CN117994819B - Human body posture monitoring system based on image data analysis - Google Patents
Human body posture monitoring system based on image data analysis Download PDFInfo
- Publication number
- CN117994819B CN117994819B CN202410385178.6A CN202410385178A CN117994819B CN 117994819 B CN117994819 B CN 117994819B CN 202410385178 A CN202410385178 A CN 202410385178A CN 117994819 B CN117994819 B CN 117994819B
- Authority
- CN
- China
- Prior art keywords
- human body
- data analysis
- posture
- unit
- preset
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000007405 data analysis Methods 0.000 title claims abstract description 99
- 238000012544 monitoring process Methods 0.000 title claims abstract description 30
- 238000001514 detection method Methods 0.000 claims abstract description 23
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 7
- 230000003068 static effect Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 3
- 230000008676 import Effects 0.000 claims description 2
- 238000000034 method Methods 0.000 description 13
- 230000009471 action Effects 0.000 description 12
- 230000036541 health Effects 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000006378 damage Effects 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 230000006872 improvement Effects 0.000 description 3
- 230000001788 irregular Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 208000028571 Occupational disease Diseases 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of human body posture monitoring, in particular to a human body posture monitoring system based on image data analysis, which comprises an image acquisition unit for acquiring human body image information from different directions, an image preprocessing unit for preprocessing the human body image information acquired by the image acquisition unit, a joint point detection unit for carrying out joint point identification detection on the preprocessed human body image information, a posture estimation unit for preliminarily estimating the human body posture based on the distribution condition of each joint point, a posture recognition unit for matching the estimated posture with a predefined posture model, a data analysis unit for judging whether the current human body posture is standard or not based on the motion trail of each joint point, and a prompt unit for sending a notification of correcting the posture based on the non-standard reason of the current human body posture determined by the data analysis unit.
Description
Technical Field
The invention relates to the technical field of human body posture monitoring, in particular to a human body posture monitoring system based on image data analysis.
Background
The human body posture monitoring has important significance in real life, can help people to correctly execute exercise actions in the field of body building, avoid injury, improve exercise effect, monitor the posture of staff in workplaces in work, prevent occupational diseases, improve work efficiency, and in addition, the human body posture monitoring can also be used for rehabilitation training and posture correction in the medical field, helps patients to recover health, is beneficial to improving the life quality of people and promotes health and productivity in general.
Chinese patent publication No.: CN115019399a discloses a human body posture detecting method, and provides a human body posture detecting method, which comprises the steps of obtaining a user action video through a camera and detecting skeletal key points of a human body, and further comprises the following steps: extracting each image frame in the user action video; judging the number of times of completion of leg lifting actions in the user action video according to the leg lifting action recording rule; calculating the detection angles among the key points and analyzing the posture of the human body; judging whether each leg lifting action is a high leg lifting action or not; calculating the standard degree of the high leg lifting action; after the whole movement is finished, calculating the whole completion degree b and the average standard degree c of the movement; according to the human body posture detection method, the high leg lifting movement posture in body building is detected, judged and performance evaluated through the human body posture estimation technology aiming at the problems that a coach is absent in household exercise, correct posture guidance is not available, the movement result cannot be evaluated and the like.
The existing human body posture monitoring system has the following problems: the existing human body posture monitoring system needs to carry out a great deal of training and optimization to adapt to different application scenes, and cannot adapt to complex environments and scenes well.
Disclosure of Invention
Therefore, the invention provides a human body posture monitoring system based on image data analysis to solve the problem that the human body posture cannot be completely and accurately identified in the prior art.
In order to achieve the above object, the present invention provides a human body posture monitoring system based on image data analysis, comprising:
The image acquisition unit comprises a plurality of image sensors and is used for acquiring human body image information from different directions;
The image preprocessing unit is connected with the image acquisition unit and is used for preprocessing the human body image information acquired by the image acquisition unit;
The joint point detection unit is connected with the image preprocessing unit and is used for identifying and detecting joint points of the preprocessed human body image information and marking the joint points;
The gesture estimation unit is connected with the joint point detection unit and is used for preliminarily estimating the human gesture based on the distribution situation of each joint point; the preliminary estimated body posture includes standing, sitting, lying, walking and running;
The gesture recognition unit is connected with the gesture estimation unit and is used for judging the specific gesture of the human body based on the matching result of the primarily estimated gesture and the corresponding predefined gesture model;
The data analysis unit is respectively connected with the joint point detection unit and the gesture recognition unit and is used for establishing a coordinate system based on the matched gesture model to determine the coordinates of each joint point in the coordinate system and determining whether the current human gesture is standard or not based on the motion track of each joint point in the coordinate system; the data analysis unit is also used for determining the cause of the nonstandard human body posture or sending a posture correcting instruction to the prompting unit under the condition that the current human body posture is judged to be nonstandard;
The prompting unit is connected with the data analysis unit and used for sending out a corresponding notification of correcting the posture based on the reason that the current human posture is not standard and determined by the data analysis unit;
the data analysis unit changes the acquisition angle to carry out secondary judgment under the condition that the movement track of each joint point is not in accordance with the preset movement track of the gesture model which is currently recognized by the data analysis unit, and determines the cause of the non-standard human gesture based on the joint points which are not in accordance with the preset movement track under the condition that the movement track of each joint point is not in accordance with the preset movement track of the gesture model which is currently recognized by the data analysis unit.
Further, the data analysis unit is configured to change the acquisition angle of the image acquisition unit for the human body under the condition that the motion track of each joint point is primarily determined to be not in accordance with the preset motion track of the gesture model currently identified by the data analysis unit, set the deflection track of each joint point in advance based on the variation of the acquisition angle and the initial acquisition angle, and secondarily determine whether the gesture of the human body is standard based on the number of the joint points, where the contact ratio of the counted actual deflection track and the preset deflection track is not in accordance with the preset contact ratio standard.
Further, the data analysis unit is configured to determine a cause of the nonstandard human body posture based on a variance of a deviation between the motion track of each joint point and the preset motion track under the condition that it is determined that the motion track of each joint point does not conform to the preset motion track of the gesture model currently identified by the data analysis unit, and select a corresponding processing mode, where the processing mode includes:
issuing a command for correcting the human body posture to the prompt unit under the condition that the variance accords with a preset variance standard;
Judging that the matching of the gesture recognition unit aiming at the gesture model is problematic under the condition that the preliminary judgment variance accords with a preset variance standard;
Determining that the posture has a change problem under the condition that the preliminary determination variance does not meet the preset variance standard, and re-determining the human posture based on the current state;
And judging that the positioning of the joint point detection unit for each joint point is problematic under the condition that the judgment variance does not accord with the preset variance standard.
Further, the data analysis unit is configured to set a plurality of adjustment modes for the matching mechanism of the gesture recognition unit to the human body gesture model under the condition that the variance of the deviation between the motion track of each joint point and the preset motion track is preliminarily determined to be in accordance with the preset variance standard, and the accuracy of the gesture recognition unit adjusted by different adjustment modes to match the human body gesture model is different.
Further, the data analysis unit is configured to switch the static posture estimation mode initiated by the posture estimation unit to the dynamic posture estimation mode when it is preliminarily determined that the variance of the motion track of each joint point and the deviation of the preset motion track do not meet the preset variance standard.
Further, the data analysis unit is configured to obtain the reference data from the cloud to perform the abnormal recognition of the human body under the condition that the human body posture still cannot be accurately recognized when the posture estimation unit is switched to the dynamic posture estimation mode.
Further, the data analysis unit is configured to provide a plurality of optimization modes for coordinates of each joint point in the coordinate system under the condition that it is determined that variances of deviations of motion trajectories of the joint points and preset motion trajectories do not meet preset variance standards, and the coordinates of the joint points are different after the adjustment modes are used for adjustment.
Further, the data analysis unit judges that the reason of the nonstandard human body posture of the monitored person is the stature reason under the condition that the coordinate optimization of each joint point is completed and the variance of the deviation between the motion trail of each joint point and the preset motion trail still does not meet the preset variance standard, and the data analysis unit guides the matching data retrieved at the cloud end to the data analysis unit based on the motion trail of each joint point so as to judge whether the motion trail of each joint point meets the standard again.
Further, the data analysis unit is used for recording current human body posture data under the condition that matching data does not exist in cloud searching, and the data analysis unit is also used for uploading the recorded human body posture data to a cloud updating cloud database.
Compared with the prior art, the method has the beneficial effects that the data analysis unit establishes a coordinate system based on the gesture model identified by the gesture identification unit, marks the joint points identified and detected by the joint point detection unit in the coordinate system, judges that the human body gesture is standard when the movement track of each joint point accords with the preset movement track of the gesture model identified currently, changes the acquisition angle to carry out secondary judgment when the movement track of each joint point is primarily judged to be not in accordance with the preset movement track of the gesture model identified currently, and determines the reason of the non-standard human body gesture based on the joint points which are not in accordance with the preset movement track when the movement track of each joint point is judged to be not in accordance with the preset movement track of the gesture model identified currently, thereby improving the judgment precision of whether the human body gesture monitoring system is standard for the current human body gesture, reducing the damage of the wrong gesture to the human body health and helping people to correctly execute the movement action.
Further, the data analysis unit changes the acquisition angle of the image acquisition unit under the condition that the movement track of each joint point is not in accordance with the preset movement track of the current recognized gesture model by the data analysis unit, the preset deflection track is set based on the variation of the acquisition angle and the initial acquisition angle, whether the human gesture is normal or not is secondarily judged based on the number of the joint points, the coincidence degree of the counted actual deflection track and the preset deflection track is not in accordance with the preset coincidence degree standard, the judgment precision of the human gesture monitoring system on whether the current human gesture is normal or not is further improved, the damage of the wrong gesture to the human health is reduced, and the correct execution of the movement action of people is helped.
Further, the data analysis unit determines the reason for the nonstandard human body posture based on the variance of the deviation between the motion track of each joint point and the preset motion track when the motion track of each joint point is not in accordance with the preset motion track of the currently-identified posture model, sends a command for correcting the human body posture to the prompting unit when the variance is in accordance with the preset variance standard, determines that the posture recognition unit is in question for matching the posture model when the preliminary determination variance is in accordance with the preset variance standard, determines that the posture is in question of changing when the preliminary determination variance is not in accordance with the preset variance standard, redetermines the human body posture based on the current state, determines that the positioning of each joint point by the joint point detection unit is in question when the determination variance is not in accordance with the preset variance standard, further improves the determination accuracy of whether the human body posture monitoring system is in accordance with the current human body posture, reduces the harm of the wrong posture to the human body health, and helps people to correctly execute the motion action.
Further, the data analysis unit is provided with a plurality of adjustment modes aiming at a matching mechanism of the gesture recognition unit for matching the human body gesture model under the condition that the variance of the deviation between the motion track of each joint point and the preset motion track is preliminarily judged to be in accordance with the preset variance standard, so that the judgment precision of the human body gesture monitoring system on the cause of the non-standard human body gesture is improved, a firm and reliable theoretical basis is provided for solving the problem of the non-standard human body gesture, and the improvement of the human body health of the system is promoted.
Further, the data analysis unit switches the initial static posture estimation mode of the posture estimation unit to the dynamic posture estimation mode under the condition that the variance of the deviation between the motion track of each joint point and the preset motion track is not in accordance with the preset variance standard, so that the judgment precision of the human posture monitoring system on the reasons of the non-standardization of the human posture is further improved, a firm and reliable theoretical basis is provided for solving the problem of the non-standardization of the human posture, and the system is promoted to improve the human health.
Further, the data analysis unit is provided with a plurality of optimization modes aiming at the coordinates of each joint point in the coordinate system under the condition that the variance of the deviation between the motion track of each joint point and the preset motion track is not in accordance with the preset variance standard, so that the judgment precision of the human body posture monitoring system on the reasons of the non-standardization of the human body posture is further improved, a firm and reliable theoretical basis is provided for solving the problem of the non-standardization of the human body posture, and the improvement of the human body health of the system is promoted.
Drawings
FIG. 1 is a block diagram of a human body posture monitoring system based on image data analysis according to an embodiment of the present invention;
FIG. 2 is a flowchart of a determination of whether the motion trail of each joint point is normal to the human body gesture by the data analysis unit according to the embodiment of the invention;
FIG. 3 is a flowchart of a secondary determination of whether a human body gesture is normalized by a data analysis unit according to an embodiment of the present invention, based on the number of articulation points where the coincidence ratio of an actual deflection track and a preset deflection track does not meet a preset coincidence ratio standard;
Fig. 4 is a flowchart of a determination of a cause of irregular human body posture by the data analysis unit according to the embodiment of the present invention based on the variance of the deviation between the motion trajectories of the joints and the preset motion trajectories.
Detailed Description
In order that the objects and advantages of the invention will become more apparent, the invention will be further described with reference to the following examples; it should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
It should be noted that, the data in this embodiment are obtained by comprehensively analyzing and evaluating the historical detection data and the corresponding historical detection results in three months before the current detection by the system of the present invention. The preset motion track of the system in the embodiment of the invention comprehensively determines the numerical value of each preset parameter standard aiming at the current monitoring according to the image information of 34027 times detected in the accumulation in the previous three months for the current human body monitoring; it can be understood by those skilled in the art that the determining manner of the system according to the present invention for the single item of parameter may be to select the value with the highest duty ratio as the preset standard parameter according to the data distribution, so long as the system according to the present invention can clearly define different specific situations in the single item determination process through the obtained value.
The above and further technical features and advantages of the present invention are described in more detail below with reference to the accompanying drawings.
Referring to fig. 1, a block diagram of a human body posture monitoring system based on image data analysis according to an embodiment of the present invention is shown; the human body posture monitoring system based on image data analysis comprises an image acquisition unit, an image preprocessing unit, a joint point detection unit, a posture estimation unit, a posture recognition unit, a data analysis unit and a prompt unit.
The image acquisition unit comprises six image sensors and is used for acquiring human body image information from different directions; the image preprocessing unit is connected with the image acquisition unit and is used for preprocessing the human body image information acquired by the image acquisition unit; the joint point detection unit is connected with the image preprocessing unit and is used for recognizing and detecting joint points of the preprocessed human body image information and marking the joint points; the gesture estimation unit is connected with the joint point detection unit and is used for preliminarily estimating the human gesture based on the distribution condition of each joint point; the preliminary estimated human body pose includes: standing, sitting, lying, walking and running; the gesture recognition unit is connected with the gesture estimation unit and is used for matching the estimated gesture with a predefined gesture model and judging a specific gesture; the data analysis unit is respectively connected with the joint point detection unit and the gesture recognition unit and is used for establishing a coordinate system based on the matched gesture model, determining the coordinates of each joint point and determining whether the current human gesture is standard or not based on the motion trail of each joint point; the data analysis unit is also used for determining the reason of the nonstandard human body posture under the condition of judging the nonstandard human body posture and sending a posture correcting command to the prompting unit; the prompting unit is connected with the data analysis unit and is used for sending out a notification of correcting the posture based on the reason that the current human posture is not standard and determined by the data analysis unit;
specifically, the human body posture model matching method comprises the following steps:
Firstly, selecting a proper mathematical model according to the characteristics and requirements of human body posture problems, and based on the complexity of a human body, adopting a model which is nonlinear;
The invention adopts 33047 times of image information for comparison and matching with the model, and the experimental data is as comprehensive and accurate as possible so as to ensure the reliability of model matching;
After experimental data are obtained, estimating parameters of the model, and adjusting the parameters of the model by using a numerical optimization method so as to fit the model with actual data as much as possible;
comparing the adjusted model with actual data, and evaluating the fitting degree between the model and the data by using statistical indexes, root Mean Square Error (RMSE), correlation coefficient and the like;
The invention uses 1000 times of image information to verify the reliability and generalization capability of the model; in general, human model matching is an iterative process that continually adjusts model parameters to better describe the actual situation.
Fig. 2 is a flowchart showing a determination of whether the motion trail of each joint point is normal to the human body gesture by the data analysis unit according to the embodiment of the present invention; the data analysis unit establishes a coordinate system based on the gesture model identified by the gesture identification unit, marks the joint points detected by the joint point detection unit in the coordinate system, and for the currently identified gesture model, the data analysis unit is provided with preset motion tracks for the joint points and a judgment mode for the human body gesture based on the actual motion tracks of the joint points, wherein:
The first judging mode is that the data analysis unit judges that the actual motion trail of each joint point accords with the preset motion trail and judges that the current human body gesture is standard; the first judging mode meets the condition that the motion trail of the joint points with more than 90% is matched with the preset motion trail;
The second judging mode is that the data analysis unit preliminarily judges that the actual motion trail of each joint point does not accord with the preset motion trail, and judges that the acquisition angle is changed to carry out secondary judgment; the second judging mode meets the condition that the motion trail of the joint points with the positions of more than 70% and less than 90% is matched with the preset motion trail;
The third judging mode is that the data analyzing unit judges that the actual motion trail of each joint point does not accord with the preset motion trail, and the data analyzing unit determines the cause of the nonstandard human body posture based on the joint points which do not accord with the preset motion trail; the third judging mode meets the condition that the motion trail of the joint points with more than 70% of the motion trail is matched with the preset motion trail;
Specifically, the invention acquires the reference for determination by collecting experimental data, and sequentially detects the motion trail of each articulation point based on the reference to determine whether the motion of the human body at the articulation point meets the preset standard, when the motion trail of a single articulation point is determined, the motion trail is primarily determined based on the determined standard, and when the coincidence ratio of the actual trail and the preset trail is between 70% and 90%, the data analysis unit cannot determine the actual state of the articulation point only through the acquired actual motion trail, so that other parameters are required to be combined to avoid the influence of external factors on the determination result.
Specifically, the invention is different from the human body posture monitoring system in the prior art in that the invention simulates the possible positions of the presumed articular point through a large amount of image information, then adjusts the detail parameters of the simulated presumption based on the comparison result of the positions of the actual articular point and the simulated presumption position, realizes the self-adaptive perfection of the simulated presumption process through a neural network algorithm, and selects the numerical range with the highest occupation ratio according to the data distribution for the preset track coincidence ratio of the articular point.
Referring to fig. 3, a flowchart of a secondary determination of whether the human body posture is normalized by the data analysis unit according to the embodiment of the present invention based on the number of the nodes where the coincidence ratio of the actual deflection track and the preset deflection track does not meet the preset coincidence ratio standard; the data analysis unit changes the acquisition angle of the image acquisition unit under the condition that the motion trail of each joint point accords with a second judgment mode, sets a preset deflection trail based on the variation of the acquisition angle and the initial acquisition angle, and judges whether the human body gesture is standard or not for the second time based on the number of the joint points, wherein the coincidence degree of the counted actual deflection trail and the preset deflection trail does not accord with the preset coincidence degree standard, and the method comprises the following steps:
the first secondary judgment mode is that the data analysis unit judges that the number of the articulation points of which the coincidence degree of the actual deflection track and the preset deflection track does not accord with the preset coincidence degree standard accords with the preset number standard, and judges that the current human body posture is standard; the first secondary judgment mode meets the condition that the number of the articulation points, the coincidence degree of the actual deflection track and the preset deflection track does not accord with the preset coincidence degree standard, is larger than the preset number standard, and the preset number standard is set to be 40; setting the preset overlap ratio standard to 90%;
The second secondary judgment mode is that the data analysis unit judges that the number of the articulation points of which the coincidence degree of the actual deflection track and the preset deflection track does not accord with the preset coincidence degree standard does not accord with the preset number standard, and judges that the current human body posture is not standard; the second secondary judgment mode meets the condition that the number of the articulation points of which the coincidence degree of the actual deflection track and the preset deflection track does not accord with the preset coincidence degree standard is smaller than or equal to a preset number standard;
Specifically, if there is a deviation between the motion track and the corresponding preset motion track and the deviation amount is in the preset section, the data analysis unit determines whether the deviation between the actual motion track and the preset motion track is caused by the acquired angle deviation based on the deviation value of the motion track and the preset motion track, so that the acquired angle for the human body is changed according to the actual deviation condition to avoid the problem, and accurately determines whether the motion of the human body at the joint point meets the preset standard through the motion track of each joint point after the avoidance.
Referring to fig. 4, a flowchart of a method for determining a cause of irregular human body posture by the data analysis unit according to the embodiment of the present invention based on variances of deviations between motion trajectories of various joints and preset motion trajectories is shown; the data analysis unit is provided with a reason judgment mode for the irregular reasons of the human body gesture based on the variance of the deviation of the movement track of each joint point and the preset movement track under the condition that the movement track of each joint point meets a third judgment mode, wherein:
The first cause judgment mode is that the data analysis unit judges that the variance of the deviation between the motion trail of each joint point and the preset motion trail meets the preset variance standard, and the data analysis unit sends a command for correcting the human body posture to the prompting unit; the first reason judging mode meets the condition that the variance of the deviation between the motion trail of each joint point and the preset motion trail is smaller than or equal to a first preset variance, and the first preset variance is set to be 30;
The second cause judgment mode is that the data analysis unit preliminarily judges that the variance of the deviation between the motion trail of each joint point and the preset motion trail meets the preset variance standard, and judges that the gesture recognition unit has a problem aiming at the matching of the gesture model; the second reason judging mode meets the condition that the variance of the deviation between the motion trail of each joint point and the preset motion trail is larger than the first preset variance and smaller than or equal to the second preset variance, and the second preset variance is set to be 50;
the third cause judgment mode is that the data analysis unit preliminarily judges that the variance of the deviation between the motion trail of each joint point and the preset motion trail does not accord with the preset variance standard, and judges that the posture has a change problem; the third factor judging mode meets the condition that the variance of the deviation between the motion trail and the preset motion trail of each joint point is larger than the second preset variance and smaller than or equal to the third preset variance, and the third preset variance is set to be 100;
The fourth reason judging mode is that the data analyzing unit judges that the variance of the deviation between the motion trail of each joint point and the preset motion trail does not accord with the preset variance standard, and judges that the joint point detecting unit has problems aiming at the positioning of each joint point; the fourth reason judgment mode meets the condition that the variance of the deviation of the motion trail of each joint point and the preset motion trail is larger than a third preset variance;
Specifically, the data analysis unit acquires the distribution condition of the tracks of all the joints of the human body based on the acquired variance of the deviation values of all the joints, and comprehensively evaluates whether the determination of the joints of the human body meets the standard based on the acquired variance, so that the smaller the variance is, the smaller the fluctuation of the deviation values of the motion tracks of all the joints and the corresponding preset motion tracks on the data level is, and the more accurate the positions of all the joints in the acquired human body is, therefore, when the variance value is minimum, the acquisition precision of the joints of the human body at the moment can be determined to meet the standard, and the reason that the human body posture does not meet the standard under the current condition is the fact that the human body posture is inconsistent with the preset posture is determined, and the data analysis unit sends a command for correcting the human body posture at the moment;
Similarly, when the collected variance is greater than the first preset variance and less than or equal to the second preset variance, only a small part of the movement tracks of the articulation points deviate from the preset movement tracks, which is caused by the difference of human bodies, and certain differences exist when the movement tracks are matched with the human body model, the embodiment adopts a form of replacing a matching mechanism to adjust the matching mechanism of rigid body transformation to the matching mechanism of a neural network, and improves the matching accuracy in a mode of sacrificing the operation speed;
When the acquired variance is larger than the second preset variance and smaller than or equal to the third preset variance, more than half of the movement tracks of the joints are deviated from the preset movement tracks, because the human body is switched to the sitting posture in a short time, for example, the standing posture is switched to the sitting posture, at the moment, the movement tracks of the joints at the upper half part of the human body are not changed greatly, but the movement tracks of the joints at the lower half part of the human body are in great ingress and egress with the standing posture, and the variance of the deviation value of each joint is great due to the hysteresis of operation, so that the human body data is required to be acquired again;
When the acquired variance is larger than the third preset variance, at this time, all the joint points of the human body have deviation in motion trail, the current human body posture and the human body posture model are not matched completely, the variance value of the deviation value of the motion trail of each joint point and the preset motion trail is large, and the data analysis unit judges that the joint point detection unit has problems aiming at the positioning of each joint point and needs to perform joint point positioning again.
With continued reference to fig. 1 to fig. 4, the data analysis unit initially determines that a problem occurs in matching the gesture recognition unit with respect to the gesture model when the variance of the deviation between the motion trajectories of the joint points and the preset motion trajectories meets the preset variance criterion, and the system initially defaults to a matching mechanism based on rigid transformation, which is mainly based on the principle of rigid transformation, and processes the human gesture model as a rigid body, specifically, firstly performs rigid transformation on the human gesture model, and then matches the transformed gesture model with the target gesture;
Under the condition that the data analysis unit preliminarily judges that the variance of the deviation of the motion trail of each joint point and the preset motion trail meets the preset variance standard, a matching mechanism adopted by the initial default of the system is switched to a matching mechanism based on a neural network, the mechanism is mainly based on the technology of the neural network, a human body posture model is taken as input, feature extraction and representation learning are carried out on the model through the neural network, and then the learned features are matched with the target posture.
With continued reference to fig. 1 to fig. 4, the data analysis unit is configured to switch the static posture estimation mode initiated by the posture estimation unit to the dynamic posture estimation mode when it is preliminarily determined that the variance of the deviation between the motion trajectories of the joint points and the preset motion trajectories does not meet the preset variance criterion;
specifically, the static posture estimation mode mainly analyzes a single image to infer the posture of the human body, and detects the position of a human body joint point to infer the posture of the human body; the dynamic posture estimation mode mainly considers the posture change of the human body in the motion process, continuous image information needs to be analyzed, and the dynamic posture is modeled and predicted based on a time sequence information method, namely, a time sequence model is established to capture the time change rule of the human body posture.
With continued reference to fig. 1 to fig. 4, the data analysis unit is configured to obtain reference data from the cloud for anomaly recognition when the posture estimation unit is switched to a dynamic posture estimation mode and still cannot accurately recognize the posture of the human body;
Specifically, when the dynamic posture estimation mode is switched to the dynamic posture estimation mode and still cannot be accurately identified, the fact that the human posture model matched with the dynamic posture estimation mode does not exist locally is judged, the abnormality of the human posture to be monitored is judged, and various abnormal information is required to be acquired from the cloud and identified.
With continued reference to fig. 1 to fig. 4, the data analysis unit sets an optimization mode for coordinates of each joint point in a coordinate system when it is determined that the variance of the deviation between the motion trail of each joint point and the preset motion trail does not meet the preset variance standard;
specifically, the embodiment of the invention optimizes the coordinates of each joint point by adopting a Newton method, constructs a secondary approximation model by calculating the first derivative and the second derivative of the human body posture model at each joint point, and then solves the minimum value point of the model as an updating direction to realize the optimization of the coordinates.
With continued reference to fig. 1 to fig. 4, after the coordinate optimization mode for each joint point is completed, if the variance of the deviation between the motion track of each joint point and the preset motion track still does not meet the preset variance standard, it is determined that the cause of the nonstandard human body posture of the monitored person is the stature reason, the data analysis unit imports the matching data retrieved at the cloud to the data analysis unit based on the motion track of each joint point to re-determine whether the motion track of each joint point meets the standard, if the cloud retrieves that the matching data does not exist, the current human body posture data is recorded, and the current human body posture data is uploaded to the cloud update cloud database.
Thus far, the technical solution of the present invention has been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of protection of the present invention is not limited to these specific embodiments. Equivalent modifications and substitutions for related technical features may be made by those skilled in the art without departing from the principles of the present invention, and such modifications and substitutions will be within the scope of the present invention.
The foregoing description is only of the preferred embodiments of the invention and is not intended to limit the invention; various modifications and variations of the present invention will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. Human posture monitoring system based on image data analysis, characterized by comprising:
The image acquisition unit comprises a plurality of image sensors and is used for acquiring human body image information from different directions;
The image preprocessing unit is connected with the image acquisition unit and is used for preprocessing the human body image information acquired by the image acquisition unit;
The joint point detection unit is connected with the image preprocessing unit and is used for identifying and detecting joint points of the preprocessed human body image information and marking the joint points;
The gesture estimation unit is connected with the joint point detection unit and is used for preliminarily estimating the human gesture based on the distribution situation of each joint point; the preliminary estimated body posture includes standing, sitting, lying, walking and running;
The gesture recognition unit is connected with the gesture estimation unit and is used for judging the specific gesture of the human body based on the matching result of the primarily estimated gesture and the corresponding predefined gesture model;
The data analysis unit is respectively connected with the joint point detection unit and the gesture recognition unit and is used for establishing a coordinate system based on the matched gesture model to determine the coordinates of each joint point in the coordinate system and determining whether the current human gesture is standard or not based on the motion track of each joint point in the coordinate system; the data analysis unit is also used for determining the cause of the nonstandard human body posture or sending a posture correcting instruction to the prompting unit under the condition that the current human body posture is judged to be nonstandard;
The prompting unit is connected with the data analysis unit and used for sending out a corresponding notification of correcting the posture based on the reason that the current human posture is not standard and determined by the data analysis unit;
the data analysis unit changes the acquisition angle to carry out secondary judgment under the condition that the movement track of each joint point is not in accordance with the preset movement track of the gesture model which is currently recognized by the data analysis unit, and determines the cause of the non-standard human gesture based on the joint points which are not in accordance with the preset movement track under the condition that the movement track of each joint point is not in accordance with the preset movement track of the gesture model which is currently recognized by the data analysis unit.
2. The human body posture monitoring system based on image data analysis according to claim 1, wherein the data analysis unit is configured to change an acquisition angle of the image acquisition unit for a human body in a case where it is primarily determined that a motion trajectory of each joint point does not conform to a preset motion trajectory of the data analysis unit for a currently recognized posture model, set a deflection trajectory preset for each of the joint points based on a variation of the acquisition angle and the initial acquisition angle, and secondarily determine whether the human body posture is normal based on the number of the joint points where a calculated coincidence degree of an actual deflection trajectory and the preset deflection trajectory does not conform to a preset coincidence degree standard.
3. The system according to claim 1, wherein the data analysis unit is configured to determine a cause of the lack of standardization of the human body posture based on a variance of a deviation of the movement locus of each joint point from the preset movement locus in a case where it is determined that the movement locus of each joint point does not conform to the preset movement locus of the gesture model currently recognized by the data analysis unit, and select a corresponding processing manner, including:
issuing a command for correcting the human body posture to the prompt unit under the condition that the variance accords with a preset variance standard;
Judging that the matching of the gesture recognition unit aiming at the gesture model is problematic under the condition that the preliminary judgment variance accords with a preset variance standard;
Determining that the posture has a change problem under the condition that the preliminary determination variance does not meet the preset variance standard, and re-determining the human posture based on the current state;
And judging that the positioning of the joint point detection unit for each joint point is problematic under the condition that the judgment variance does not accord with the preset variance standard.
4. The human body posture monitoring system based on image data analysis according to claim 3, wherein the data analysis unit is configured to provide a plurality of adjustment modes for the matching mechanism of the posture recognition unit to match the human body posture model in the case that the variance of the deviation between the motion track of each joint point and the preset motion track is preliminarily determined to be in accordance with the preset variance standard, and the accuracy of the posture recognition unit adjusted by using different adjustment modes to match the human body posture model is different.
5. The human body posture monitoring system based on image data analysis according to claim 3, wherein the data analysis unit is configured to switch the initial static posture estimation mode of the posture estimation unit to the dynamic posture estimation mode if it is preliminarily determined that the variance of the deviation of the motion trajectories of the respective joint points from the preset motion trajectories does not meet the preset variance criterion.
6. The system according to claim 5, wherein the data analysis unit is configured to obtain reference data from the cloud for human body anomaly recognition if the human body pose cannot be accurately recognized even if the pose estimation unit is switched to the dynamic pose estimation mode.
7. The human body posture monitoring system based on image data analysis according to claim 3, wherein the data analysis unit is configured to set a plurality of optimization modes for the coordinates of each joint point in the coordinate system in case that it is determined that the variance of the deviation between the motion trail of each joint point and the preset motion trail does not meet the preset variance standard, and the coordinates of the joint points are different after being adjusted by using different adjustment modes.
8. The human body posture monitoring system based on image data analysis according to claim 7, wherein the data analysis unit determines that the cause of the non-standardization of the human body posture of the monitored person is a stature cause in the case where the coordinate optimization for each joint point is completed and the variance of the deviation of the movement track of each joint point from the preset variance criterion is not met, and the data analysis unit imports the matching data retrieved at the cloud to the data analysis unit based on the movement track of each joint point to re-determine whether the movement track of each joint point meets the criterion.
9. The system according to claim 8, wherein the data analysis unit is configured to record current human body posture data when no matching data exists in the cloud search, and the data analysis unit is further configured to upload the recorded human body posture data to the cloud update cloud database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410385178.6A CN117994819B (en) | 2024-04-01 | 2024-04-01 | Human body posture monitoring system based on image data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410385178.6A CN117994819B (en) | 2024-04-01 | 2024-04-01 | Human body posture monitoring system based on image data analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117994819A CN117994819A (en) | 2024-05-07 |
| CN117994819B true CN117994819B (en) | 2024-06-07 |
Family
ID=90892660
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410385178.6A Active CN117994819B (en) | 2024-04-01 | 2024-04-01 | Human body posture monitoring system based on image data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117994819B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20020078826A (en) * | 2001-04-10 | 2002-10-19 | 현대자동차주식회사 | Calibration method of active safety system using Camera, and error detection method of the system |
| CN103810478A (en) * | 2014-02-21 | 2014-05-21 | 广东小天才科技有限公司 | Sitting posture detection method and device |
| CN105105757A (en) * | 2015-07-29 | 2015-12-02 | 南开大学 | Wearable human motion gesture track recording and assessment device |
| CN108764066A (en) * | 2018-05-08 | 2018-11-06 | 南京邮电大学 | A kind of express delivery sorting working specification detection method based on deep learning |
| WO2021237914A1 (en) * | 2020-05-27 | 2021-12-02 | 大连成者科技有限公司 | Sitting posture monitoring system based on monocular camera sitting posture recognition technology |
| CN115497596A (en) * | 2022-11-18 | 2022-12-20 | 深圳聚邦云天科技有限公司 | Human body motion process posture correction method and system based on Internet of things |
| CN116978127A (en) * | 2023-08-10 | 2023-10-31 | 天翼物联科技有限公司 | Body-building posture correction method, device, equipment and medium based on posture estimation |
-
2024
- 2024-04-01 CN CN202410385178.6A patent/CN117994819B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20020078826A (en) * | 2001-04-10 | 2002-10-19 | 현대자동차주식회사 | Calibration method of active safety system using Camera, and error detection method of the system |
| CN103810478A (en) * | 2014-02-21 | 2014-05-21 | 广东小天才科技有限公司 | Sitting posture detection method and device |
| CN105105757A (en) * | 2015-07-29 | 2015-12-02 | 南开大学 | Wearable human motion gesture track recording and assessment device |
| CN108764066A (en) * | 2018-05-08 | 2018-11-06 | 南京邮电大学 | A kind of express delivery sorting working specification detection method based on deep learning |
| WO2021237914A1 (en) * | 2020-05-27 | 2021-12-02 | 大连成者科技有限公司 | Sitting posture monitoring system based on monocular camera sitting posture recognition technology |
| CN115497596A (en) * | 2022-11-18 | 2022-12-20 | 深圳聚邦云天科技有限公司 | Human body motion process posture correction method and system based on Internet of things |
| CN116978127A (en) * | 2023-08-10 | 2023-10-31 | 天翼物联科技有限公司 | Body-building posture correction method, device, equipment and medium based on posture estimation |
Non-Patent Citations (1)
| Title |
|---|
| 基于双向LSTM的复杂环境下实时人体姿势识别;周意乔;徐昱琳;;仪器仪表学报;20200315(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117994819A (en) | 2024-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109871750B (en) | Gait recognition method based on skeleton diagram sequence abnormal joint repair | |
| US8135220B2 (en) | Face recognition system and method based on adaptive learning | |
| JP5285575B2 (en) | Human behavior determination device and program thereof | |
| JP5753966B2 (en) | Face image registration apparatus and method | |
| KR101434768B1 (en) | Moving object tracking system and moving object tracking method | |
| CN109460702A (en) | Passenger's abnormal behaviour recognition methods based on human skeleton sequence | |
| JP6973258B2 (en) | Image analyzers, methods and programs | |
| CN110674785A (en) | Multi-person posture analysis method based on human body key point tracking | |
| CN107220596B (en) | Human body posture estimation method based on cascade error correction mechanism | |
| US11858140B2 (en) | Robot system and supplemental learning method | |
| WO2009111499A2 (en) | Dynamic object classification | |
| CN112016409A (en) | Deep learning-based process step specification visual identification determination method and system | |
| US20220414291A1 (en) | Device for Defining a Sequence of Movements in a Generic Model | |
| CN113192105A (en) | Method and device for tracking multiple persons and estimating postures indoors | |
| CN114625251A (en) | Interaction method and device based on VR, computer equipment and storage medium | |
| JP2021086322A5 (en) | ||
| CN117994819B (en) | Human body posture monitoring system based on image data analysis | |
| CN110929555B (en) | Face recognition method and electronic device using same | |
| KR20210041330A (en) | Fastening tool system and control method thereof | |
| Hu et al. | Multi-user identification and efficient user approaching by fusing robot and ambient sensors | |
| CN117137435B (en) | Rehabilitation action recognition method and system based on multi-mode information fusion | |
| CN115586748B (en) | Mobile intelligent flexible motion control system and method thereof | |
| CN112307897A (en) | Pet tracking method based on local feature recognition and adjacent frame matching in community monitoring scene | |
| CN113195177B (en) | Robot control device, robot system, and robot control method | |
| CN115205958A (en) | Method for detecting human body action and feeding back by using mobile equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |