CN113743255A - Neural network-based child sitting posture identification and correction method and system - Google Patents
Neural network-based child sitting posture identification and correction method and system Download PDFInfo
- Publication number
- CN113743255A CN113743255A CN202110950268.1A CN202110950268A CN113743255A CN 113743255 A CN113743255 A CN 113743255A CN 202110950268 A CN202110950268 A CN 202110950268A CN 113743255 A CN113743255 A CN 113743255A
- Authority
- CN
- China
- Prior art keywords
- sitting posture
- child
- time
- neural network
- sitting
- 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.)
- Pending
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 52
- 238000012937 correction Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000002159 abnormal effect Effects 0.000 claims abstract description 28
- 238000004891 communication Methods 0.000 claims abstract description 19
- 238000010586 diagram Methods 0.000 claims description 29
- 230000002093 peripheral effect Effects 0.000 claims description 14
- 238000012544 monitoring process Methods 0.000 claims description 13
- 230000003993 interaction Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 8
- 230000004438 eyesight Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 4
- 238000005452 bending Methods 0.000 claims description 3
- 230000002746 orthostatic effect Effects 0.000 claims description 3
- 210000000707 wrist Anatomy 0.000 claims description 3
- 230000036544 posture Effects 0.000 claims 57
- 230000000694 effects Effects 0.000 abstract description 3
- 210000002414 leg Anatomy 0.000 description 3
- 210000003127 knee Anatomy 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 206010058314 Dysplasia Diseases 0.000 description 1
- 208000000875 Spinal Curvatures Diseases 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000009223 counseling Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 210000003414 extremity Anatomy 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 230000004379 myopia Effects 0.000 description 1
- 208000001491 myopia Diseases 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention discloses a child sitting posture identification and correction method and system based on a neural network, which comprises the following steps of S1, setting interval time, light reminding time and heavy reminding time; s2, when the sitting posture image of the child is obtained, recognizing the current sitting posture of the child and outputting a sitting posture recognition result; s3, judging based on the sitting posture identification result, repeating the step S2 if the sitting posture identification result is normal sitting posture, repeating the step S2 if the sitting posture identification result is abnormal sitting posture, and entering the step S4; s4, timing at the time point of the abnormal sitting posture; s5, acquiring timing time and judging; if the timing time is longer than the mild reminding time and shorter than the severe reminding time, outputting sitting posture correction prompt information; if the timing time is longer than the heavy reminding time, outputting prompt information to remind a child guardian; and S6, the guardian communicates with the child in a remote communication mode to correct the sitting posture. This application has the effect of correction children's study position of sitting that can be better.
Description
Technical Field
The application relates to the field of intelligent home control systems and human health, in particular to a children sitting posture recognition and correction method and system based on a neural network.
Background
In recent years, the number of children with myopia and spinal curvature caused by improper sitting posture is gradually increasing, and the improper sitting posture is one of the main causes of the visual defects and the spinal dysplasia of the children and is very harmful to the healthy development of the children. Because many parents are because of the reason of working, parents can not be fine supervise children at the scene and study, can not know children's study state and study duration at any time, more can not correct children's bad position of sitting at any time.
At present, most of the related products for correcting the sitting posture of children in the market adopt seats with special structural designs. In the mode, the aim of correcting the sitting posture is difficult to achieve because the child feels to maintain the sitting posture to a great extent, and the child is young and has difficulty in maintaining self-control.
Disclosure of Invention
In order to guarantee cultivation of healthy learning habits of children, strengthen sitting posture correction of the children during learning and promote attention of parents to the children, the application provides a children sitting posture identification correction method and system based on a neural network.
In a first aspect, the application provides a child sitting posture identification and correction method based on a neural network, which adopts the following technical scheme:
a children sitting posture identification and correction method based on a neural network comprises the following steps,
s1, setting interval time, light reminding time and heavy reminding time, wherein the value of the interval time is smaller than that of the light reminding time, and the value of the light reminding time is smaller than that of the heavy reminding time;
s2, when the interval time arrives and the child sitting posture image is obtained, identifying the current sitting posture of the child based on the child sitting posture image and outputting a sitting posture identification result, wherein the sitting posture identification result comprises an orthostatic sitting posture and an anomalstatic sitting posture;
s3, judging based on the sitting posture identification result,
if the sitting posture recognition result is normal sitting posture, the step S2 is repeated to keep monitoring,
if the sitting posture identification result is the abnormal sitting posture, the step S2 is repeated and the process goes to the step S4;
s4, starting timing at the time point when the abnormal sitting posture is judged;
s5, acquiring timing time, and judging based on the light reminding time and the heavy reminding time;
if the timing time is longer than the mild reminding time and shorter than the severe reminding time, outputting sitting posture correction prompt information, wherein the sitting posture correction prompt information is preset information and comprises voice reminding and vibration reminding;
if the timing time is greater than the heavy reminding time, outputting prompt information to remind a child guardian;
s6, after the child guardian obtains the prompt information, the child guardian communicates with the child in a remote communication mode
To complete the correction.
By adopting the technical scheme, after the sitting posture recognition result of the child is obtained, the time for keeping the child in the abnormal sitting posture is timed, and different processing is performed after the mild reminding time and the severe reminding time are respectively reached, so that the sitting posture condition of the child during learning can be more intelligently monitored; and interval time, heavy reminding time and light reminding time all can be through the mode input of predetermineeing in advance, time measuring is carried out monitoring to children's position of sitting, monitoring time can be according to the more nimble setting of different children's the condition, reach light reminding time back at the abnormal holding time of position of sitting, send out prompt information automatically and remind children to correct, reach heavy reminding time back, inform guardian to correct, can guarantee the cultivation to children's healthy learning habit, strengthen the position of sitting when children learn and correct, promote father and mother's attention to children, with the position of sitting of better correction children when studying.
Optionally, in step S2, the current sitting posture of the child is identified based on the sitting posture image of the child
The specific method for posture and outputting the sitting posture recognition result comprises the following steps,
s21, calling the child sitting posture image and inputting the child sitting posture image into a neural network;
s22, detecting and marking unoccluded human body key points such as the head, the wrist or the joints in the sitting posture image of the child by the neural network, and performing multiple predictions by neural network convolution to obtain each key point hot spot diagram;
s23, connecting all key points through color gamut continuity and truncation forced continuity optimization algorithm by using color gamut and key point distance information to form an initial human posture tree diagram;
s24, based on the initial human body posture tree diagram, predicting lost key points to form a final human body posture tree diagram, and distinguishing each trunk through the connection of different key points;
s25, predicting the human body posture through the angle between the connecting lines in the final human body posture tree diagram to obtain a sitting posture recognition result;
and S26, outputting a sitting posture identification result.
By adopting the technical scheme, the neural network firstly acquires the key point thermodynamic diagram through the child sitting posture image, then based on the key point thermodynamic diagram, the color gamut continuity and cutoff forced continuity optimization algorithm is used for connecting each key point in the key point thermodynamic diagram to obtain the initial human posture tree diagram, then the predicted key point is obtained, the final human posture tree diagram is obtained through the predicted key point and the initial human posture tree diagram, the sitting posture recognition result is obtained through the angle between each connecting line in the final human posture tree diagram and is output, so that the child sitting posture in the child sitting posture image is recognized, and the child sitting posture recognition method has the effect of being capable of recognizing the child sitting posture.
Optionally, in step S24, the method for predicting the missing keypoint includes: and performing mutual learning and conjecture on the obtained dendrogram and the key point algorithm conjecture to obtain the lost key point.
Optionally, the child sitting posture image at least contains human body features of 70% of the human body in the longitudinal direction, the human body should contain 70% of the human body features in the transverse direction, the human body information should occupy more than twenty percent of pixels in the child sitting posture image, and the obtained human body key points are not less than 60% of key points required by posture estimation.
By adopting the technical scheme, the obtained key points of the human body are not less than 60% of the key points required by posture estimation, and the key points in the child sitting posture image can be more accurately obtained.
Optionally, the change of the sitting posture of the child is determined based on the sitting posture recognition result, and when the sitting posture of the child is changed from the normal sitting posture to the abnormal sitting posture or the sitting posture of the child is changed from the abnormal sitting posture to the normal sitting posture, the time of the current moment is recorded and stored.
Through adopting above-mentioned technical scheme, if children's position of sitting changes, no matter change to the abnormal sitting position or change to the normal sitting position from the abnormal sitting position from the normal sitting position, can both reflect children's position of sitting change habit when studying, through the record and save, children's guardian can obtain children's position of sitting change habit in the backtracking of later stage to adjust interval time, light reminding time and heavy reminding time.
Optionally, in step S2, if the neural network does not detect a human body key point in the child sitting posture image, the current time is saved, and the time is saved again after the human body key point is detected, so as to count the time that the child has not learned.
Through adopting above-mentioned technical scheme, children's guardian can judge children's study custom through the children who counts not study time to strengthen the relation between parents children, better counseling children study.
Optionally, the out-of-position posture includes an excessive back flexion, a head too close to the table, a knee bending and placing one leg on the other, and so on.
In a second aspect, the application provides a child sitting posture recognition and correction system based on a neural network, which adopts the following technical scheme:
a children sitting posture recognition and correction system based on a neural network comprises a computer control end and an internet intelligent terminal, wherein the computer control end is in communication connection with the internet intelligent terminal; the method is characterized in that:
the computer control end comprises a camera module, a computer vision neural network system, a timing timer, a peripheral control system, a loudspeaker and a microphone; wherein the content of the first and second substances,
the camera module is used for capturing the current child image and outputting a child sitting posture image;
the computer vision neural network system is preset with a trained neural network, and the neural network takes a child sitting posture image captured by the camera module as input and carries out processing to obtain a child sitting posture recognition result;
the timing timer is used for timing, and the interval time, the light reminding time and the heavy reminding time are all preset in the timing timer;
the peripheral control system is used for calling the camera module, the loudspeaker and the microphone;
the loudspeaker plays prompt voice or user dialogue information when the sitting posture information of the child is abnormal sitting posture;
the microphone is used for inputting user audio;
the internet intelligent terminal comprises a cloud server and a remote intelligent terminal; wherein the content of the first and second substances,
the cloud server is used for storing, calculating and processing various effective information of the PC terminal; and the remote intelligent terminal finally performs data interaction with the computer control terminal through the cloud server, and the effective information comprises sitting posture information of the children at certain interval time points, the duration of the children who do not learn, reminding information and communication information.
By adopting the technical scheme, the computer control end and the internet intelligent terminal can be matched for use, the computer control end controls the camera module, the loudspeaker and the microphone through the peripheral control module, wherein the camera module can acquire a sitting posture image of a child, the microphone captures audio of the child, the computer vision neural network system can process the sitting posture image of the child, and then the loudspeaker is used for playing prompt voice or user conversation information; a cloud server and a remote intelligent terminal in the internet intelligent terminal can store information, realize data interaction, transmit communication information, reminding information and the like of a guardian; the system can realize the method to complete the detection and correction of the sitting posture of the child.
Optionally, a WIFI communication gateway, a router and an Internet terminal are further arranged, and the computer control end and the Internet intelligent terminal are interconnected through the WIFI communication gateway, the router and the Internet terminal to perform data interaction.
Optionally, the remote intelligent terminal includes an intelligent device such as a mobile phone, a computer, and a tablet computer.
Through adopting above-mentioned technical scheme, children's guardian can monitor children's learning condition and position of sitting condition at any time to timely correction.
In summary, the present application includes at least one of the following beneficial technical effects: after the sitting posture recognition result of the child is obtained, the time for the child to keep the abnormal sitting posture is timed, different processing is carried out after the mild reminding time and the severe reminding time are respectively reached, and the sitting posture condition of the child during learning can be monitored more intelligently; and interval time, heavy reminding time and light reminding time all can be through the mode input of predetermineeing in advance, time measuring is carried out monitoring to children's position of sitting, monitoring time can be according to the more nimble setting of different children's the condition, reach light reminding time back at the abnormal holding time of position of sitting, send out prompt information automatically and remind children to correct, reach heavy reminding time back, inform guardian to correct, can guarantee the cultivation to children's healthy learning habit, strengthen the position of sitting when children learn and correct, promote father and mother's attention to children, with the position of sitting of better correction children when studying.
Drawings
FIG. 1 is a block flow diagram of the method of the present invention.
Fig. 2 is a block flow diagram of step S2 of the method of the present invention.
Fig. 3 is a schematic view of the arrangement of the image pickup apparatus in the present invention.
FIG. 4 is a schematic diagram of the system of the present invention.
Detailed Description
The present application is described in further detail below with reference to figures 1-4.
The embodiment of the application discloses a child sitting posture identification and correction method and system based on a neural network.
A children sitting posture identification correction method based on a neural network refers to fig. 1 and 2, and comprises the following steps:
and S1, setting interval time, light reminding time and heavy reminding time, wherein the value of the interval time is less than that of the light reminding time, and the value of the light reminding time is less than that of the heavy reminding time. The interval time, the light reminding time and the heavy reminding time can be preset by a child guardian, and in the preset process, the child guardian can set the interval time, the light reminding time and the heavy reminding time according to the sitting posture habit of the child; for example, in the performance of children at ordinary times, the condition that the position of sitting is bad appears comparatively easily, then all set up interval time, mild warning time and severe warning time to the short time, can improve the frequency of correcting children's position of sitting like this to better correction children's position of sitting. If children in the performance at ordinary times, the bad condition of position of sitting can appear only occasionally, can adjust interval time, slight warning time and heavy warning time all to lengthen, can also not disturb children's learning process when monitoring children's position of sitting like this.
S2, acquiring a sitting posture image of the child when the interval time arrives, recognizing the current sitting posture of the child based on the sitting posture image of the child and outputting a sitting posture recognition result, wherein the sitting posture recognition result comprises an orthostatic sitting posture and an anomalposition sitting posture; the abnormal sitting posture includes the condition that the back is excessively bent, the head is too close to the table surface, one leg is bent to knee and placed on the other leg, and the like, and the normal sitting posture excludes the sitting posture.
Timing is carried out when monitoring is started, when monitoring time is consistent with interval time after monitoring is started, a child sitting posture image is obtained, and the obtained image is a color image. And when the child sitting posture image is obtained, the device for obtaining the image is arranged as shown in fig. 3, so that the obtained child sitting posture image at least comprises human body features of 70% of the longitudinal direction of a human body, the transverse direction of the human body at least comprises 70% of the human body features, human body information occupies more than twenty percent of pixels in the child sitting posture image, and the key points of the human body which can be obtained are not less than 60% of the key points required by posture estimation.
Referring to fig. 2, the specific method of recognizing the current sitting posture of the child based on the child sitting posture image and outputting a sitting posture recognition result in step S2 includes the steps of:
s21, calling the child sitting posture image and inputting the child sitting posture image into a neural network;
and S22, detecting and marking unoccluded human body key points such as the head, the wrist or the joints in the sitting posture image of the child by the neural network, and performing multiple predictions by the neural network convolution to obtain each key point hot spot diagram.
And S23, connecting all the key points through a color gamut continuity and truncation forced continuity optimization algorithm by using the color gamut and key point distance information to form an initial human posture tree diagram.
And S24, predicting lost key points based on the initial human body posture tree diagram to form a final human body posture tree diagram, and distinguishing each trunk through the connection of different key points. In the process of obtaining the key points, the key points may be lost due to the shielding of the table and the chair on one part of the limbs of the child, and the obtained initial human posture tree diagram is inaccurate; at this time, the lost key point is obtained by means of mutual learning and conjecture of the obtained tree diagram and the key point algorithm conjecture. After the lost key points and the existing key points are connected, a final human body posture tree diagram can be formed, and the posture of the child closest to the actual situation can be reflected.
And S25, connecting all key points in the final human posture tree diagram, and predicting the human posture through the angle between all connecting lines in the final human posture tree diagram to obtain a sitting posture recognition result. After the key points are connected into the connecting line, the body posture of the child kept in the currently input child sitting posture image can be preliminarily predicted and judged through angle analysis of the connecting line.
And S26, outputting a sitting posture identification result. The two sitting posture identification results, namely the normal sitting posture or the abnormal sitting posture, are output through an output layer of the neural network.
And S3, judging based on the sitting posture identification result, and directly judging the sitting posture identification result after the sitting posture identification result is output by the neural network so as to judge whether prompt information needs to be sent out. The determination process is as follows:
if the sitting posture identification result is normal sitting posture, repeating the step S2 to keep monitoring;
if the sitting posture recognition result is the abnormal sitting posture, the process proceeds to step S4 while repeating step S2.
S4, starting timing at the time point when the abnormal sitting posture is judged; through the timing, can take place to record children's position of sitting change time when changing at children's position of sitting, judge the severity that children's position of sitting changes through the timing time to carry out corresponding correction.
S5, acquiring timing time, and judging based on the light reminding time and the heavy reminding time;
if the timing time is longer than the mild reminding time and shorter than the severe reminding time, outputting sitting posture correction prompt information, wherein the sitting posture correction prompt information is preset information and comprises voice reminding and vibration reminding.
In a specific embodiment, when the sitting posture information of the child is identified as the abnormal sitting posture, the sitting posture correction prompt information is output, and the sitting posture correction prompt information can be preset by a guardian of the child and is set to be a relatively gentle reminding voice, such as 'sitting posture adjustment and back straightening', so that the adverse psychology of the child can be relieved.
If the timing time is longer than the heavy reminding time, outputting prompt information to remind a child guardian;
s6, after the child guardian obtains the prompt information, the child guardian communicates with the child in a remote communication mode
To complete the correction.
After the sitting posture recognition result of the child is obtained, the time for the child to keep the abnormal sitting posture is timed, different processing is carried out after the mild reminding time and the severe reminding time are respectively reached, and the sitting posture condition of the child during learning can be monitored more intelligently; and interval time, heavy reminding time and light reminding time all can be through the mode input of predetermineeing in advance, time measuring is carried out monitoring children's position of sitting, monitoring time can be according to the more nimble setting of different children's the condition, reach light reminding time after the retention time of position of sitting anomaly, automatic suggestion information of sending out reminds children to correct, reach heavy reminding time after, inform guardian to correct, can guarantee the cultivation to children's healthy study custom, strengthen the position of sitting when children study and correct, promote father and mother's attention to children, with the better position of sitting of correction children when studying
And judging the change of the sitting posture of the child based on the sitting posture identification result, and recording and storing the time of the current moment when the sitting posture of the child is changed from the normal sitting posture to the abnormal sitting posture or the sitting posture of the child is changed from the abnormal sitting posture to the normal sitting posture.
After the system is started, if the neural network does not detect the human body key point in the child sitting posture image, the current time is saved, and the time is saved again after the human body key point is detected, so that the non-learning time of the child is counted. By detecting the key points, if the key points are not detected, it can be stated that no children exist in the camera area, timing is carried out after the key points are not detected, a guardian can more conveniently acquire the learning condition of the children, and information prompting learning is output at the moment.
In a specific embodiment, the computer performs timing after detecting the key point, and when the sitting posture information of the child is identified that the sitting time exceeds 60 minutes, the system plays a reminding voice twice to remind the child to have a rest. When the sitting posture information of the children is identified as that the sitting time exceeds 120 minutes, the system continuously plays reminding voice to remind the children to have a rest. Meanwhile, when the standing of the child is detected, namely the key point is not detected, the computer judges that the time is more than 5 minutes in an unmanned state, and then the child can enter the learning again. Through setting up the suggestion, can remind children to have a rest after the study time overlength to form better learning habit.
This embodiment still provides a children's position of sitting discernment correction system based on neural network, refers to fig. 4, including computer control end and internet intelligent terminal, computer control end and internet intelligent terminal communication connection. A WIFI communication gateway is arranged in the computer control end, a router and an Internet terminal are arranged in the Internet intelligent terminal, and the computer control end and the Internet intelligent terminal are interconnected through the WIFI communication gateway, the router and the Internet terminal to perform data interaction.
Referring to fig. 4, the computer control terminal includes a PC computer, a camera module, a computer vision neural network system, a timing timer, a peripheral control system, a speaker and a microphone; the PC computer is used for providing software and hardware basis required by program operation, and peripheral devices such as a camera module, a loudspeaker, a microphone and the like are connected with the PC computer and can carry out data interaction; the computer visual neural network system, the timing timer and the peripheral control system are programs installed in the PC computer, and functions can be called and realized by using the PC computer.
The camera module is used for capturing the current image of the child and outputting a sitting posture image of the child after capturing; the camera module adopts a high-definition camera, can capture color images and input the color images into a PC computer, and the acquired images are one frame of images and can be directly processed by a computer vision neural network system.
The trained neural network is preset in the computer vision neural network system, and the neural network takes the child sitting posture image captured by the camera module as input and carries out processing to obtain a child sitting posture recognition result.
And the timing timer is used for timing, and the interval time, the light reminding time and the heavy reminding time in the method are all preset in the timing timer.
The peripheral control system is used for calling the camera module, the loudspeaker and the microphone; the PC computer is provided with a driver of various peripherals, and the peripheral control system can directly call the peripherals through the driver.
The loudspeaker plays prompt voice or user conversation information when the sitting posture information of the child is abnormal sitting posture; a microphone for entering user audio. When the guardian corrects the sitting posture of the child in a communication mode, the loudspeaker can output user conversation information, and the child can reply to the guardian through the microphone.
Referring to fig. 4, the internet intelligent terminal includes a cloud server and a remote intelligent terminal; the cloud server is used for storing, calculating and processing various effective information of the PC terminal; existing servers may be leased directly.
The remote intelligent terminal comprises intelligent equipment such as a mobile phone, a computer, a tablet personal computer and the like, performs data interaction with the computer control terminal through the cloud server, and performs data interaction with the cloud server by downloading and installing a corresponding program on the remote intelligent terminal; the interactive effective information comprises sitting posture information of the children at certain time intervals, the duration of the children who do not learn, reminding information and communication information.
In a specific embodiment, when a child and a parent have a voice conversation, the computer control terminal obtains voice input information of the child conversation through the microphone module, and related information reaches the internet intelligent terminal through the WI-FI communication gateway and reaches the remote intelligent terminal through the internet intelligent terminal, namely, a guardian's remote device such as a PC, a mobile phone and the like for playing.
The computer control end can call the loudspeaker to play the voice conversation information from the guardian through the peripheral control system. The conversation information from the guardian reaches the Internet intelligent terminal from the remote equipment of the guardian, reaches the built-in network card of the computer control end through the router and the Internet terminal or is connected to the built-in network card through a wire, and finally can be called by the peripheral control module.
Through setting up remote intelligent terminal, various information when children's guardian can be convenient acquireing children study to judge whether children have the position of sitting unusual or some other behaviors, the help guardian that can be better pays close attention to children, with the position of sitting of better correction children when studying.
The above embodiments are preferred embodiments of the present application, and the protection scope of the present application is not limited by the above embodiments, so: all equivalent changes made according to the structure, shape and principle of the present application shall be covered by the protection scope of the present application.
Claims (10)
1. A child sitting posture identification and correction method based on a neural network is characterized by comprising the following steps: comprises the following steps of (a) carrying out,
s1, setting interval time, light reminding time and heavy reminding time, wherein the value of the interval time is smaller than that of the light reminding time, and the value of the light reminding time is smaller than that of the heavy reminding time;
s2, when the interval time arrives and the child sitting posture image is obtained, identifying the current sitting posture of the child based on the child sitting posture image and outputting a sitting posture identification result, wherein the sitting posture identification result comprises an orthostatic sitting posture and an anomalstatic sitting posture;
s3, judging based on the sitting posture identification result,
if the sitting posture recognition result is normal sitting posture, the step S2 is repeated to keep monitoring,
if the sitting posture identification result is the abnormal sitting posture, the step S2 is repeated and the process goes to the step S4;
s4, starting timing at the time point when the abnormal sitting posture is judged;
s5, acquiring timing time, and judging based on the light reminding time and the heavy reminding time;
if the timing time is longer than the mild reminding time and shorter than the severe reminding time, outputting sitting posture correction prompt information, wherein the sitting posture correction prompt information is preset information and comprises voice reminding and vibration reminding;
if the timing time is greater than the heavy reminding time, outputting prompt information to remind a child guardian;
and S6, after the child guardian acquires the prompt information, the child guardian communicates with the child in a remote communication mode to finish correction.
2. The neural network-based child sitting posture recognition correction method according to claim 1, characterized in that: in step S2, the specific method for recognizing the current sitting posture of the child based on the sitting posture image of the child and outputting the sitting posture recognition result includes,
s21, calling the child sitting posture image and inputting the child sitting posture image into a neural network;
s22, detecting and marking unoccluded human body key points such as the head, the wrist or the joints in the sitting posture image of the child by the neural network, and performing multiple predictions by neural network convolution to obtain each key point hot spot diagram;
s23, connecting all key points through color gamut continuity and truncation forced continuity optimization algorithm by using color gamut and key point distance information to form an initial human posture tree diagram;
s24, based on the initial human body posture tree diagram, predicting lost key points to form a final human body posture tree diagram, and distinguishing each trunk through the connection of different key points;
s25, predicting the human body posture through the angle between the connecting lines in the final human body posture tree diagram to obtain a sitting posture recognition result;
and S26, outputting a sitting posture identification result.
3. The neural network-based child sitting posture recognition correction method according to claim 2, characterized in that: in step S24, the method for predicting the missing keypoints includes: and performing mutual learning and conjecture on the obtained dendrogram and the key point algorithm conjecture to obtain the lost key point.
4. The neural network-based child sitting posture recognition correction method according to claim 3, wherein: the child sitting posture image at least comprises human body features which are 70% of the human body in the longitudinal direction, the human body features which are 70% of the human body features in the transverse direction, the human body information occupies more than twenty percent of pixels in the child sitting posture image, and the human body key points which can be obtained are not less than 60% of key points required by posture estimation.
5. The neural network-based child sitting posture recognition correction method according to claim 1, characterized in that: and judging the change of the sitting posture of the child based on the sitting posture identification result, and recording and storing the time of the current moment when the sitting posture of the child is changed from the normal sitting posture to the abnormal sitting posture or the sitting posture of the child is changed from the abnormal sitting posture to the normal sitting posture.
6. The neural network-based child sitting posture recognition correction method according to claim 1, characterized in that: in step S2, if the neural network does not detect a human body key point in the child sitting image, the current time is saved, and the time is saved again after the human body key point is detected, so as to count the time that the child has not learned.
7. The neural network-based child sitting posture recognition correction method according to claim 1, characterized in that: the abnormal sitting postures include excessive bending of the back, too close head to the table, bending one leg and placing the other leg in the upper sitting posture.
8. A children sitting posture recognition and correction system based on a neural network for realizing the method of any one of claims 1-7 comprises a computer control terminal and an Internet intelligent terminal, wherein the computer control terminal is in communication connection with the Internet intelligent terminal; the method is characterized in that:
the computer control end comprises a camera module, a computer vision neural network system, a timing timer, a peripheral control system, a loudspeaker and a microphone; wherein the content of the first and second substances,
the camera module is used for capturing the current child image and outputting a child sitting posture image;
the computer vision neural network system is preset with a trained neural network, and the neural network takes a child sitting posture image captured by the camera module as input and carries out processing to obtain a child sitting posture recognition result;
the timing timer is used for timing, and the interval time, the light reminding time and the heavy reminding time are all preset in the timing timer;
the peripheral control system is used for calling the camera module, the loudspeaker and the microphone;
the loudspeaker plays prompt voice or user dialogue information when the sitting posture information of the child is abnormal sitting posture;
the microphone is used for inputting user audio;
the internet intelligent terminal comprises a cloud server and a remote intelligent terminal; wherein the content of the first and second substances,
the cloud server is used for storing, calculating and processing various effective information of the PC terminal; and the remote intelligent terminal finally performs data interaction with the computer control terminal through the cloud server, and the effective information comprises sitting posture information of the children at certain interval time points, the duration of the children who do not learn, reminding information and communication information.
9. The neural network-based child sitting posture recognition correcting system of claim 8, wherein: the intelligent terminal is also provided with a WIFI communication gateway, a router and an Internet terminal, and the computer control end is interconnected with the Internet intelligent terminal through the WIFI communication gateway, the router and the Internet terminal so as to carry out data interaction.
10. The neural network-based child sitting posture recognition correcting system of claim 8, wherein: the remote intelligent terminal comprises intelligent equipment such as a mobile phone, a computer and a tablet personal computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110950268.1A CN113743255A (en) | 2021-08-18 | 2021-08-18 | Neural network-based child sitting posture identification and correction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110950268.1A CN113743255A (en) | 2021-08-18 | 2021-08-18 | Neural network-based child sitting posture identification and correction method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113743255A true CN113743255A (en) | 2021-12-03 |
Family
ID=78731669
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110950268.1A Pending CN113743255A (en) | 2021-08-18 | 2021-08-18 | Neural network-based child sitting posture identification and correction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113743255A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494441A (en) * | 2022-04-01 | 2022-05-13 | 广东机电职业技术学院 | Grape and picking point synchronous identification and positioning method and device based on deep learning |
| CN114973423A (en) * | 2022-07-28 | 2022-08-30 | 聊城市飓风工业设计有限公司 | Warning method and system for sitting posture monitoring of child learning table |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104916093A (en) * | 2015-06-05 | 2015-09-16 | 重庆智韬信息技术中心 | Method for realizing intelligent correction prompting of students' sitting postures |
| CN111091046A (en) * | 2019-10-28 | 2020-05-01 | 北京灵鹦科技有限公司 | User bad sitting posture correction system based on machine vision |
| CN111178280A (en) * | 2019-12-31 | 2020-05-19 | 北京儒博科技有限公司 | Human body sitting posture identification method, device, equipment and storage medium |
| CN112084878A (en) * | 2020-08-14 | 2020-12-15 | 浙江工业大学 | Method for judging attitude standard degree of operator |
| CN112364694A (en) * | 2020-10-13 | 2021-02-12 | 宁波大学 | Human body sitting posture identification method based on key point detection |
-
2021
- 2021-08-18 CN CN202110950268.1A patent/CN113743255A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104916093A (en) * | 2015-06-05 | 2015-09-16 | 重庆智韬信息技术中心 | Method for realizing intelligent correction prompting of students' sitting postures |
| CN111091046A (en) * | 2019-10-28 | 2020-05-01 | 北京灵鹦科技有限公司 | User bad sitting posture correction system based on machine vision |
| CN111178280A (en) * | 2019-12-31 | 2020-05-19 | 北京儒博科技有限公司 | Human body sitting posture identification method, device, equipment and storage medium |
| CN112084878A (en) * | 2020-08-14 | 2020-12-15 | 浙江工业大学 | Method for judging attitude standard degree of operator |
| CN112364694A (en) * | 2020-10-13 | 2021-02-12 | 宁波大学 | Human body sitting posture identification method based on key point detection |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494441A (en) * | 2022-04-01 | 2022-05-13 | 广东机电职业技术学院 | Grape and picking point synchronous identification and positioning method and device based on deep learning |
| CN114494441B (en) * | 2022-04-01 | 2022-06-17 | 广东机电职业技术学院 | Grape and picking point synchronous identification and positioning method and device based on deep learning |
| CN114973423A (en) * | 2022-07-28 | 2022-08-30 | 聊城市飓风工业设计有限公司 | Warning method and system for sitting posture monitoring of child learning table |
| CN114973423B (en) * | 2022-07-28 | 2022-10-25 | 聊城市飓风工业设计有限公司 | Warning method and system for sitting posture monitoring of child learning table |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113743255A (en) | Neural network-based child sitting posture identification and correction method and system | |
| CN107292271B (en) | Learning monitoring method and device and electronic equipment | |
| CN103400476B (en) | A kind of sitting posture prompting method and device | |
| CN111079513B (en) | Posture reminding method and device, mobile terminal and storage medium | |
| CN112614399A (en) | Dance teaching equipment based on virtual reality and teaching method thereof | |
| CN108877885B (en) | Personal education fitness management system and method based on thermal imager and intelligent bracelet | |
| CN113297938A (en) | Sitting posture monitoring method and device, electronic equipment and storage medium | |
| KR100862352B1 (en) | Apparatus for automatic power control of video appliance and the method thereof | |
| CN111734974A (en) | Intelligent desk lamp with sitting posture reminding function | |
| CN111931608A (en) | Operation management method and system based on student posture and student face recognition | |
| CN102804204A (en) | Geture recognition using chroma- keying | |
| CN111402096A (en) | Online teaching quality management method, system, equipment and medium | |
| CN111047832A (en) | Mobile equipment with sitting posture adjusting function and using method thereof | |
| CN111967376A (en) | Pose identification and detection method based on neural network | |
| CN112017495A (en) | Information interaction method and system based on learning terminal | |
| CN107368808B (en) | Method and device for detecting reading state of child | |
| CN108416988A (en) | A kind of sitting posture prompting method and device | |
| CN116959189A (en) | Pet behavior correction method and pet behavior correction system based on camera device | |
| CN108766486B (en) | Control method and device and electronic equipment | |
| CN110933501A (en) | Child eye protection method for TV device, TV device with child eye protection function and system | |
| CN111310629A (en) | Operation process monitoring system | |
| CN214067935U (en) | Vision sitting posture identification correction reminding device | |
| CN111325273A (en) | Deep learning model establishing method and system based on user autonomous calibration | |
| CN112891165B (en) | Intelligent training system for realizing naked eye 3D amblyopia rehabilitation through intelligent handheld mobile terminal | |
| CN112972218B (en) | Naked eye 3D amblyopia rehabilitation intelligent training system |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211203 |