is stored in the console's local database using an MD5 hash table.

[0013] As a preferred solution of the console sedentary reminder method based on machine vision described in the present invention, it further includes operating state adaptation: Get the process name of the console focus window through the operating system API; When the process name belongs to the preset business software list, the alarm threshold duration is automatically extended by 1.5 times; When the focus window is the browser and the domain name is not a working platform, the original alarm threshold is restored.

[0014] As a preferred solution of the console sedentary reminder method based on machine vision described in the present invention, when the health intervention instruction in step S6 is executed: Sends a height adjustment signal to the console lift mechanism, driving the desktop to rise to standing mode; Send instructions to the environmental control module via the Modbus protocol to increase the illumination of the workstation area to 500 lux.

[0015] As a preferred solution of the console sedentary reminder method based on machine vision described in the present invention, wherein: the visual auxiliary feature is dynamically updated: Every 30 minutes of effective working time, the color range of the work clothes is re-extracted; When the wearable device is detected to be removed or replaced, the outline polygon data is updated in real time.

[0016] The beneficial effects of the present invention are as follows: the present invention has a breakthrough fusion of the main biometric authentication channel and the secondary visual auxiliary feature channel, and can still stably bind the operator's identity in complex scenarios such as facial occlusion, sideways, and overlapping of multiple people; the dynamic weight of the work uniform color gamut compensates for lighting changes, and the Fréchet distance of the equipment contour accurately matches the geometric features, eliminating the timing interruption problem caused by identity confusion from the root; micro-movements are quantified by displacement standard deviation, sitting posture types are distinguished by spinal bending angle, and non-working time is filtered by business window status. The three mechanisms work together to eliminate pseudo-sedentary states; timing is accumulated only when the operator maintains a focused work posture to avoid invalid reminders interfering with emergency dispatch tasks.

[0017] The sedentary trigger instruction of the present invention is deeply integrated into the smart ecology of the enterprise. The lifting mechanism automatically switches to standing mode to relieve lumbar pressure. The environmental module adjusts the illumination of the workstation to reduce visual fatigue, dynamically updates visual features, and refreshes the work uniform color gamut every 30 minutes to ensure all-weather recognition stability, forming a monitoring-intervention-optimization closed loop; reuses the existing hardware of the console, such as cameras / biometrics / electronic nameplates, and realizes functional transition through algorithm upgrades rather than hardware modifications. The edge computing architecture ensures local processing of video stream data to meet the dual requirements of privacy and real-time performance of the dispatch center; at the same time, the introduction of a work status perception mechanism fits the high-intensity duty characteristics of the dispatcher; dynamic feature updates meet the needs of shift changes; the essence of the present invention is to seamlessly embed health management into the workflow rather than adding additional operational burden. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 This is a flow chart of the machine vision-based console sedentary reminder method in Example 1. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0023] Example 1, reference Figure 1 This embodiment provides a console sedentary reminder method based on machine vision, comprising the following steps: Step S1, establishing the primary identity index: generating a unique operator identity by logging into the terminal's biometric module or electronic nameplate system through the console; The establishment of the identity primary index in step S1 includes: The biometric module is a fingerprint or iris sensor, which generates an identity token containing the work number, shift, and position code after successful authentication; The electronic nameplate system transmits the employee ID and device binding code via the near-field communication protocol, which is mapped to the identity master index; Step S2, capturing video data: using the console's built-in camera to collect the workstation area video stream at a set frequency; Step S3: Extract visual auxiliary features: Based on the video frames associated with the identity primary index, identify the operator's unique static feature set, including the color range of the work clothes, the outline of the device worn, and the geometric features of the hairstyle; Extracting visual auxiliary features in step S3 includes: Color gamut range extraction for work clothes: Take the first frame image after identity binding, convert the operator's upper body ROI area into HSV space, and cluster to extract the average value and variance range of the main hues; Wearable device contour locking: Generates a region of interest based on the head and shoulders key points, and fits the polygon vertex sequence of the headphones / glasses after Canny edge detection; Hairstyle geometric feature construction: Principal component analysis is performed on the circumscribed ellipse of the head, and the ratio of the length of the first principal component to the minor axis is used as the feature quantity; Visual Assist features are dynamically updated: Every 30 minutes of effective working time, the color range of the work clothes is re-extracted; When the wearable device is detected to be removed or replaced, the contour polygon data is updated in real time; Step S4, decoupling multi-target poses: for overlapping human targets in the video stream, the skeleton nodes are segmented using partial convolution masks and assigned to corresponding identity primary indexes based on visual auxiliary features; Decoupling the multi-target postures in step S4 includes: When skeleton detection outputs overlapping objects, a binary mask of each object is generated and expanded; Perform matching calculation on nodes within the mask area; Associate the local node set with the highest matching degree to the corresponding identity primary index; In step S4, during the matching degree calculation process of decoupling multiple target postures, when the skeleton detection outputs overlapping human targets, a binary mask of each target is generated and a morphological dilation process is performed to separate the adhesion areas. For each candidate skeleton node set within the mask area, a multi-feature fusion matching degree calculation is performed, including: Color gamut distribution of work clothes based on identity master index storage , HSV hue histogram, extract the hue histogram of the same dimension for the ROI area where the candidate node set is located , use Bhattacharyya coefficient to calculate similarity: , in, The histogram bin index ranges from 1 to 16. Indicates the reference color gamut in The normalized frequency value of each bin, Indicates the candidate region in The normalized frequency value of each bin; Introducing lighting robustness weights : , in, is the variance of the reference color gamut distribution, the constant term is the maximum allowable variance threshold, an empirical calibration value; Wearable device polygon vertex sequence stored based on identity primary index , for the candidate region edge point sequence , sample 50 points and calculate the Fréchet distance : , in, and is a continuous parameterized function of the two curves, is a parameter variable in the interval [0,1], represents the Euclidean distance operator; Normalize and convert to similarity score: , Among them, the constant term is the distance attenuation coefficient, in mm⁻¹, Define the silhouette significance weight as: , in, is the total length of the reference profile, in mm, and the constant is the maximum length threshold, in mm; Calculate the similarity between the candidate region and the reference hairstyle geometric features: , in, is the ratio of the length and short axis of the reference hairstyle, is the ratio of the major and minor axes of the hairstyle in the candidate area, and the constant is the Gaussian kernel standard deviation; View adaptation weight: ,in, is the angle between the head normal and the camera optical axis, in radians; Fusion of three types of feature scores: , in, is the comprehensive matching degree; after traversing all candidate node sets: if the highest matching degree satisfies , associate the node set to the corresponding identity primary index, otherwise it is marked as an unidentified target, where the constant term is the matching threshold, and the parameters involved in the formula are set according to the actual scenario; Specifically, a multi-level feature fusion mechanism significantly improves identity recognition accuracy in occluded scenes. Color matching uses dynamic weights to compensate for lighting changes. Contour matching accurately captures geometric differences through curve parameterization. Hairstyle features are combined with viewpoint weights to reduce the misjudgment rate. A comprehensive judgment threshold effectively filters low-quality matches and avoids timing deviations caused by identity misassociation. Step S5, calculating the sitting time: when the displacement change of the skeleton node associated with the identity primary index is continuously lower than the activity threshold, the effective sitting time is accumulated; The calculation of the sitting time in step S5 includes: The three-dimensional coordinates of the sacral nodes were obtained in a 10-second cycle; Calculate the standard deviation of the node displacement in the XY plane over 6 consecutive cycles; When the standard deviation is lower than the set threshold and the spinal curvature angle is in the range of 20°-40°, it is determined to be effective sitting; In step S5, when calculating the sitting time, the effective sedentary time is determined as follows: when the skeleton node associated with the identity primary index continues to exist, the following determination process is executed, and only the sitting behavior in the working state is accumulated as the effective sedentary time: The three-dimensional coordinates of the sacral node are collected at a cycle of 10 seconds , calculate the displacement fluctuations in the intrinsic XY plane for 6 consecutive cycles, 60 seconds: , in, is the displacement vector of adjacent periods, is its module length in mm, is the average displacement, is the index period, 5 and 6 are the consecutive period numbers; Judgment conditions: When When the body is considered to have no significant displacement, is the empirical threshold; Based on the spatial coordinates of the sacrum S, cervical spine C, and head T, construct a vector and calculate the real-time bending angle: , , , in, is the vector from sacrum to cervical vertebra, unit is mm, is the vector from the cervical spine to the top of the head, in mm, represents the vector dot product, is the vector modulus, is the spinal curvature angle, in degrees, is the conversion factor from radians to degrees; When the following conditions are met at the same time, the cumulative effective sedentary time is: , in, The xy in the code indicates that only the XY plane components are calculated. The business software status is verified in real time through the operating system API. If any condition fails, the timing is paused until it is met again for 60 consecutive seconds. The business software list is stored in the console's local database using an MD5 hash table. Specifically, effective sedentary time is determined through composite physiological characteristics, significantly improving the accuracy of state recognition. The displacement standard deviation quantifies micro-movements of the body to avoid misjudging brief posture adjustments as leaving the seat. The spinal curvature angle accurately distinguishes between a sitting posture of 20°-40° and a forward working posture of <20° or a backward resting posture of >40°. Combined with the business software status, non-working periods are filtered to ensure that the timing results truly reflect occupational health risks.

[0024] Step S6, triggering health intervention: when the effective sedentary time exceeds a preset threshold, a prompt instruction is sent to the corresponding console terminal; When the health intervention instruction in step S6 is executed: Sends a height adjustment signal to the console lift mechanism, driving the desktop to rise to standing mode; Send instructions to the environmental control module via the Modbus protocol to increase the illumination of the workstation area to 500 lux; The console sedentary reminder method further includes an operation state adaptation step: Get the process name of the console focus window through the operating system API; When the process name belongs to the preset business software list, the alarm threshold duration is automatically extended by 1.5 times; When the focus window is the browser and the domain name is not a working platform, the original alarm threshold is restored.

[0025] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A console sedentary reminder method based on machine vision, characterized in that: include, Step S1, establishing the primary identity index: generating a unique operator identity by logging into the terminal's biometric module or electronic nameplate system through the console; Step S2, capturing video data: using the console's built-in camera to collect the workstation area video stream at a set frequency; Step S3: Extract visual auxiliary features: Based on the video frames associated with the identity primary index, identify the operator's unique static feature set, including the color range of the work clothes, the outline of the device worn, and the geometric features of the hairstyle; Step S4, decoupling multi-target poses: for overlapping human targets in the video stream, the skeleton nodes are segmented using partial convolution masks and assigned to corresponding identity primary indexes based on visual auxiliary features; Step S5, calculating the sitting time: when the displacement change of the skeleton node associated with the identity primary index is continuously lower than the activity threshold, the effective sitting time is accumulated; Step S6, triggering health intervention: when the effective sedentary time exceeds a preset threshold, a prompt instruction is sent to the corresponding console terminal.

2. The method for reminding a user of sedentary behavior based on machine vision according to claim 1, wherein: The establishment of the identity primary index in step S1 includes: The biometric module is a fingerprint or iris sensor, which generates an identity token containing the work number, shift, and position code after successful authentication; The electronic nameplate system transmits employee ID and device binding code via near-field communication protocol, which is mapped to the identity master index.

3. The method for reminding a user of sedentary behavior based on machine vision according to claim 1, wherein: Extracting visual auxiliary features in step S3 includes: Color gamut range extraction for work clothes: Take the first frame image after identity binding, convert the operator's upper body ROI area into HSV space, and cluster to extract the average value and variance range of the main hues; Wearable device contour locking: Generates a region of interest based on the head and shoulders key points, and fits the polygon vertex sequence of the headphones / glasses after Canny edge detection; Hairstyle geometric feature construction: Principal component analysis is performed on the circumscribed ellipse of the head, and the ratio of the length of the first principal component to the minor axis is used as the feature quantity.

4. The method for reminding a user of sedentary behavior based on machine vision according to claim 1, wherein: Decoupling the multi-target postures in step S4 includes: When skeleton detection outputs overlapping objects, a binary mask of each object is generated and expanded; Perform matching calculation on nodes within the mask area; Associate the local node set with the highest matching degree to the corresponding identity primary index.

5. The method for reminding a user of sedentary behavior based on machine vision according to claim 4, wherein: In step S4, during the decoupled multi-target pose matching calculation process, when the skeleton detection outputs overlapping human targets, a binary mask of each target is generated and a morphological dilation process is performed to separate the adhesion areas. For each candidate skeleton node set within the mask area, a multi-feature fusion matching calculation is performed, including: Color gamut distribution of work clothes based on identity master index storage , HSV hue histogram, extract the hue histogram of the same dimension for the ROI area where the candidate node set is located , use Bhattacharyya coefficient to calculate similarity: ， in, The histogram bin index ranges from 1 to 16. Indicates the reference color gamut in The normalized frequency value of each bin, Indicates the candidate region in The normalized frequency value of each bin; Introducing lighting robustness weights : ， in, is the variance of the reference color gamut distribution, the constant term is the maximum allowable variance threshold, an empirical calibration value; Wearable device polygon vertex sequence stored based on identity primary index , for the candidate region edge point sequence , sample 50 points and calculate the Fréchet distance : ， in, and is a continuous parameterized function of the two curves, is a parameter variable in the interval [0,1], represents the Euclidean distance operator; Normalize and convert to similarity score: ， Among them, the constant term is the distance attenuation coefficient, in mm⁻¹, Define the silhouette significance weight as: ， in, is the total length of the reference profile, in mm, and the constant is the maximum length threshold, in mm; Calculate the similarity between the candidate region and the reference hairstyle geometric features: ， in, is the ratio of the length and short axis of the reference hairstyle, is the ratio of the major and minor axes of the hairstyle in the candidate area, and the constant is the Gaussian kernel standard deviation; View adaptation weight: ,in, is the angle between the head normal and the camera optical axis, in radians; Fusion of three types of feature scores: ， in, is the comprehensive matching degree; after traversing all candidate node sets: if the highest matching degree satisfies , associate the node set to the corresponding identity primary index, otherwise it is marked as an unidentified target, where the constant term is the matching threshold.

6. The method for reminding a user of sedentary behavior based on machine vision according to claim 1, wherein: The calculation of the presence time in step S5 includes: The three-dimensional coordinates of the sacral nodes were obtained in a 10-second cycle; Calculate the standard deviation of the node displacement in the XY plane over 6 consecutive cycles; When the standard deviation is lower than the set threshold and the spinal curvature angle is in the range of 20°-40°, it is judged as effective sedentary.

7. The method for reminding a user of sedentary behavior based on machine vision according to claim 6, wherein: In step S5, during the calculation of the sitting time, the effective sedentary time is determined as follows: when the skeleton node associated with the identity primary index persists, the following determination process is executed, and only the sitting behavior in the working state is accumulated as the effective sedentary time: The three-dimensional coordinates of the sacral node are collected at a cycle of 10 seconds , calculate the displacement fluctuations in the intrinsic XY plane for 6 consecutive cycles, 60 seconds: ， in, is the displacement vector of adjacent periods, is its module length in mm, is the average displacement, is the index period, 5 and 6 are the consecutive period numbers; Judgment conditions: When When the body is considered to have no significant displacement, is the empirical threshold; Based on the spatial coordinates of the sacrum S, cervical spine C, and head T, construct a vector and calculate the real-time bending angle: ， ， ， in, is the vector from sacrum to cervical vertebra, unit is mm, is the vector from the cervical spine to the top of the head, in mm, represents the vector dot product, is the vector modulus, is the spinal curvature angle, in degrees, is the conversion factor from radians to degrees; When the following conditions are met at the same time, the cumulative effective sedentary time is: ， in, The xy in the calculation indicates that only the XY plane component is calculated. The business software status is verified in real time through the operating system API. If any condition fails, the timing is paused until it is met again for 60 consecutive seconds. The business software list is stored in the console's local database using an MD5 hash table.

8. The method for reminding a user of sedentary behavior based on machine vision according to claim 1, wherein: Further includes operation state adaptation: Get the process name of the console focus window through the operating system API; When the process name belongs to the preset business software list, the alarm threshold duration is automatically extended by 1.5 times; When the focus window is the browser and the domain name is not a working platform, the original alarm threshold is restored.

9. The method for reminding a user of prolonged sitting on a console based on machine vision according to claim 1, wherein: When the health intervention instruction in step S6 is executed: Sends a height adjustment signal to the console lift mechanism, driving the desktop to rise to standing mode; Send instructions to the environmental control module via the Modbus protocol to increase the illumination of the workstation area to 500 lux.

10. The method for reminding a user of sedentary behavior based on machine vision according to claim 3, wherein: Visual Assist features are dynamically updated: Every 30 minutes of effective working time, the color range of the work clothes is re-extracted; When the wearable device is detected to be removed or replaced, the outline polygon data is updated in real time.

Images