CN113903058A - Intelligent control system based on regional personnel identification - Google Patents
Intelligent control system based on regional personnel identification Download PDFInfo
- Publication number
- CN113903058A CN113903058A CN202111374022.0A CN202111374022A CN113903058A CN 113903058 A CN113903058 A CN 113903058A CN 202111374022 A CN202111374022 A CN 202111374022A CN 113903058 A CN113903058 A CN 113903058A
- Authority
- CN
- China
- Prior art keywords
- image
- control system
- signal
- system based
- unit
- 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
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Geometry (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an intelligent control system based on regional personnel identification, which comprises an image acquisition unit, an image analysis unit and a control unit, wherein the image acquisition unit is used for acquiring images; the image acquisition unit acquires signals and transmits the signals to the image analysis unit, and the image analysis unit transmits analysis results to the controlled equipment to transmit control instructions; the image analysis unit is an edge calculation processor which comprises a high-performance algorithm analysis module; the high-performance algorithm analysis module adopts a concurrent operation mode of a production line and a data transmission mode of a shared memory, so that the flow from transmission to calculation analysis of the whole image can be completed in a shorter time, the response speed of the system is improved, the accuracy of collision judgment of the portrait and the working area is ensured by acquiring the images of all areas on the spot and analyzing the portrait in the image through the algorithm, the work of the controlled equipment is stopped accurately and timely, and the safety of man-machine interaction is fully ensured.
Description
Technical Field
The invention relates to the technical field of recognition only, in particular to an intelligent control system based on regional personnel recognition.
Background
At present of automation popularization, man-machine interaction is common in daily life, in factories where some automation devices are popularized, man-machine safety operation is a ring which cannot be ignored, man-machine safety needs an interaction system to make accurate and quick judgment so as to effectively guarantee personal safety of workers, but the existing man-machine interaction safety system is difficult to achieve the characteristics of high efficiency, wide applicability, accuracy and the like.
Disclosure of Invention
The invention aims to provide an intelligent control system which outputs a result in real time after high-performance calculation and analysis and can realize specific equipment control according to the result of real-time data so as to ensure the safety of human-computer interaction.
In order to achieve the purpose, the invention provides an intelligent control system based on regional personnel identification, which comprises an image acquisition unit, an image analysis unit and a control unit; the image acquisition unit acquires signals and transmits the signals to the image analysis unit, and the image analysis unit transmits analysis results to the controlled equipment to transmit control instructions;
the image analysis unit is an edge calculation processor which comprises a high-performance algorithm analysis module; the high-performance algorithm analysis module comprises a video frame acquisition module, an image preprocessing module, an image analysis module and a region entry judgment module;
the video frame acquisition module acquires a picture of a scene needing to be analyzed currently from the image acquisition unit;
the image preprocessing module stores, screens and integrates the acquired image data;
the image analysis module realizes tasks of image classification, segmentation and target detection;
the region entry judging module judges the behavior of people in the analyzed image and discriminates whether the people in the designated region enter or leave a region critical point; the method realizes the process analysis of the safety of the designated area or the specific task of the area.
Further, the image acquisition module acquires field data in real time through image acquisition equipment and transmits a numerical value to the image analysis unit; the image acquisition device comprises at least one camera.
Furthermore, the control unit is a circuit control device and comprises a relay, the relay is in signal connection with the edge calculation processor, the relay receives the processing information from the edge calculation processor, corresponding instruction information is transmitted to the controlled unit, and the controlled unit is controlled in real time.
Furthermore, the signal transmission mode of the relay comprises a single voltage intensity signal or a dry junction electric signal, and the signal acts on a heartbeat signal and a switch signal;
the heartbeat signal is used as a judgment basis for judging the normal operation of the equipment, and the edge calculation processor sends a signal to the relay in real time through the GPIO port under the condition that the equipment normally operates; when the equipment has abnormal operation, the heartbeat signal is disconnected, and the condition is indicated to the signal unit in time, so that the controlled equipment is prevented from being in an error state for a long time due to an error signal.
Further, the video frame acquisition module acquires a picture of a scene to be analyzed from the camera, and completes matching of the camera and the corresponding picture.
Furthermore, in the image analysis module, the image classification task realizes the discrimination and the distinguishing of certain categories of articles and realizes the function of replacing human vision judgment; the target detection task realizes position judgment and state judgment of dynamic or static personnel or articles in the image designated area, thereby realizing the functions of behavior recognition analysis and the like; the image segmentation task realizes morphological analysis and position judgment of a specific object.
Further, the region entering judging module judges whether the person is in the region or not based on a distinguishing person algorithm and mainly divides the region into a region division part and detects the position of the person.
Furthermore, the region division mode is divided by different pictures acquired by different installation modes of image acquisition equipment in an actual scene, and the installation modes comprise a suspended ceiling direct-view installation mode, a high-point wall surface overlook installation mode and a horizontal head-up installation mode; the method for defining the designated area comprises a straight line setting method which is used for setting a straight line type boundary; the rectangular frame setting mode is used for judging the entrance and exit of square, fan-shaped and irregular multi-deformation areas; and the irregular polygon setting mode is used for judging the progress of the irregular sector and irregular polygon areas.
Further, the personnel position detection mainly depends on a deep learning algorithm to realize human body designated part detection, and the deep learning algorithm comprises a YooloV 3, a YooloV 4, a YooloV 5, a YooloX and a fast _ RCNN target detection common algorithm; the designated parts of the human body comprise a human head, a human face, hands, feet and a human body detection area;
the implementation steps of the deep learning algorithm comprise image acquisition, image drying and selection, image annotation, image preprocessing, image analysis, model pruning, model parameter adjustment and model storage; the method comprises the steps of detecting a human body appointed part by using a specific deep learning model to obtain position information of the human body appointed part in an image, marking the position information in a rectangular frame mode, wherein the position information comprises xmin, ymin, xmax, ymax and the like, comparing the detection information of the specific position with preset region position information, and determining whether a person is in a region or not according to the relative relationship between the current appointed human body part and the set region by comparing the coordinate point size of the specific position with the coordinate point size of the specific position, the intersection area of the rectangular frame size of the specific position with the polygonal region, and the position relationship between the coordinate point and a straight line.
Furthermore, the personnel position detection also comprises a monocular distance measurement scheme which comprises the steps of camera intrinsic parameter calibration and camera distortion elimination, wherein the intrinsic parameter and the extrinsic parameter of the camera are obtained through a series of calibration and a small amount of data measurement, and the intrinsic parameter and the extrinsic parameter are applied to the distance measurement scheme of an actual scene to obtain the actual distance of a target; and determining the specific position information of the human body in the real scene by combining the ranging scheme and the personnel detection scheme.
Compared with the prior art, the invention has the advantages that: the high-performance algorithm analysis module adopts a concurrent operation mode of a production line and a data transmission mode of a shared memory, so that the flow from the transmission of the whole image to the calculation analysis can be completed in a shorter time, the response speed of the system is improved, and the safety operation of a human machine is ensured.
The learning algorithm of the high-performance algorithm analysis module is suitable for various types and has wider adaptability, and the learning algorithm can be applied to various intelligent factories and meets the requirements of universality and applicability of real-time scene change.
The intelligent control system of the invention relies on various video image acquisition and algorithm to analyze the portrait in the image, the conflict between the portrait and the working area is judged accurately, and then the controlled equipment is stopped accurately and timely, the controlled equipment can be a factory mechanical arm, an automatic door opening and closing and the like.
The learning algorithm of the high-performance algorithm analysis module is suitable for various types and has wider adaptability, and the learning algorithm can be applied to various intelligent factories and meets the requirements of universality and applicability of real-time scene change.
The intelligent control system provided by the invention relies on various video image acquisition and algorithm to analyze the portrait in the image, the judgment of the conflict between the portrait and the working area is accurate, and further the controlled equipment can be stopped accurately and timely, the controlled equipment can be a factory mechanical arm, a door can be automatically opened and closed, and the like, so that the safety of human-computer interaction is sufficiently guaranteed. .
Drawings
FIG. 1 is a flow chart of an intelligent control system based on regional personnel identification according to the present invention;
FIG. 2 is a schematic diagram of distance measurement in an image and an actual scene according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be further described below.
As shown in fig. 1, the present invention provides an intelligent control system based on regional personnel identification, which includes an image acquisition unit, an image analysis unit and a control unit; the image acquisition unit acquires signals and transmits the signals to the image analysis unit, and the image analysis unit transmits analysis results to the controlled equipment to transmit control instructions;
the image analysis unit is an edge calculation processor which comprises a high-performance algorithm analysis module; the high-performance algorithm analysis module comprises a video frame acquisition module, an image preprocessing module, an image analysis module and a region entry judgment module;
the control unit is a circuit control device and comprises a relay, the relay is in signal connection with the edge calculation processor, the relay receives processing information from the edge calculation processor, corresponding instruction information is transmitted to the controlled unit, and the controlled unit is controlled in real time. The signal transmission mode of the relay comprises a single voltage intensity signal or a dry junction electric signal, and the signal acts on a heartbeat signal and a switch signal; the heartbeat signal is used as a judgment basis for judging the normal operation of the equipment, and the edge calculation processor sends a signal to the relay in real time through the GPIO port under the condition that the equipment normally operates; when the equipment has abnormal operation, the heartbeat signal is disconnected, and the condition is indicated to the signal unit in time, so that the controlled equipment is prevented from being in an error state for a long time due to an error signal.
The image acquisition module acquires field data in real time through image acquisition equipment and transmits the numerical value to the image analysis unit; the image acquisition device comprises at least one camera.
In the image analysis unit, a video frame acquisition module acquires a picture of a scene to be analyzed from a camera and completes matching of the camera and the corresponding picture;
the image preprocessing module stores, screens and integrates the acquired image data;
the image analysis module realizes the tasks of image classification, segmentation and target detection; the image classification task realizes the discrimination and the distinguishing of certain categories of articles and realizes the function of replacing human vision judgment; the target detection task realizes position judgment and state judgment of dynamic or static personnel or articles in the image designated area, thereby realizing the functions of behavior recognition analysis and the like; the image segmentation task realizes morphological analysis and position judgment of a specific object.
The region entry judging module judges the behavior of people in the analyzed image, and the region entry judging module discriminates whether the people in the designated region enter or leave a region critical point based on a distinguishing person algorithm; the process analysis of the safety of the designated area or the specific task of the area is realized; the method mainly comprises a region division part and personnel position detection.
The region entering judging module judges whether the personnel is in the region or not based on a distinguishing personnel algorithm and mainly divides the region into a region division part and personnel position detection.
The region division mode is divided by different pictures acquired by different installation modes of image acquisition equipment in an actual scene, and the installation modes comprise a suspended ceiling direct-view installation mode, a high-point wall surface overlook installation mode and a horizontal head-up installation mode; the method for defining the designated area comprises a straight line setting method which is used for setting a straight line type boundary; the rectangular frame setting mode is used for judging the entrance and exit of square, fan-shaped and irregular multi-deformation areas; and the irregular polygon setting mode is used for judging the progress of the irregular sector and irregular polygon areas.
The personnel position detection mainly depends on a deep learning algorithm to realize the detection of the designated part of the human body, and the deep learning algorithm comprises a common target detection algorithm of YoloV3, YoloV4, YoloV5, YoloX and Faster _ RCNN; the designated parts of the human body comprise a human head, a human face, hands, feet and a human body detection area;
the implementation steps of the deep learning algorithm comprise image acquisition, image drying and selection, image annotation, image preprocessing, image analysis, model pruning, model parameter adjustment and model storage; the method comprises the steps of detecting a human body appointed part by using a specific deep learning model to obtain position information of the human body appointed part in an image, marking the position information in a rectangular frame mode, wherein the position information comprises xmin, ymin, xmax, ymax and the like, comparing the detection information of the specific position with preset region position information, and determining whether a person is in a region or not according to the relative relationship between the current appointed human body part and the set region by comparing the coordinate point size of the specific position with the coordinate point size of the specific position, the intersection area of the rectangular frame size of the specific position with the polygonal region, and the position relationship between the coordinate point and a straight line.
The personnel position detection also comprises a monocular distance measurement scheme which comprises the steps of camera intrinsic parameter calibration and camera distortion elimination, wherein the intrinsic parameter and the extrinsic parameter of the camera are obtained through a series of calibration and a small amount of data measurement, and the intrinsic parameter and the extrinsic parameter are applied to the distance measurement scheme of an actual scene to obtain the actual distance of a target; and determining the specific position information of the human body in the real scene by combining the ranging scheme and the personnel detection scheme.
Example 1:
the method is mainly used for preventing casualty caused by the fact that people appear in the peripheral area of mechanical arm movement in the operation process of the mechanical arm.
The selected cameras in the scheme are all network cameras, the installation modes of the two cameras are wall high-point overlooking installation, the resolution ratio of the cameras is set to be 640x480, and the frame rate of the cameras for acquiring pictures is 25 FPS.
According to the scheme, the picture of the field area acquired by the camera is larger than or equal to the motion area of the field mechanical arm, and the picture acquired by the camera is limited by adopting a rectangular frame mode under the condition that the picture is larger than a preset area, so that a dangerous area and a safe area are further defined, and invalid signals are avoided.
In the scheme, a deep learning detection algorithm is adopted to realize the position detection of the personnel, a specific logic judgment principle is adopted to realize the position relation between the personnel position and the formulated area, and whether the current personnel have potential danger or not and whether a division alarm signal is needed or not is confirmed.
The deep learning algorithm adopted in the scheme is YooloX, the size of the image input by the network is 640x640, and the detection effect of the remote target is enhanced by a specific augmentation mode of the original training data set aiming at the situation of the field environment.
According to the scheme, the head of a person is selected as a detection target, whether the head of the person wears a safety helmet or a formulated hat is subjected to detail processing, and the head detection efficiency of different head accessories is optimized;
in the scheme, in view of the fact that the installation mode of the camera and the actual size of the field area are possibly larger than the preset condition, a certain angle deviation exists in the scheme of limiting the picture of the camera by adopting the rectangular frame, so that the distance between the human face and the camera needs to be increased as a secondary judgment condition, and the original training data labeling needs to be the labeling standard of the human face frame.
According to the scheme, a deep learning face detection model is used for acquiring face positions and label information (xmin, ymin, xmax, ymax and label), the height of a face on an image, namely height-ymax-ymin, can be acquired by using the face position information, standard length information of a target is acquired according to the face label information, a camera is used for shooting a face conversion function of a picture and an actual scene, the actual face length is converted into the standard length of the image, the standard length is compared with the measured height, and the finally measured actual distance is acquired by using the conversion function of the face proportion and the actual distance.
The principle of measuring and calculating the distance between the image and the actual scene in the scheme is shown in fig. 2, and essentially belongs to the principle of pinhole imaging, and the conversion formula of the height of the face detection frame on the image and the distance between the face and the camera in the actual scene can be realized by utilizing the property of proportion of three sides of the similar triangle. The formula is as follows:
distance: the actual distance between the face and the camera;
focus: calibrating parameters of the camera, wherein the parameters need to be calibrated according to the type of the camera, the parameter setting of the camera and the installation mode of the camera;
height _ of _ standard: the standard height of the human face, the parameter is determined according to the change of the actual scene, and the actual situation of the field personnel is known; height _ of _ detection: the height of the face detection frame.
According to the scheme, the comprehensive camera mounting mode adopts a specific area limiting mode of a rectangular frame and a scheme of calculating the distance between the face and the camera, so that the logic judgment on whether a person in a designated area enters or not is realized, different distance settings are adopted for different positions in an acquired image in an actual scene, and a detailed customized area is realized.
Personnel status determination
The personnel state judgment mainly comprises human body position detection, personnel detail detection and relevant video frame information comparison, wherein the personnel state judgment comprises but is not limited to human body motion state analysis, human body posture analysis and the like. The human motion state analysis is mainly used for tracking and judging the main action track of the human so as to prejudge the subsequent motion of the human; the human body posture analysis focuses on actual actions of a part of a human body, including but not limited to hands, legs and the like, wherein the hand posture analysis can judge real-time behaviors of the human body, such as civilized behaviors, criminal behaviors or specific gesture signals.
The human body position detection is mainly realized by depending on a monocular distance measurement principle and a deep learning algorithm, the shape of the human body is detected by the deep learning algorithm including but not limited to target detection algorithms such as YoloV3, YoloV4, YoloV5, YoloX, fast _ RCNN and the like, and then the position information of the person is obtained by a monocular distance measurement method. The implementation steps of the deep learning algorithm include, but are not limited to, image acquisition, image screening, image annotation, image preprocessing, image analysis, model pruning, model parameter adjustment, model storage and the like. The detection of the human body position is realized by using a deep learning model suitable for the current scene, and position information is usually marked by using a rectangular frame mode, wherein the position information comprises information such as (xmin, ymin, xmax, ymax) and the like. By using the determined human body position information, human body detail information can be further detected by using a deep learning model. .
The related video frame information comparison mainly depends on continuous acquisition of real-time data by the image acquisition module and matching of time sequence information, the data analysis module and the logic judgment module are used for acquiring the processing results of front and rear frames of the current frame by combining the time sequence information, integrating and analyzing the action or behavior of the formulated part of the human body, and optionally, the control unit is used for outputting formulated signals according to the analysis results to realize the control of corresponding equipment.
Signal control module
Heartbeat output
The intelligent control based on the regional personnel identification realizes the real-time control of the regional equipment depending on the real-time result of the regional personnel identification, and the control comprises but is not limited to door opening, counting, reminding, warning, shutdown and other effects. In order to ensure real-time effectiveness in the partial scenes, real-time operability of the region personnel identification algorithm must be ensured, and the control unit needs to acquire a specific signal in real time to ensure the correct operation state of the personnel identification algorithm, so that the effect can be realized through the heartbeat signal.
When the emergency state of the equipment in a specific area is adjusted by aiming at certain control signals, the area identification algorithm determines that the current area state does not need to be changed under the normal condition, a complex signal does not need to be sent to the control system in real time, only a heartbeat signal which is simple with the control system needs to be continuously sent, the normal operation of the area personnel identification algorithm is explained to the control system, the state of the area equipment does not need to be changed, and meanwhile, the condition that the state of the area equipment is unchanged all the time when the area personnel identification algorithm is abnormal is effectively avoided, so that the abnormal condition is generated.
The output of the heartbeat depends on the GPIO port of the edge computing device to output high and low levels.
Switching signal output
The intelligent control system based on the regional personnel identification finally acts on the specific equipment to realize the on-off or state conversion of the specific equipment, and the functions are usually realized by utilizing a switch signal.
The type of switching signal depends on the requirements of the actual controlled device, including but not limited to dry contact signal, high and low level signal, specified voltage signal. Wherein, the dry contact signal and the high-low level signal can be directly output to the control unit through the GPIO port of the edge device; the specified voltage signal needs to be converted by a specific voltage relay.
Web service module
Parameter configuration
The parameter configuration service is divided into system parameter configuration and business parameter configuration,
the system parameter configuration mainly aims at the relevant parameter setting of the edge terminal equipment, the state of the edge terminal equipment in current operation can be effectively checked through a system parameter configuration interface, and the opening and closing of different edge terminal equipment or abnormal equipment troubleshooting and the like are realized; meanwhile, aiming at multiple standardized algorithms configured by single edge terminal equipment, the system parameter configuration interface can be set by different algorithms in time, and actual application scenes of different edge terminal equipment can be customized in detail. For the image acquisition unit, the designated camera equipment can realize the configuration of camera standard parameters, including but not limited to focal length setting, image resolution setting, image acquisition frame rate setting and the like, and the camera parameter setting can be effectively realized depending on a system parameter configuration interface.
The service parameter configuration interface is mainly applied to the configuration of parameters required in the operation process of different algorithms, and includes but is not limited to relevant settings such as model selection, algorithm input image size, camera calibration parameter file configuration, result storage files and the like.
Image calibration
The image calibration service is mainly applied to the application of a person identification algorithm in a designated area, the person identification algorithm in the designated area depends on the acquisition of a camera of an image acquisition unit to a picture in the designated area, and the equipment control or the specific behavior analysis of the designated area is realized through an image analysis unit, a logic judgment unit and a control unit. For the area condition obtained by the camera, the full picture obtained by the camera in the actual application scene is not necessarily the specified area, so that the picture obtained by the camera needs to be limited by the specified area, and at this time, the image calibration service can be adopted to realize the limitation of the specified area.
Aiming at a specific personnel intrusion recognition algorithm, the installation mode of a camera and a limiting scheme of a site specific area are integrated, detailed customization needs to be carried out aiming at different image area distances so as to realize customized management aiming at different areas, and distance setting can be carried out in image calibration service at the moment.
Results display
The result display service mainly performs display of different algorithm results and display of real-time pictures of the image acquisition unit.
The above description is only a preferred embodiment of the present invention, and does not limit the present invention in any way. It will be understood by those skilled in the art that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. An intelligent control system based on regional personnel identification is characterized by comprising an image acquisition unit, an image analysis unit and a control unit; the image acquisition unit acquires signals and transmits the signals to the image analysis unit, and the image analysis unit transmits analysis results to the controlled equipment to transmit control instructions;
the image analysis unit is an edge calculation processor which comprises a high-performance algorithm analysis module; the high-performance algorithm analysis module comprises a video frame acquisition module, an image preprocessing module, an image analysis module and a region entry judgment module;
the video frame acquisition module acquires a picture of a scene needing to be analyzed currently from the image acquisition unit;
the image preprocessing module stores, screens and integrates the acquired image data;
the image analysis module realizes tasks of image classification, segmentation and target detection;
the region entry judging module judges the behavior of people in the analyzed image and discriminates whether the people in the designated region enter or leave a region critical point; the method realizes the process analysis of the safety of the designated area or the specific task of the area.
2. The intelligent control system based on regional personnel identification according to claim 1, wherein the image acquisition module acquires field data in real time through an image acquisition device and transmits the values to the image analysis unit; the image acquisition device comprises at least one camera.
3. The intelligent control system based on regional personnel identification of claim 1, wherein the control unit is a circuit control device and comprises a relay, the relay is in signal connection with the edge computing processor, the relay receives the processing information from the edge computing processor, transmits corresponding instruction information to the controlled unit, and controls the controlled unit in real time.
4. The intelligent control system based on regional personnel identification is characterized in that the signal transmission mode of the relay comprises a single voltage intensity signal or a dry junction electric signal, and the signal is acted on a heartbeat signal and a switch signal;
the heartbeat signal is used as a judgment basis for judging the normal operation of the equipment, and the edge calculation processor sends a signal to the relay in real time through the GPIO port under the condition that the equipment normally operates; when the equipment has abnormal operation, the heartbeat signal is disconnected, and the condition is indicated to the signal unit in time, so that the controlled equipment is prevented from being in an error state for a long time due to an error signal.
5. The intelligent control system based on regional personnel identification of claim 1, wherein the video frame acquisition module acquires a picture of a scene to be analyzed from a camera and completes matching of the camera and its corresponding picture.
6. The intelligent control system based on regional personnel identification according to claim 1, wherein in the image analysis module, the image classification task realizes discrimination and differentiation of certain categories of articles and realizes a function of replacing human vision judgment; the target detection task realizes position judgment and state judgment of dynamic or static personnel or articles in the image designated area, thereby realizing the functions of behavior recognition analysis and the like; the image segmentation task realizes morphological analysis and position judgment of a specific object.
7. The intelligent control system based on regional personnel identification of claim 1, wherein the regional entry decision module is based on a distinguishing personnel algorithm to decide whether personnel are in a region or not, and is mainly divided into a regional division part and personnel position detection.
8. The intelligent control system based on regional personnel identification as claimed in claim 7, wherein the region division mode is divided by different pictures collected by different installation modes of the image collection device in the actual scene, and the installation modes comprise a suspended ceiling direct-view installation mode, a high-point wall surface top-view installation mode and a horizontal head-up installation mode; the method for defining the designated area comprises a straight line setting method which is used for setting a straight line type boundary; the rectangular frame setting mode is used for judging the entrance and exit of square, fan-shaped and irregular multi-deformation areas; and the irregular polygon setting mode is used for judging the progress of the irregular sector and irregular polygon areas.
9. The intelligent control system based on regional personnel identification as claimed in claim 7, wherein the personnel position detection mainly relies on deep learning algorithm to realize human body designated part detection, the deep learning algorithm comprises yoolov 3, yoolov 4, yoolov 5, yoolox and fast _ RCNN target detection common algorithm; the designated parts of the human body comprise a human head, a human face, hands, feet and a human body detection area;
the implementation steps of the deep learning algorithm comprise image acquisition, image drying and selection, image annotation, image preprocessing, image analysis, model pruning, model parameter adjustment and model storage; the method comprises the steps of detecting a human body appointed part by using a specific deep learning model to obtain position information of the human body appointed part in an image, marking the position information in a rectangular frame mode, wherein the position information comprises xmin, ymin, xmax, ymax and the like, comparing the detection information of the specific position with preset region position information, and determining whether a person is in a region or not according to the relative relationship between the current appointed human body part and the set region by comparing the coordinate point size of the specific position with the coordinate point size of the specific position, the intersection area of the rectangular frame size of the specific position with the polygonal region, and the position relationship between the coordinate point and a straight line.
10. The intelligent control system based on regional personnel identification according to claim 9, wherein the personnel position detection further comprises a monocular distance measurement scheme comprising a camera intrinsic parameter calibration and a camera distortion removal step, wherein the camera intrinsic parameter calibration and the camera distortion removal step are used for obtaining the camera intrinsic and extrinsic parameters through a series of calibration and a small amount of data measurement, and the camera intrinsic and extrinsic parameters are applied to the distance measurement scheme of the actual scene to obtain the actual distance of the target; and determining the specific position information of the human body in the real scene by combining the ranging scheme and the personnel detection scheme.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111374022.0A CN113903058A (en) | 2021-11-19 | 2021-11-19 | Intelligent control system based on regional personnel identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111374022.0A CN113903058A (en) | 2021-11-19 | 2021-11-19 | Intelligent control system based on regional personnel identification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113903058A true CN113903058A (en) | 2022-01-07 |
Family
ID=79194709
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111374022.0A Pending CN113903058A (en) | 2021-11-19 | 2021-11-19 | Intelligent control system based on regional personnel identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113903058A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114581769A (en) * | 2022-01-19 | 2022-06-03 | 西南交通大学 | A method for identifying houses under construction based on unsupervised clustering |
| CN115083050A (en) * | 2022-06-09 | 2022-09-20 | 国网冀北综合能源服务有限公司 | Multi-condition judgment system and method for personnel existence in specific space |
| CN115953741A (en) * | 2023-03-14 | 2023-04-11 | 江苏实点实分网络科技有限公司 | Edge computing system and method based on embedded algorithm |
| CN116403284A (en) * | 2023-04-07 | 2023-07-07 | 北京奥康达体育产业股份有限公司 | Wisdom running examination training system based on bluetooth transmission technology |
| CN116862980A (en) * | 2023-06-12 | 2023-10-10 | 上海玉贲智能科技有限公司 | Target detection frame position optimization correction method, system, medium and terminal for image edge |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103618297A (en) * | 2013-07-17 | 2014-03-05 | 国家电网公司 | High-efficiency and high-reliability fault processing method aiming at complex power distribution network |
| CN111038258A (en) * | 2019-11-19 | 2020-04-21 | 上海赫千电子科技有限公司 | Display control method for digital instrument for vehicle |
| CN111341068A (en) * | 2020-03-02 | 2020-06-26 | 北京四利通控制技术股份有限公司 | Drilling site dangerous area early warning system and method based on deep learning |
| CN111860282A (en) * | 2020-07-15 | 2020-10-30 | 中国电子科技集团公司第三十八研究所 | Method and system for passenger flow statistics of subway section and pedestrian retrograde detection |
| CN112232211A (en) * | 2020-10-16 | 2021-01-15 | 山东科技大学 | An intelligent video surveillance system based on deep learning |
| CN112635045A (en) * | 2020-12-26 | 2021-04-09 | 西安科锐盛创新科技有限公司 | Intelligent monitoring system |
-
2021
- 2021-11-19 CN CN202111374022.0A patent/CN113903058A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103618297A (en) * | 2013-07-17 | 2014-03-05 | 国家电网公司 | High-efficiency and high-reliability fault processing method aiming at complex power distribution network |
| CN111038258A (en) * | 2019-11-19 | 2020-04-21 | 上海赫千电子科技有限公司 | Display control method for digital instrument for vehicle |
| CN111341068A (en) * | 2020-03-02 | 2020-06-26 | 北京四利通控制技术股份有限公司 | Drilling site dangerous area early warning system and method based on deep learning |
| CN111860282A (en) * | 2020-07-15 | 2020-10-30 | 中国电子科技集团公司第三十八研究所 | Method and system for passenger flow statistics of subway section and pedestrian retrograde detection |
| CN112232211A (en) * | 2020-10-16 | 2021-01-15 | 山东科技大学 | An intelligent video surveillance system based on deep learning |
| CN112635045A (en) * | 2020-12-26 | 2021-04-09 | 西安科锐盛创新科技有限公司 | Intelligent monitoring system |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114581769A (en) * | 2022-01-19 | 2022-06-03 | 西南交通大学 | A method for identifying houses under construction based on unsupervised clustering |
| CN115083050A (en) * | 2022-06-09 | 2022-09-20 | 国网冀北综合能源服务有限公司 | Multi-condition judgment system and method for personnel existence in specific space |
| CN115953741A (en) * | 2023-03-14 | 2023-04-11 | 江苏实点实分网络科技有限公司 | Edge computing system and method based on embedded algorithm |
| CN116403284A (en) * | 2023-04-07 | 2023-07-07 | 北京奥康达体育产业股份有限公司 | Wisdom running examination training system based on bluetooth transmission technology |
| CN116403284B (en) * | 2023-04-07 | 2023-09-12 | 北京奥康达体育产业股份有限公司 | Wisdom running examination training system based on bluetooth transmission technology |
| CN116862980A (en) * | 2023-06-12 | 2023-10-10 | 上海玉贲智能科技有限公司 | Target detection frame position optimization correction method, system, medium and terminal for image edge |
| CN116862980B (en) * | 2023-06-12 | 2024-01-23 | 上海玉贲智能科技有限公司 | Target detection frame position optimization correction method, system, medium and terminal for image edge |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113903058A (en) | Intelligent control system based on regional personnel identification | |
| CN110188724B (en) | Method and system for helmet positioning and color recognition based on deep learning | |
| CN101465033B (en) | Automatic tracking recognition system and method | |
| US8036425B2 (en) | Neural network-controlled automatic tracking and recognizing system and method | |
| US8432450B2 (en) | Surveillance system having status detection module, method for self-monitoring of an observer, and computer program | |
| CN201278180Y (en) | Automatic tracking recognition system | |
| CN107133564B (en) | Tooling cap detection method | |
| CN101635834A (en) | Automatic tracking and identifying system controlled by nerve | |
| CN102164270A (en) | Intelligent video monitoring method and system capable of exploring abnormal events | |
| CN112084813B (en) | Abnormal target detection method, device and storage medium | |
| CN103929592A (en) | All-dimensional intelligent monitoring equipment and method | |
| CN112396658A (en) | Indoor personnel positioning method and positioning system based on video | |
| US10183843B2 (en) | Monitoring of step rollers and maintenance mechanics of passenger conveyors | |
| CN112235537A (en) | A kind of safety early warning method for on-site operation of substation | |
| RU83675U1 (en) | VIDEO MONITORING SYSTEM | |
| CN115527158A (en) | Method and device for detecting abnormal behaviors of personnel based on video monitoring | |
| CN107295308A (en) | A kind of supervising device based on computer room recognition of face and alarm | |
| CN112532927A (en) | Intelligent safety management and control system for construction site | |
| CN119339354A (en) | Intelligent obstacle detection system for driverless subway based on multimodal AI sensors | |
| CN118258399A (en) | A method for indoor human detection based on multi-sensor fusion | |
| CN113314230A (en) | Intelligent epidemic prevention method, device, equipment and storage medium based on big data | |
| CN120047907B (en) | Urban rail transit emergency treatment method and system based on visual large model | |
| US20250244741A1 (en) | Real-Time Collision Detection and Illuminated Guidance System Using Computer Vision | |
| CN113239772A (en) | Personnel gathering early warning method and system in self-service bank or ATM environment | |
| JPH0460880A (en) | Moving body discrimination and analysis controlling 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: 20220107 |