CN113538873A - AI position of sitting corrects camera based on image recognition technology - Google Patents
AI position of sitting corrects camera based on image recognition technology Download PDFInfo
- Publication number
- CN113538873A CN113538873A CN202110860054.5A CN202110860054A CN113538873A CN 113538873 A CN113538873 A CN 113538873A CN 202110860054 A CN202110860054 A CN 202110860054A CN 113538873 A CN113538873 A CN 113538873A
- Authority
- CN
- China
- Prior art keywords
- module
- image
- camera
- frame
- sitting posture
- 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
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/18—Status alarms
- G08B21/24—Reminder alarms, e.g. anti-loss alarms
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/80—Camera processing pipelines; Components thereof
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Abstract
The invention discloses an AI sitting posture correction camera based on an image recognition technology, which comprises a camera module, an AI recognition module, a communication module and an outgoing loudspeaker module, wherein: a camera module: for acquiring image data; an AI identification module: a related algorithm for sitting posture identification is built in, and images collected by the camera module are predicted in real time; a communication module: the system consists of an Ethernet, a 4G communication module or a 5G communication module and provides network transmission service for an upper layer; the external transmission horn module: when the AI identification module identifies an incorrect sitting posture, relevant voice prompts are played and corrected in time. The invention can identify various bad sitting postures, more comprehensively help users to correct various bad sitting postures, is suitable for different crowds, has wide coverage, can keep images to be checked on a client, is convenient for users to know own postures, and is beneficial to the situation of unmanned supervision.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an AI sitting posture correction camera based on an image recognition technology.
Background
The current sitting posture correcting products in the market mainly take physical products such as a sitting posture corrector and a sitting posture correcting brace, need to be manually adjusted and set or worn, have a certain auxiliary effect, and prevent the problems of humpback, myopia and the like.
The prior art has the following defects:
the body structures of people have certain differences, and the correction equipment and the wearable equipment can only correct partial people and specific partial postures and cannot completely cover people of different ages and different body shapes.
The service life of the corrector and the wearing equipment is limited and the corrector and the wearing equipment are easy to damage.
And the physical structures of the corrector and the wearable device may have potential safety hazards.
Disclosure of Invention
The invention aims to provide an AI sitting posture correction camera based on an image recognition technology, which can solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme:
the utility model provides a camera is corrected to AI position of sitting based on image recognition technology, includes camera module, AI identification module, communication module, biography loudspeaker module outward, wherein:
a camera module: for acquiring image data;
an AI identification module: a related algorithm for sitting posture identification is built in, and images collected by the camera module are predicted in real time;
a communication module: the system consists of an Ethernet, a 4G communication module or a 5G communication module and provides network transmission service for an upper layer;
the external transmission horn module: when the AI identification module identifies an incorrect sitting posture, relevant voice prompts are played and corrected in time.
Further, the workflow of the AI sitting posture correction camera is as follows: can initialize camera, communication after the circular telegram, pass outward loudspeaker module, the camera is the image now in real time, and the image input that will gather carries out real-time prediction to AI identification module, when discerning the position of sitting incorrect, calls the relevant suggestion of passing outward loudspeaker module broadcast to and the just position of sitting of just holding in time, call communication module simultaneously and encrypt the back with the not just picture of position of sitting and upload the high in the clouds, convenient statistics and user look over.
Further, the AI identification module comprises a plug-in module and an image identification module, wherein the plug-in module is used for coding image frame data collected by the camera and then plug-in to a network, and the image identification module is used for loading a related neural network model to identify the image frames in real time.
Further, the overall flow of the AI identification module is as follows: the camera module captures two image frames with different resolutions, the image frame of the high-resolution video frame channel is used for video plug flow, the low-resolution video frame channel is used for identifying the image frame identification module, the image frame of the high-resolution video frame channel is firstly subjected to geometric deformation correction by the geometric distortion correction module and then enters the 2D image operation module for inversion, then the 2D image interface module integrates the result identified by the image frame identification module, the image frame is zoomed by the 2D image operation module and is stored in the queue for waiting for the encoding of the encoding module, the encoding module encodes the video frame into H264/H265 code stream and finally pushes the encoded data to the network, the image frame of the low-resolution video frame channel is firstly zoomed or inverted by the 2D image operation module and then is stored in the queue and then is input to the image frame identification module for identification, and inputting the result identified by the image frame identification module into the image drawing module for drawing, and calling the NPU at the bottom layer to identify the input image frame with the specified format after the specified neural network model is loaded by the image frame identification module.
Further, the flow of the plug-flow module is as follows:
after a sensor of a camera captures a video frame, inputting the image frame into a geometric distortion correction module for geometric distortion correction, inputting the corrected data frame into a 2D image operation module for rotation, then inputting the data frame into a 2D image interface module for operations such as drawing, font rendering and the like, then inputting the data frame of the previous step into a 2D image operation module for scaling, storing the scaled data frame into a queue, waiting for a coding module to code the data frame into an H264/H265 code stream, and finally pushing the code stream into a network by a mainstream spanning tree protocol.
Further, the flow of the image recognition module is as follows:
video frames captured by the camera are input to the 2D image operation module to be stored in the queue after being rotated or zoomed, after the neural network model is loaded by the image frame identification module, image frame data are obtained from the queue to be detected in real time, after the result is detected, the image frame data are input to the image drawing module to be drawn, the drawing result is input to the 2D image interface module to be synthesized with the image frames output by the 2D image operation module.
Further, the 2D graphics operations module provides 2D graphics operations including rectangle filling, bitmap copying, image scaling, image blending.
Further, the 2D graphics interface module includes drawing, filling, font rendering, and image loading.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention checks the human sitting posture in real time through an image recognition technology, and reminds and corrects the sitting posture by voice when the human body has bad sitting postures such as bending down, humpback and the like. Simple to use does not need any correction equipment of installation and dress correction equipment, only needs the camera to aim at the human body and can use, and the various bad position of sitting of distinguishable multiple position of sitting, more comprehensive help user correct various bad position of sitting, is applicable to the crowd of different ages, stature, posture, and it is wide to cover the crowd.
2. The camera module can keep images to be checked on the client side when detecting the bad sitting posture, so that a user can know the posture conveniently.
3. The invention can call the camera to check the video on the client, can remotely monitor, is beneficial to remotely checking under the condition of no supervision, and is convenient for parents to remotely check under the condition of no supervision of children.
4. The built-in algorithm model of the camera module can be independently learned, and the recognition result is continuously optimized.
Drawings
Fig. 1 is a schematic structural diagram of an AI sitting posture correction camera according to the present invention;
FIG. 2 is a schematic structural diagram of an AI identification module according to the present invention;
FIG. 3 is a flow chart of the AI identification module of the present invention;
FIG. 4 is a flow diagram of a plug flow module of the present invention;
FIG. 5 is a flow chart of the image recognition module of the present invention.
In the figure: 100. a camera module; 200. an AI identification module; 300. a communication module; 400. an outgoing horn module; 201. a plug flow module; 202. and an image identification module.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, an AI sitting posture correcting camera based on an image recognition technology includes a camera module 100, an AI recognition module 200, a communication module 300, and an outgoing speaker module 400, wherein: the camera module 100: the system is mainly responsible for acquiring real-time images; the AI identification module 200: a related algorithm for sitting posture identification is built in, and images collected by the camera module 100 are predicted in real time; the communication module 300: the system consists of an Ethernet, a 4G communication module or a 5G communication module and provides network transmission service for an upper layer; the outgoing horn module 400: when the AI identification module 200 identifies an incorrect sitting posture, the associated voice prompt is played and the posture is corrected.
The working process of the AI sitting posture correction camera is as follows: can initialize the camera after the circular telegram, communication, biography loudspeaker module outward, the real-time present image of camera, the image input that will gather carries out real-time prediction to AI identification module 200, when discerning the position of sitting incorrect, calls and passes loudspeaker module 400 broadcast relevant suggestion outward to and the time correction position of sitting, call communication module 300 simultaneously and encrypt the back with the not just right picture of position of sitting and upload the high in the clouds, convenient statistics and user look over.
Referring to fig. 2, the AI identification module 200 includes a plug-in module 201 and an image identification module 202, where the plug-in module 201 is configured to encode and plug-in image frame data acquired by a camera to a network, and the image identification module 202 is configured to load a relevant neural network model to identify image frames in real time.
Referring to fig. 3, the overall flow of the AI identification module 200 is as follows: the camera module 100 captures two image frames with different resolutions, the image frame of the high resolution video frame channel is used for video plug flow, the low resolution video frame channel is used for image frame identification module to identify, the image frame of the high resolution video frame channel is firstly subjected to geometric deformation correction by the geometric distortion correction module, then enters the 2D image operation module to be inverted, then integrates the result identified by the image frame identification module in the 2D image interface module, is zoomed by the 2D image operation module and is stored in the queue to wait for the encoding of the encoding module, the encoding module encodes the video frame into H264/H265 code stream, and finally pushes the encoded data into the network, the image frame of the low resolution video frame channel is firstly zoomed or inverted by the 2D image operation module, then is stored in the queue and then is input into the image frame identification module to be identified, and inputting the result identified by the image frame identification module into the image drawing module for drawing, and calling the NPU at the bottom layer to identify the input image frame with the specified format after the specified neural network model is loaded by the image frame identification module.
Referring to fig. 4, the flow of the plug flow module 201 is as follows:
after a sensor of a camera captures a video frame, inputting the image frame into a geometric distortion correction module for geometric distortion correction, inputting the corrected data frame into a 2D image operation module for rotation, then inputting the data frame into a 2D image interface module for operations such as drawing, font rendering and the like, then inputting the data frame of the previous step into a 2D image operation module for scaling, storing the scaled data frame into a queue, waiting for a coding module to code the data frame into an H264/H265 code stream, and finally pushing the code stream into a network by a mainstream spanning tree protocol.
Referring to fig. 5, the flow of the image recognition module 202 is as follows:
video frames captured by the camera are input to the 2D image operation module to be stored in the queue after being rotated or zoomed, after the neural network model is loaded by the image frame identification module, image frame data are obtained from the queue to be detected in real time, after the result is detected, the image frame data are input to the image drawing module to be drawn, the drawing result is input to the 2D image interface module to be synthesized with the image frames output by the 2D image operation module.
The high resolution video frame channel and the low resolution video frame channel are video frame channels with different resolutions captured by the camera sensor. The geometric distortion correction module performs geometric distortion correction on the input image.
The 2D graphics operations module provides 2D graphics operations including rectangle filling, bitmap copying, image scaling, image blending. The 2D graphics interface module includes drawing, filling, font rendering, and image loading.
The AI identification module 200 handles the image of gathering the camera and discerns, when discerning bad position of sitting, then calls through the loudspeaker module 400 of biography outward and put loudspeaker broadcast pronunciation and remind the correction position of sitting to transmit image and result to high in the clouds through communication module 300 and preserve. The communication module 300 is connected to the local area network in a wired or wireless manner, and transmits the audio and video data to the cloud for storage, so that a user can check the data on the client.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be able to cover the technical solutions and the inventive concepts of the present invention within the technical scope of the present invention.
Claims (8)
1. The AI sitting posture correction camera based on the image recognition technology is characterized by comprising a camera module (100), an AI recognition module (200), a communication module (300) and an outgoing loudspeaker module (400), wherein:
camera module (100): for acquiring image data;
AI identification module (200): a related algorithm for sitting posture identification is built in, and images collected by the camera module (100) are predicted in real time;
communication module (300): the system consists of an Ethernet, a 4G communication module or a 5G communication module and provides network transmission service for an upper layer;
outgoing horn module (400): when the AI recognition module (200) recognizes an incorrect sitting posture, relevant voice prompts are played and corrected in time.
2. The AI sitting posture rectification camera based on the image recognition technology as claimed in claim 1, wherein the workflow of the AI sitting posture rectification camera is as follows: can initialize the camera after the circular telegram, communication, pass outward loudspeaker module, the real-time present image of camera, the image input that will gather carries out real-time prediction to AI identification module (200), when discerning the position of sitting incorrect, transfer and pass outward loudspeaker module (400) and broadcast relevant suggestion to and the time correction position of sitting, transfer communication module (300) simultaneously and encrypt the back with the not just picture of position of sitting and upload to the high in the clouds, convenient statistics and user look over.
3. The AI sitting posture correction camera based on the image recognition technology as claimed in claim 1, wherein the AI recognition module (200) comprises a stream pushing module (201) and an image recognition module (202), wherein the stream pushing module (201) is used for encoding and then pushing the image frame data collected by the camera to the network, and the image recognition module (202) is used for loading a relevant neural network model to perform real-time recognition on the image frame.
4. The AI sitting posture correction camera based on the image recognition technology as claimed in claim 1, wherein the overall flow of the AI recognition module (200) is as follows: the camera module (100) captures two image frames with different resolutions, the image frame of the high-resolution video frame channel is used for video plug flow, the low-resolution video frame channel is used for identifying the image frame identification module, the image frame of the high-resolution video frame channel is firstly subjected to geometric deformation correction by the geometric distortion correction module and then enters the 2D image operation module for inversion, then the 2D image interface module integrates the result identified by the image frame identification module, the image frame is zoomed by the 2D image operation module and stored in the queue for waiting for the encoding of the encoding module, the encoding module encodes the video frame into an H264/H265 code stream and finally pushes the encoded data to the network, the image frame of the low-resolution video frame channel is firstly zoomed or inverted by the 2D image operation module and then stored in the queue and then input to the image frame identification module for identification, and inputting the result identified by the image frame identification module into the image drawing module for drawing, and calling the NPU at the bottom layer to identify the input image frame with the specified format after the specified neural network model is loaded by the image frame identification module.
5. The AI sitting posture correction camera based on image recognition technology as claimed in claim 3, wherein the flow of the push flow module (201) is as follows:
after a sensor of a camera captures a video frame, inputting the image frame into a geometric distortion correction module for geometric distortion correction, inputting the corrected data frame into a 2D image operation module for rotation, then inputting the data frame into a 2D image interface module for operations such as drawing, font rendering and the like, then inputting the data frame of the previous step into a 2D image operation module for scaling, storing the scaled data frame into a queue, waiting for a coding module to code the data frame into an H264/H265 code stream, and finally pushing the code stream into a network by a mainstream spanning tree protocol.
6. The AI sitting posture correction camera based on image recognition technology as claimed in claim 3, wherein the flow of the image recognition module (202) is as follows:
video frames captured by the camera are input to the 2D image operation module to be stored in the queue after being rotated or zoomed, after the neural network model is loaded by the image frame identification module, image frame data are obtained from the queue to be detected in real time, after the result is detected, the image frame data are input to the image drawing module to be drawn, the drawing result is input to the 2D image interface module to be synthesized with the image frames output by the 2D image operation module.
7. The AI sitting posture rectification camera based on the image recognition technology as claimed in any one of claims 4-6, wherein the 2D graphics operation module provides 2D graphics operations including rectangle filling, bitmap copying, image scaling, image blending.
8. The AI sitting posture rectification camera based on the image recognition technology as claimed in any one of claims 4-6, wherein the 2D graphics interface module includes drawing, filling, font rendering and image loading.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110860054.5A CN113538873A (en) | 2021-07-28 | 2021-07-28 | AI position of sitting corrects camera based on image recognition technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110860054.5A CN113538873A (en) | 2021-07-28 | 2021-07-28 | AI position of sitting corrects camera based on image recognition technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113538873A true CN113538873A (en) | 2021-10-22 |
Family
ID=78089582
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110860054.5A Pending CN113538873A (en) | 2021-07-28 | 2021-07-28 | AI position of sitting corrects camera based on image recognition technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113538873A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116260990A (en) * | 2023-05-16 | 2023-06-13 | 合肥高斯智能科技有限公司 | AI asynchronous detection and real-time rendering method and system for multipath video streams |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109274883A (en) * | 2018-07-24 | 2019-01-25 | 广州虎牙信息科技有限公司 | Posture antidote, device, terminal and storage medium |
| CN110796077A (en) * | 2019-10-29 | 2020-02-14 | 湖北民族大学 | Attitude motion real-time detection and correction method |
| CN111601088A (en) * | 2020-05-27 | 2020-08-28 | 大连成者科技有限公司 | Sitting posture monitoring system based on monocular camera sitting posture identification technology |
| US20200394817A1 (en) * | 2019-06-17 | 2020-12-17 | Qisda Corporation | Lamp and method for detecting a sitting posture of a user |
| CN112149472A (en) * | 2019-06-28 | 2020-12-29 | 广州芊泓运动科技有限公司 | Artificial intelligence-based limb action recognition and comparison method |
-
2021
- 2021-07-28 CN CN202110860054.5A patent/CN113538873A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109274883A (en) * | 2018-07-24 | 2019-01-25 | 广州虎牙信息科技有限公司 | Posture antidote, device, terminal and storage medium |
| US20200394817A1 (en) * | 2019-06-17 | 2020-12-17 | Qisda Corporation | Lamp and method for detecting a sitting posture of a user |
| CN112149472A (en) * | 2019-06-28 | 2020-12-29 | 广州芊泓运动科技有限公司 | Artificial intelligence-based limb action recognition and comparison method |
| CN110796077A (en) * | 2019-10-29 | 2020-02-14 | 湖北民族大学 | Attitude motion real-time detection and correction method |
| CN111601088A (en) * | 2020-05-27 | 2020-08-28 | 大连成者科技有限公司 | Sitting posture monitoring system based on monocular camera sitting posture identification technology |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116260990A (en) * | 2023-05-16 | 2023-06-13 | 合肥高斯智能科技有限公司 | AI asynchronous detection and real-time rendering method and system for multipath video streams |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103460250B (en) | Image procossing based on object of interest | |
| WO2018077142A1 (en) | Panoramic video processing method, device and system | |
| CN105072327B (en) | A kind of method and apparatus of the portrait processing of anti-eye closing | |
| CN105979224A (en) | Head mount display, video output device and video processing method and system | |
| JP2016129416A (en) | Method for dynamically adapting video image parameter for facilitating subsequent application | |
| CN101930535A (en) | Human face detection and tracking device | |
| US9001181B2 (en) | Image processing apparatus, image processing method, and computer program product | |
| CN111696039B (en) | Image processing method and device, storage medium and electronic equipment | |
| CN102474593A (en) | Teleconferencing device and image display processing method | |
| CN102088593B (en) | MPEG4 (Moving Pictures Experts Group 4) compressed video transmission communication system and method based on Bluetooth 3.0 specification | |
| CN113538873A (en) | AI position of sitting corrects camera based on image recognition technology | |
| CN110291774A (en) | A kind of image processing method, equipment, system and storage medium | |
| CN104504852A (en) | Image-based forewarning infant safety monitor and infant safety judging method | |
| CN108076303A (en) | A kind of video image display method and device | |
| CN107197192B (en) | A kind of method and system for face video in compressed video communication | |
| CN106899810A (en) | A kind of mine video image fusion method and device | |
| CN110753243A (en) | Image processing method, image processing server and image processing system | |
| JP3550046B2 (en) | Camera image transmission method and camera image transmission system | |
| JPH0338188A (en) | Moving picture transmission system | |
| CN112887293A (en) | Streaming media processing method and device and electronic equipment | |
| JP5934993B2 (en) | Image transmission apparatus and image transmission / reception system | |
| JPH0879721A (en) | Video telephone system | |
| CN205987159U (en) | Live all -in -one of medical treatment video | |
| CN111491166A (en) | Dynamic compression system and method based on content analysis | |
| JP3197894B2 (en) | Image transmission device |
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 |
Application publication date: 20211022 |
|
| RJ01 | Rejection of invention patent application after publication |