CN111814587A

CN111814587A - Human behavior detection method, teacher behavior detection method, and related system and device

Info

Publication number: CN111814587A
Application number: CN202010561453.7A
Authority: CN
Inventors: 王原原
Original assignee: Zhejiang Dahua Technology Co Ltd
Current assignee: Zhejiang Dahua Technology Co Ltd
Priority date: 2020-06-18
Filing date: 2020-06-18
Publication date: 2020-10-23
Anticipated expiration: 2040-06-18
Also published as: CN111814587B

Abstract

The application discloses a human behavior detection method, a teacher behavior detection method, a relevant system and a relevant device, wherein the human behavior detection method comprises the following steps: the monitoring device acquires a monitoring image comprising a target human body and determines a target area in the monitoring image; positioning a target human body in the monitoring image, and determining the position information of the target human body; judging whether the target human body is in the target area or not based on the position information; if the target human body is in the target area, determining at least one characteristic information of the orientation and the action of the target human body based on the monitoring image; determining the action of the target human body by using preset key points of the target human body; and determining the current behavior of the target human body according to the at least one characteristic information and the action of the target human body. By means of the method, the target human body appearing in the monitoring image can be detected in real time, and the accuracy of behavior judgment on the target human body can be effectively improved.

Description

Human behavior detection method, teacher behavior detection method, and related system and device

Technical Field

The application relates to the technical field of human behavior detection, in particular to a human behavior detection method, a teacher behavior detection method, a relevant system and a relevant device.

Background

Nowadays, in the field of video monitoring, it is becoming more and more important to perform behavior detection on a human body appearing in a monitored image and to perform statistical analysis on the behavior of the human body. The current behavior of the target human body in the monitored image is judged and then classified, so that the behavior characteristics of the target human body, such as performance, teaching effect or motion posture, can be evaluated according to corresponding behavior statistics. For example, in schools, especially in primary schools, junior high schools and high schools, since teachers have a lot of activities on the platform, such as explaining to the Power Point (slide) or performing teaching work by writing on a board, in actual teaching, especially in teaching video recording and broadcasting, the learning status of students is affected by the back-to-back status of teachers and the posture of writing on a board, and at this time, the teaching supervision group cannot be positioned all around, so that the long-term and effective supervision on the behavior of teachers cannot be performed, and finally the evaluation on the teachers is affected.

In recent years, with the development of the internet, the application of automatically detecting the behavior of a target human body through a monitoring device and then performing statistical evaluation is also greatly developed. However, in the conventional video monitoring method, the behavior of the target human body appearing in the monitored area is rarely detected in real time to perform statistical analysis after judgment and classification, and effective and accurate detection of the human body behavior cannot be performed.

Disclosure of Invention

The technical problem mainly solved by the application is to provide a human behavior detection method, a teacher behavior detection method, a relevant system and a relevant device.

In order to solve the above technical problem, the first technical solution adopted by the present application is: provided is a human behavior detection method, wherein the human behavior detection method comprises the following steps: the monitoring device acquires a monitoring image comprising a target human body and determines a target area in the monitoring image; positioning a target human body in the monitoring image, and determining the position information of the target human body; judging whether the target human body is in the target area or not based on the position information; if the target human body is in the target area, determining at least one characteristic information of the orientation and the action of the target human body based on the monitoring image; determining the action of the target human body by using preset key points of the target human body; and determining the current behavior of the target human body according to the at least one characteristic information and the action of the target human body.

The step of determining at least one of the orientation and the motion of the target human body based on the monitored image comprises: and carrying out classification prediction on the orientation and/or action type of the target human body in the monitored image by utilizing a classification model of the deep learning network to obtain the orientation and/or action of the target human body.

The method for determining the action of the target human body by using the preset key points of the target human body comprises the following steps: and detecting the positions of preset key points of the target human body by using a deep learning network, and classifying the behavior types of the target human body according to the positions of the preset key points to obtain the action of the target human body.

The step of determining the current behavior of the target human body according to the at least one type of feature information and the motion of the target human body specifically includes: and determining the current behavior of the target human body according to the at least one type of characteristic information, and correcting the determined result of the current behavior of the target human body through the action of the target human body to obtain the classification result of the current behavior of the target human body.

Before the step of performing classification prediction on the orientation and/or action type of the target human body in the monitored image by using a classification model of the deep learning network to obtain the orientation and/or action of the target human body, the method further comprises the following steps: carrying out target human body detection on the target area to obtain a characteristic image comprising a target human body detection frame; expanding the area of the target human body detection frame in the characteristic image according to a preset proportional coefficient so as to intercept the target image in the expanded target human body detection frame; the method comprises the following steps of utilizing a classification model of a deep learning network to classify and predict the orientation and/or action type of a target human body in a monitored image, and obtaining the orientation and/or action of the target human body, wherein the classification model comprises the following steps: and carrying out classification prediction on the orientation and/or action type of the target human body in the target image by utilizing a classification model of the deep learning network to obtain the orientation and/or action of the target human body.

The steps of acquiring a monitoring image including a target human body by a monitoring device and determining a target area from the monitoring image specifically include: the monitoring device monitors a set monitoring area in real time to acquire a monitoring image including a target human body when the target human body appears in the set monitoring area; and defining a target area in the monitoring image according to the activity area of the target human body.

The steps of positioning a target human body in the monitored image and determining the position information of the target human body comprise: and positioning the target human body in the monitoring image by using the trained detection network model so as to determine the position information of the target human body.

After the step of determining the current behavior of the target human body according to the at least one type of feature information and the action of the target human body, the method further comprises the following steps: and respectively counting the occurrence frequency, single duration and total duration of each current behavior of the target human body within a set time, and displaying the counting result.

In order to solve the above technical problem, the second technical solution adopted by the present application is: provided is a teacher behavior detection method, wherein the teacher behavior detection method comprises the following steps: the method comprises the steps that a monitoring device obtains a monitoring image including a teacher and determines a platform area in the monitoring image; positioning the teacher in the monitoring image, and determining the position information of the teacher; judging whether the teacher is in the platform area or not based on the position information; if the teacher is in the platform area, determining whether the teacher faces the student and whether to write at least one feature information in the board writing based on the monitoring image; determining the actions of the teacher by using the human body preset key points of the teacher; and determining whether the teacher belongs to the writing board writing state currently according to the at least one characteristic information and the teacher action.

After the step of determining whether the teacher belongs to the writing board writing state currently according to the at least one type of feature information and the teacher's action, the method further comprises the following steps: and respectively counting the occurrence frequency, single duration and total duration of the writing board of the teacher facing to students, the non-writing board of the back facing students and the writing board of the back facing students in the set time, and displaying the counting result.

In order to solve the above technical problem, the third technical solution adopted by the present application is: the human behavior detection system comprises an intelligent terminal and a camera connected with the intelligent terminal; the camera is used for acquiring a monitoring image including a target human body; the intelligent terminal is used for receiving the monitoring image sent by the camera, determining a target area in the monitoring image, positioning a target human body in the monitoring image, determining position information of the target human body, judging whether the target human body is in the target area or not based on the position information, determining at least one feature information of orientation and action of the target human body based on the monitoring image when the target human body is determined to be in the target area, determining action of the target human body by utilizing preset key points of the target human body, and determining current action of the target human body according to the at least one feature information and the action of the target human body.

In order to solve the above technical problem, a fourth technical solution adopted by the present application is: the teacher behavior detection system comprises an intelligent terminal and a camera connected with the intelligent terminal; the camera is used for acquiring a monitoring image including a teacher; the intelligent terminal is used for receiving the monitoring image sent by the camera, determining a platform area in the monitoring image, positioning a teacher in the monitoring image, determining position information of the teacher, judging whether the teacher is in the platform area, determining whether the teacher faces the student or not and whether the teacher writes at least one type of feature information in a blackboard writing or not when determining that the teacher is in the platform area, determining actions of the teacher by using preset key points of a human body of the teacher, and determining whether the teacher belongs to a blackboard writing state or not according to the at least one type of feature information and the actions of the teacher.

In order to solve the above technical problem, a fifth technical solution adopted by the present application is: providing an intelligent terminal, wherein the intelligent terminal comprises a memory and a processor which are coupled with each other; the memory stores program data; the processor is configured to execute the program data to implement the detection method as described in any one of the above.

In order to solve the above technical problem, a sixth technical solution adopted in the present application is: there is provided a computer readable storage medium having stored thereon program data executable by a processor to implement a detection method as described in any one of the above.

The beneficial effect of this application is: different from the prior art, the human behavior detection method in the application acquires a monitoring image including a target human body through a monitoring device, determines a target area in the monitoring image, positions the target human body in the monitoring image, determines position information of the target human body, judges whether the target human body is in the target area based on the position information, determines at least one feature information of orientation and action of the target human body based on the monitoring image if the target human body is in the target area, determines action of the target human body by utilizing preset key points of the target human body, determines current action of the target human body according to the at least one feature information and the action of the target human body, can detect real-time action of the target human body appearing in the target area of the monitoring image, and combines two ways of action detection and judgment of the target human body, the accuracy of final behavior judgment is effectively improved.

Drawings

FIG. 1 is a schematic flow chart of a first embodiment of a human behavior detection method according to the present application;

FIG. 2 is a schematic flow chart of a human behavior detection method according to a second embodiment of the present application;

FIG. 3 is a schematic flow chart of a human behavior detection method according to a third embodiment of the present application;

FIG. 4 is a flowchart of a first embodiment of a teacher behavior detection method according to the present application;

FIG. 5 is a flowchart of a second embodiment of a teacher behavior detection method according to the present application;

FIG. 6 is a flowchart illustrating a specific application scenario of the teacher behavior detection method according to the present application;

FIG. 7 is a schematic structural diagram of an embodiment of a human behavior detection system according to the present application;

FIG. 8 is a schematic block diagram of an embodiment of a teacher behavior detection system of the present application;

FIG. 9 is a schematic structural diagram of an embodiment of an intelligent terminal according to the present application;

FIG. 10 is a schematic structural diagram of an embodiment of a computer-readable storage medium of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, 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 application.

Referring to fig. 1, fig. 1 is a schematic flow chart of a human behavior detection method according to a first embodiment of the present application. The embodiment comprises the following steps:

s11: the monitoring device acquires a monitoring image including a target human body and determines a target area in the monitoring image.

Specifically, the monitoring device acquires a monitoring image of a target human body in a monitoring area of the monitoring device to further define a target area in the acquired monitoring image, for example, a detection frame is correspondingly generated in a set area in the monitoring image, so as to determine an area in the detection frame as the target area.

The monitoring image may be a monitoring video, or may be a monitoring picture obtained by extracting a frame from the monitoring video, which is not limited in this application.

The target human body is a target object for detecting human body behaviors by the monitoring device, the target area is a designated area in the monitored image, and the area can be designated by a user or identified by a machine. Taking a target human body as a teacher and a monitoring area of the monitoring device as a classroom as an example, the designated area, that is, the target area, is mainly a platform area for the teacher to perform teaching activities.

Optionally, the monitoring device may be an intelligent electronic device with a video monitoring function, for example, one of intelligent electronic devices such as an intelligent camera, an intelligent robot, and an unmanned aerial vehicle, or a monitoring device formed by combining a camera and an intelligent terminal that establishes a wireless or wired communication connection with the camera, where the intelligent terminal may be one of a mobile phone, a tablet computer, a server, and the like, and the application does not limit this.

S12: and positioning the target human body in the monitoring image and determining the position information of the target human body.

Specifically, the target human body in the monitored image is subjected to human body detection and positioning to determine the position information of the target human body in the monitored image. In a specific embodiment, the monitoring image is input into a trained detection network model integrated inside the monitoring device, so as to obtain the position information of the target human body in the monitoring image through the detection network model.

S13: whether the target human body is within the target area is determined based on the position information.

Further, the monitoring device determines whether the target human body is in the target area in the monitored image based on the position information of the target human body in the monitored image, for example, determines whether the position information of the target human body in the monitored image matches the position information of the target area in the monitored image, that is, determines whether the position information of the target area in the monitored image includes the position information of the target human body in the monitored image or the position information of the center point of the target human body, so as to further determine whether the target human body is in the target area.

In another embodiment, the monitoring device generates a detection frame and a human body detection frame corresponding to the target region in the monitored image and the region surrounded by the joints of the target human body, so as to determine whether the target human body is in the target region by determining whether the human body detection frame is in the detection frame.

Wherein S14 is performed if the target human body is within the target region, and S17 is performed if the target human body is not within the target region.

S14: and determining at least one characteristic information of the orientation and the action of the target human body based on the monitoring image.

Specifically, when it is determined that the target human body is in the target area of the monitoring image, at least one feature information of the orientation and behavior of the target human body is further determined based on the monitoring image, for example, the monitoring image is input into a trained deep network learning model to determine whether the target human body faces or faces away from the monitoring lens or a specified direction, and the current behavior of the target human body is classified to determine whether the current behavior of the target human body belongs to one of the set actions, so as to determine at least one feature information of the orientation and the behavior of the target human body.

S15: and determining the action of the target human body by using the preset key points of the target human body.

Specifically, when the current behavior of the target human body is expressed as a set behavior, the corresponding joint points generally have specific relative position relationships, for example, the important joint points such as the head, wrist, elbow, shoulder, and ankle of the target human body are expressed as specific relative position relationships corresponding to the current behavior of the target human body, so that the current behavior of the target human body can be determined by detecting information of the preset key points of the target human body, that is, the relative position relationships of the important joint points of the target human body.

The preset key points are positions of 17 key points corresponding to all human bones and sequentially comprise a head, a neck, a left shoulder joint point, a right shoulder joint point, a left elbow joint, a right elbow joint, a left wrist joint, a right wrist joint, a left chest, a right chest, a left hip, a right hip, a pelvis, a left knee joint, a right knee joint, a left ankle and a right ankle of a human body.

S16: and determining the current behavior of the target human body according to the at least one characteristic information and the action of the target human body.

Specifically, after determining at least one of the orientation and the motion of the target human body based on the monitored image, the motion of the target human body is further determined by using the preset key points of the target human body, so that when the current motion of the target human body determined by using the preset key points of the target human body is inconsistent with the motion of the target human body determined based on the monitored image, a result with a low recall rate or confidence coefficient in classification results given based on the monitored image is deleted, thereby improving the accuracy of classifying the behavior type of the target human body and finally determining the current motion of the target human body. For example, when the current action of the target human body is determined to be the writing board based on the monitoring image through the trained deep learning network classification model, the preset key points of the target human body are further utilized to detect and classify the current action of the target human body again, so that when the current action of the target human body is determined not to be the writing board and the recall rate of the current action given by the deep learning network classification model as the writing board result is low, the current action given based on the monitoring image is deleted as the writing board classification result, the classification result given based on the monitoring image is continuously added and deleted, and the accuracy of classifying the action type of the target human body is further improved.

S17: and continuously keeping real-time monitoring.

Specifically, when the target human body is judged not to be in the target area of the monitored image based on the position information of the target human body in the monitored image, the real-time monitoring is continuously maintained, and the detection and judgment are not carried out on the behavior action of the target human body.

Different from the prior art, the human behavior detection method in the application acquires a monitoring image including a target human body through a monitoring device, determines a target area in the monitoring image, positions the target human body in the monitoring image, determines position information of the target human body, judges whether the target human body is in the target area or not based on the position information, determines at least one feature information of orientation and action of the target human body based on the monitoring image if the target human body is in the target area, determines action of the target human body by using a preset key point of the target human body, determines current action of the target human body according to the at least one feature information and the action of the target human body, can detect real-time action of the target human body appearing in the target area of the monitoring image, and detects and classifies the current action of the target human body based on the monitoring image, and further, by detecting and classifying preset key points of the target human body, when the inconsistency between the preset key points and the classification result given based on the monitoring image is determined, the classification result with lower recall rate or confidence coefficient in the classification result given based on the monitoring image is deleted, so that the accuracy of behavior judgment finally performed on the target human body is effectively improved.

Referring to fig. 2, fig. 2 is a schematic flow chart of a human behavior detection method according to a second embodiment of the present application. The embodiment comprises the following steps:

s21: the monitoring device monitors the set monitoring area in real time so as to acquire a monitoring image including the target human body when the target human body appears in the set monitoring area.

Specifically, the monitoring device monitors the set monitoring area in real time to obtain a monitoring image including the target human body when detecting that the target human body appears in the set monitoring area, for example, to capture a monitoring video including the target human body, or to frame the monitoring video to obtain a corresponding monitoring picture including the target human body, so as to send the monitoring picture to a processing hub of the monitoring device.

The set monitoring area is the maximum monitoring area that can be captured after the monitoring device is installed in a designated area, and the designated area is determined by a user.

S22: and defining a target area in the monitoring image according to the activity area of the target human body.

Specifically, to detect the behavior of the target human body, a main activity area of the target human body in a monitoring area of a monitoring device is determined first, or a main observation area of the target human body in the monitoring area of the monitoring device is determined according to the behavior and motion type of the behavior detection performed on the target human body, so as to determine the main activity area as the target area, and then the target area is defined in a corresponding monitoring image, for example, a detection frame is generated correspondingly in the activity area of the target human body in the monitoring image, so as to determine an area in the detection frame as the target area.

S23: and positioning the target human body in the monitoring image by using the trained detection network model so as to determine the position information of the target human body.

Specifically, the acquired monitoring image is input into a trained detection network model integrated in the monitoring device, so that the position information of the target human body in the monitoring image is obtained through the detection network model.

S24: whether the target human body is within the target area is determined based on the position information.

Further, the monitoring device determines whether the target human body is in the target region in the determined monitoring image based on the position information of the target human body in the monitoring image, for example, determines whether the position information of the target human body in the monitoring image matches with the position information of the target region in the monitoring image, that is, determines whether the position information of the target region in the monitoring image includes the position information of the target human body in the monitoring image or the position information of the center point of the target human body, so as to further determine whether the target human body is in the target region.

In another embodiment, the monitoring device generates a detection frame and a human body detection frame at the target area and the position of the target human body in the monitored image respectively, so as to determine whether the target human body is in the target area by determining whether the human body detection frame is in the detection frame.

Wherein S25 is performed if the target human body is within the target region, and S29 is performed if the target human body is not within the target region.

S25: and carrying out classification prediction on the orientation and/or action type of the target human body in the monitored image by utilizing a classification model of the deep learning network to obtain the orientation and/or action of the target human body.

Specifically, when the target human body is determined to be in the target area of the monitored image, the monitored image is input into a classification model of a trained deep learning network, so as to perform classification prediction on the orientation of the target human body in the monitored image through the classification model, for example, the condition that the target human body is facing or facing away from a monitoring lens or a specified direction. The type of the current behavior and action of the target human body in the monitored image can be classified and predicted through the classification model, so that the direction and/or action of the target human body can be obtained.

The classification prediction of the orientation of the target human body and the classification prediction of the motion type of the target human body may be classification models using the same deep learning network or classification models using different deep learning networks, which is not limited in the present application.

The training method of the classification model of the deep learning network comprises the following steps: the method comprises the steps of obtaining a monitoring image with the direction and the action of a target human body marked, inputting the target image into a preset network model, giving a corresponding classification prediction result through the preset network model, and further training the preset network model through the classification prediction result and the marking type of the target human body in the monitoring image, so that the classification model of the deep learning network is obtained.

S26: and detecting the positions of preset key points of the target human body by using a deep learning network, and classifying the behavior types of the target human body according to the positions of the preset key points to obtain the action of the target human body.

And further, inputting the monitoring image which comprises the target human body and is in the target area in the monitoring image into a deep learning network so as to detect the position of the preset key point of the target human body through the deep learning network.

When the current behavior of the target human body is represented as a set type of behavior, the corresponding joint points generally have a specific relative position relationship, for example, important joint points such as the head, the wrist joint, the elbow joint, the shoulder joint and the ankle of the target human body correspond to the current behavior of the target human body and are represented as a specific relative position relationship, for example, when the target human body is writing a blackboard-writing, the position of the wrist joint of the target human body in the standing direction of the target human body is higher than the position of the shoulder joint or the elbow joint of the target human body, so that the current behavior of the target human body can be determined by detecting the position of preset key points of the target human body, that is, by detecting the relative position relationship of the important joint points of the target human body.

The deep learning network can classify the behavior type of the target human body according to the detected position of the preset key point, for example, the deep learning network can establish a plane coordinate system by using a vertex of the upper left corner or the lower left corner of the monitoring image as an origin, detect the coordinate value of the corresponding preset key point on the target human body in the plane coordinate system, and further classify the behavior type of the target human body according to the preset position corresponding relation and the coordinate value of the preset key point, so as to determine the action of the target human body. The preset key points can be used for reasonably selecting important joint points on the target human body as required, such as any two or more of the important joint points of the head, the wrist joint, the elbow joint, the shoulder joint, the ankle and the like of the target human body, and the application does not limit the important joint points.

S27: and determining the current behavior of the target human body according to the at least one type of characteristic information, and correcting the determined result of the current behavior of the target human body through the action of the target human body to obtain the classification result of the current behavior of the target human body.

Specifically, the behavior types of the target human body in the monitored image are classified by using a classification model of the deep learning network, so as to determine the current behavior of the target human body according to at least one kind of feature information in the orientation and the motion of the target human body, and further the behavior types of the target human body are classified by using the position of a preset key point of the target human body detected by using the deep learning network, so that the motion of the target human body is obtained to correct the classification result of the current behavior of the target human body determined according to the at least one kind of feature information, for example, the classification result with low recall rate in the current behavior of the target human body determined according to the at least one kind of feature information is added or deleted, so as to improve the accuracy of classifying the behavior types of the target human body, and thus the final classification result of the current behavior of the target human body is obtained.

S28: and respectively counting the occurrence frequency, single duration and total duration of each current behavior of the target human body within a set time, and displaying the counting result.

Furthermore, the monitoring device respectively counts the occurrence frequency, single duration and total duration of each behavior action of the target human body, such as the behavior actions of setting actions towards a specified direction, back to the specified direction and back to the specified direction, within a set time, and displays the statistical result through a display screen on the monitoring device, so that the behavior of the target human body within the set time can be analyzed or scored according to the statistical result.

The set time can be any reasonable time such as 45 minutes or 60 minutes, and can be set reasonably by a user according to needs, and the comparison is not limited in the application.

S29: and continuously keeping real-time monitoring.

Referring to fig. 3, fig. 3 is a schematic flow chart of a human behavior detection method according to a third embodiment of the present application. The human behavior detection method of the present embodiment is a flowchart of a detailed embodiment of the human behavior detection method in fig. 2, and includes the following steps:

s31: the monitoring device monitors the set monitoring area in real time so as to acquire a monitoring image including the target human body when the target human body appears in the set monitoring area.

S32: and defining a target area in the monitoring image according to the activity area of the target human body.

S33: and positioning the target human body in the monitoring image by using the trained detection network model so as to determine the position information of the target human body.

S34: whether the target human body is within the target area is determined based on the position information.

S31, S32, S33, and S34 are the same as S21, S22, S23, and S24 in fig. 2, respectively, for details, please refer to S21, S22, S23, S24, and the related text descriptions thereof, which are not repeated herein.

S35: and carrying out target human body detection on the target area to obtain a characteristic image comprising a target human body detection frame.

Specifically, when the target human body is determined to be in the target area of the monitored image, the target human body is detected in the target area, so that a detection frame is correspondingly generated in an area surrounded by all joint points of the target human body, and the characteristic image comprising the target human body detection frame is obtained.

S36: and expanding the area of the target human body detection frame in the characteristic image according to a preset proportional coefficient so as to intercept the target image in the expanded target human body detection frame.

Further, the area of the target human body detection frame in the feature image is enlarged according to a preset scale factor, for example, the two ends of the target human body detection frame in the target human body standing direction are respectively extended and enlarged by 10%, and the two ends of the target human body detection frame in the direction perpendicular to the target human body standing direction are respectively extended and enlarged by 25%, so as to intercept the image in the enlarged target human body detection frame, thereby obtaining a corresponding target image.

It can be understood that the region defined by the current joint points of the target human body is determined as the target human body detection frame, obviously, the target human body detection frame cannot completely contain each behavior action which may occur in the subsequent target human body, such as extending the arm or the leg, and the screenshot is performed after the area of the target human body detection frame is enlarged, so that each behavior action of the target human body can be effectively displayed in the screenshot, and a margin is left.

In other embodiments, the preset scaling factor for expanding the target human body detection frame may also be other combinations, which may be set by the user as needed, so as to effectively include each possible behavior of the target human body, and the comparison is not limited in the present application.

S37: and carrying out classification prediction on the orientation and/or action type of the target human body in the target image by utilizing a classification model of the deep learning network to obtain the orientation and/or action of the target human body.

Further, after the target image is acquired, the target image is input into a classification model of a trained deep learning network, so as to perform classification prediction on the orientation of the target human body in the target image through the classification model, for example, the situation that the target human body is facing or facing away from a monitoring lens or a specified direction. The type of the current behavior and action of the target human body in the monitored image can be classified and predicted through the classification model, so that the direction and/or action of the target human body can be obtained.

The training method of the classification model of the deep learning network comprises the following steps: the method comprises the steps of obtaining a target image with the direction and the action of a target human body marked, inputting the target image into a preset network model, giving a corresponding classification prediction result through the preset network model, and training the preset network model through the classification prediction result and the marking type of the target human body in the target image, so that the classification model of the deep learning network is obtained.

It can be understood that compared with the method of directly classifying and predicting the orientation and/or the type of the motion of the target human body in the monitored image by using a classification model of a deep learning network, corresponding model training is performed on the target image, and a classification result is given by the classification model, the motion characteristics of the target human body in the target image can be obviously highlighted, so that the accuracy of giving the classification result to the orientation and/or the type of the motion of the target human body can be effectively improved, and the calculation amount of a corresponding monitoring device is effectively reduced.

S38: and detecting the positions of preset key points of the target human body by using a deep learning network, and classifying the behavior types of the target human body according to the positions of the preset key points to obtain the action of the target human body.

S39: and determining the current behavior of the target human body according to the at least one type of characteristic information, and correcting the determined result of the current behavior of the target human body through the action of the target human body to obtain the classification result of the current behavior of the target human body.

S310: and respectively counting the occurrence frequency, single duration and total duration of each current behavior of the target human body within a set time, and displaying the counting result.

S311: and continuously keeping real-time monitoring.

S38, S39, S310, and S311 are the same as S26, S27, S28, and S29 in fig. 2, respectively, and please refer to S26, S27, S28, S29 and their related text descriptions, which are not repeated herein.

Referring to fig. 4, fig. 4 is a flowchart illustrating a teacher behavior detection method according to a first embodiment of the present application. The teacher behavior detection method of the embodiment is a specific application of the human behavior detection method. The embodiment comprises the following steps:

s41: the monitoring device acquires a monitoring image including a teacher and determines a platform area in the monitoring image.

Specifically, the monitoring device acquires a monitoring image of a teacher appearing in a monitoring area thereof to further define a platform area in the acquired monitoring image, for example, a detection frame is generated corresponding to the platform area of the teacher performing teaching activities in the monitoring image, so as to determine an area in the detection frame as the platform area.

S42: and positioning the teacher in the monitoring image and determining the position information of the teacher.

Specifically, human body detection and positioning are carried out on the teacher in the monitoring image so as to determine the position information of the teacher in the monitoring image. In a specific embodiment, the monitoring image is input into a trained detection network model integrated inside the monitoring device, so as to obtain the position information of the teacher in the monitoring image through the detection network model.

S43: and judging whether the teacher is in the platform area or not based on the position information.

Further, the monitoring device judges whether the teacher is in the platform area in the monitoring image based on the position information of the teacher in the monitoring image, for example, judges whether the position information of the teacher in the monitoring image matches with the position information of the platform area in the monitoring image, that is, judges whether the position information of the platform area in the monitoring image includes the position information of the teacher in the monitoring image or the position information of the center point of the image corresponding to the teacher, so as to further judge whether the target human body is in the platform area.

In another embodiment, the monitoring device generates a detection frame and a human body detection frame corresponding to the platform area in the monitoring image and the area surrounded by the joints of the teacher, so as to determine whether the teacher is in the platform area by determining whether the human body detection frame is in the detection frame.

If the teacher is in the lecture area, S44 is executed, and if the teacher is not in the lecture area, S47 is executed.

S44: determining whether the teacher faces the student and whether to write at least one feature information in the board book based on the monitoring image.

Specifically, when it is determined that the teacher is within the podium area of the monitor image, it is further determined whether the teacher faces the student based on the monitor image, for example, the monitor image is input into a trained deep web learning model to determine whether the teacher faces the student, and it is determined whether the teacher is writing a blackboard book, thereby determining whether the teacher faces the student and whether the teacher is writing at least one of feature information of the blackboard book.

S45: and determining the actions of the teacher by using the human body preset key points of the teacher.

Specifically, when the teacher is facing away from the student and writing on the blackboard, the corresponding important joint points generally have a specific relative position relationship, for example, the position of the wrist joint of the left hand or the right hand in the standing direction of the teacher is higher than the position of the shoulder joint or the elbow joint corresponding to the teacher, so that the current movement of the teacher can be determined by detecting the position of the preset key point of the human body of the teacher, that is, by detecting the relative position relationship of the important joint points of the teacher.

S46: and determining whether the teacher belongs to the writing board writing state currently according to the at least one characteristic information and the teacher action.

Specifically, in practical teaching, especially in teaching video recording and broadcasting, the learning state of students is affected due to the back-to-back state of teachers and the writing posture of writing on a board, and at the moment, the teaching supervision group cannot locate the problems in all directions, so that the behaviors of the teachers cannot be effectively supervised for a long time, the evaluation of the teachers is finally affected, and the recording and broadcasting education plays an important role in modern education. The recorded broadcast education mainly records the teaching process of a teacher through a video so as to play the video for students to study, but the back-to-back behavior of the teacher, the posture of writing on a writing board and other behaviors have great influence on the teaching effect.

In one embodiment, after determining whether the teacher belongs to at least one of feature information facing students and being written in blackboard writing based on the monitoring image, the action of the teacher is further determined by using the preset human body key points of the teacher, so that when the current action of the teacher determined by using the preset human body key points of the teacher is inconsistent with the action of the teacher determined based on the monitoring image, the result with low recall rate or confidence coefficient in the classification result given based on the monitoring image is deleted, the accuracy of classifying the action types of the teacher is improved, and the current action of the teacher is finally determined.

S47: and continuously keeping real-time monitoring.

Specifically, when it is judged that the teacher is not in the podium area of the monitored image based on the position information of the teacher in the monitored image, the real-time monitoring is continuously maintained without performing detection judgment on the behavior action of the teacher.

Referring to fig. 5, fig. 5 is a flowchart illustrating a teacher behavior detection method according to a second embodiment of the present application. The embodiment comprises the following steps:

s51: the monitoring device monitors the set monitoring area in real time so as to acquire a monitoring image including the teacher when the teacher appears in the set monitoring area.

Specifically, the monitoring device monitors the set monitoring area in real time, so that when it is detected that the teacher appears in the set monitoring area, a monitoring image including the teacher is obtained, for example, a monitoring video including the teacher is captured, or the monitoring video is subjected to frame extraction to obtain a corresponding monitoring picture including the teacher, so that the monitoring picture is sent to a processing hub of the monitoring device.

Here, the set monitoring area refers to a maximum monitoring area that can be photographed after the monitoring apparatus is installed in a designated area, for example, a classroom, and the designated area is determined by a user.

S52: and defining a platform area in the monitoring image according to the activity area of the teacher.

Specifically, to detect the behavior of the teacher, a main activity area of the teacher in a monitoring area of the monitoring device is first determined to define a platform area in the monitoring image, for example, a detection frame is generated corresponding to the platform area of the teacher in the monitoring image for teaching activity, so as to determine an area in the detection frame as the platform area.

S53: and positioning the teacher in the monitoring image by using the trained detection network model to determine the position information of the teacher.

Specifically, the acquired monitoring image is input into a trained detection network model integrated in the monitoring device, so that the position information of the teacher in the monitoring image is obtained through the detection network model.

S54: and judging whether the teacher is in the platform area or not based on the position information.

Further, the monitoring device determines whether the teacher is in the platform area in the monitored image based on the position information of the teacher in the monitored image, for example, determines whether the position information of the teacher in the monitored image matches the position information of the platform area in the monitored image, that is, determines whether the position information of the platform area in the monitored image includes the position information of the teacher in the monitored image or the position information of the center point of the image corresponding to the teacher, so as to further determine whether the teacher is in the platform area.

If the teacher is in the lecture area, S55 is executed, and if the teacher is not in the lecture area, S59 is executed.

S55: and classifying and predicting the type of the state that whether the teacher faces the students and/or the writing board in the monitored image by using a classification model of the deep learning network to obtain the state that whether the teacher faces the students and/or the writing board.

Specifically, when the teacher is determined to be in the platform area of the monitoring image, the monitoring image is input into a classification model of a trained deep learning network, so that whether the teacher faces the students in the monitoring image is classified and predicted through the classification model. And the classification model can be used for classifying and predicting the state of whether the teacher writes the blackboard writing in the monitored image so as to acquire the state of whether the teacher faces the students and/or writes the blackboard writing.

The classification prediction of whether the teacher faces the students and the classification prediction of whether the teacher writes on the blackboard writing state can be the same classification model of the deep learning network or different classification models of the deep learning networks, and the application does not limit the classification prediction.

The training method of the classification model of the deep learning network comprises the following steps: the method comprises the steps of obtaining a monitoring image of a labeled teacher facing students, a teacher facing away from students and not writing on a blackboard, and a teacher facing away from students and writing on a blackboard, inputting the monitoring image into a preset network model, giving a corresponding classification prediction result through the preset network model, and training the preset network model through the given classification prediction result and a labeling type made for the teacher in the monitoring image, so that a classification model of the deep learning network is obtained.

In another embodiment, before S55, the method further includes: when the podium area of the monitoring image is determined, the podium area is subjected to human body detection, so that the area defined by all the joint points of the teacher correspondingly generates a detection frame, and the characteristic image comprising the human body detection frame is obtained. Further, the area of the human body detection frame in the feature image is enlarged according to a preset scale factor, for example, the two ends of the human body detection frame in the standing direction of the teacher are respectively extended by 10%, and the two ends of the human body detection frame in the direction perpendicular to the standing direction of the teacher are respectively extended by 25%, so as to intercept the image in the enlarged human body detection frame, thereby obtaining a corresponding target image.

After the target image is acquired, the target image is input into a classification model of a trained deep learning network, so that whether a teacher faces students and/or the type of writing state in the target image is classified and predicted through the classification model, and whether the teacher faces the students and/or the writing state is acquired.

S56: and detecting the positions of the preset key points of the teacher by using the deep learning network, and classifying the behavior types of the teacher according to the positions of the preset key points to obtain the actions of the teacher.

And further, a teacher is included, and the monitoring image of the teacher in the speaking area in the monitoring image is input into the deep learning network so as to detect the position of the preset key point of the teacher through the deep learning network.

When the teacher writes a blackboard-writing to the back of the student, the corresponding joint points generally have a specific relative position relationship, for example, the position of the wrist joint of the left hand or the right hand of the teacher in the standing direction of the teacher is higher than the position of the shoulder joint or the elbow joint corresponding to the teacher, so that whether the teacher is writing the blackboard-writing currently can be determined by detecting the position of the preset key point of the teacher, that is, by detecting the relative position relationship of the important joint points of the teacher.

The deep learning network can classify the behavior types of the teachers according to the positions of the preset key points of the teachers, for example, the deep learning network can establish a plane coordinate system by using a top left vertex or a bottom left vertex of the monitoring image as an original point, detect coordinate values of the corresponding preset key points of the teachers in the plane coordinate system, and further classify whether the teachers write blackboard writing according to the corresponding relation between the relative sizes of the corresponding coordinate values and whether the teachers write blackboard writing and the coordinate values of the important joint points of the teachers, so that the actions of the teachers are determined.

Specifically, in one embodiment, when it is detected that the ordinate value of the teacher's right wrist joint is greater than the ordinate value of the teacher's right shoulder joint or its right elbow joint, then it is determined that the teacher is writing on a blackboard.

S57: and determining whether the teacher belongs to the writing board writing state currently according to the at least one type of feature information, and correcting the determined classification result of whether the teacher belongs to the writing board writing state currently through the action of the teacher so as to obtain the classification result of whether the teacher belongs to the writing board writing state currently.

Specifically, a classification model of the deep learning network is used for classifying whether a teacher in a monitoring image belongs to a writing board book or not to obtain a classification result of whether the teacher belongs to the writing board book or not, and further the classification result of whether the teacher belongs to the writing board book or not is corrected by detecting the classification result of the position of a preset key point of the teacher to the state of whether the teacher writes the writing board book or not, for example, the classification result of whether the teacher belongs to the writing board book or not is deleted, wherein the classification result is determined according to the monitoring image that the current behavior of the teacher belongs to a non-writing board book facing away from students, such as explaining PPT to the students or having low recall rate in writing board books facing away from the students, so that the accuracy of classifying the behavior types of the teacher is improved, and the final classification result of the current behavior of the teacher is obtained.

S58: and respectively counting the occurrence frequency, single duration and total duration of the writing board of the teacher facing to students, the non-writing board of the back facing students and the writing board of the back facing students in the set time, and displaying the counting result.

Further, when detecting that the teacher is currently facing the student, or facing back to the student without writing on the blackboard, or facing back to the student writing on the blackboard, the monitoring device times the current behavior of the teacher, further counts the times of each behavior of the teacher in the set time, and respectively counts the single duration and the total duration, displays the counting result through a display screen on the monitoring device, and can provide the teaching supervision and guidance group for analyzing and evaluating the teaching activities of the teacher.

S59: and continuously keeping real-time monitoring.

In an embodiment, please refer to fig. 6, where fig. 6 is a flowchart illustrating a specific application scenario of the teacher behavior detection method according to the present application. After the monitoring device is started to perform video monitoring on the monitored area, the monitoring device is started to enter a human behavior detection mode, and as shown in fig. 6, S61 is sequentially executed to start; taking S62 when the target human body is detected to be present in the monitoring area, where taking the target human body as a teacher as an example, when the corresponding monitoring video including the teacher is acquired, the monitoring video is subjected to frame extraction to acquire the corresponding monitoring image.

Further executing S63, inputting the monitoring image into a deep learning network model to perform human body detection on the target human body through the deep learning network model, defining a podium regular region in the monitoring image, and generating a human body detection frame corresponding to the region enclosed by each joint of the target human body, then executing S64 to determine whether the human body detection frame is in the podium regular region in the monitoring image.

When the monitoring device determines that the human body detection frame is in the platform regular area in the monitored image, S65 and S67 are respectively performed, and the human body detection frame is expanded and captured according to a preset proportionality coefficient, for example, 10% of the two ends of the human body detection frame are respectively expanded in the direction in which the teacher stands, and 25% of the two ends of the human body detection frame are respectively expanded in the direction perpendicular to the direction in which the teacher stands, so as to obtain a corresponding target image.

And after the human body detection frame is expanded and the screenshot is carried out, respectively executing S66 and S68, wherein the step of expanding and screenshot the human body detection frame in S65 and S67 can be the same step, so that after the screenshot is obtained, the obtained target image is respectively input into the trained key point network model and the classification model of the deep learning network, and after the classification of the state whether a teacher in the target image faces a student or faces away from the student writing board is carried out, the step of executing S69 is carried out, and classification results given by the two network models are fused to remove the classification result with low recall rate.

And further executing S610 to obtain the final results of classifying the current behaviors of the teacher, and after the times of occurrence, the single duration and the total duration of each classification result in the set time are respectively subjected to statistical analysis, ending the human behavior detection task.

Based on the general inventive concept, the present application further provides a human behavior detection system, please refer to fig. 7, and fig. 7 is a schematic structural diagram of an embodiment of the human behavior detection system of the present application. The human behavior detection system 71 includes an intelligent terminal 711 and a camera 712 connected to the intelligent terminal 711.

The camera 712 is configured to obtain a monitoring image at least including a target human body, and specifically, the monitoring image including the target human body may be obtained by performing real-time monitoring shooting on a monitoring area of the camera 712 installed in a designated area, so that when the monitoring image including the target human body is obtained, the monitoring image is input into the intelligent terminal 711.

The intelligent terminal 711 is configured to receive the monitoring image sent by the camera 712, and determine a target area in the monitoring image, so as to locate a target human body in the monitoring image and determine position information of the target human body.

The intelligent terminal 711 further determines whether the target human body is in the target area based on the position information, so as to determine at least one of orientation and motion of the target human body based on the monitoring image when determining that the target human body is in the target area, and determine the motion of the target human body by using preset key points of the target human body, so as to be able to finally determine the current behavior of the target human body according to the at least one of the feature information and the motion of the target human body.

In another embodiment, the camera 712 may also be integrated into the intelligent terminal 711, for example, after a monitoring image at least including a target human body is acquired directly through the camera 712 carried by the intelligent terminal 711, such as an unmanned aerial vehicle or an intelligent robot, the monitoring image is processed by the processor of the intelligent terminal 711.

Optionally, the intelligent terminal 711 counts the occurrence frequency, the single duration and the total duration of each current behavior of the target human body within a set time, and displays the statistical result through a display screen of the intelligent terminal 711.

Optionally, the intelligent terminal 711 may be one of a mobile phone, a tablet computer, a server, and the like, which is not limited in this application.

Based on the general inventive concept, the present application further provides a teacher behavior detection system, please refer to fig. 8, and fig. 8 is a schematic structural diagram of an embodiment of the teacher behavior detection system of the present application. The teacher behavior detection system 81 includes an intelligent terminal 811 and a camera 812 connected to the intelligent terminal 811.

The camera 812 is configured to acquire a monitoring image at least including a teacher, and specifically, the monitoring image at least including a teacher may be acquired by monitoring and shooting an area in a classroom through the camera 812 installed at a set position in the classroom, and is input to the intelligent terminal 811.

The intelligent terminal 811 is configured to receive the monitoring image sent by the camera 812, determine a corresponding platform area in the monitoring image, locate a teacher in the monitoring image, determine position information of the teacher, and determine whether the teacher is in the platform area, so as to further determine whether the teacher faces a student and belongs to at least one type of feature information in a writing tablet state when determining that the teacher is in the platform area, and determine an action of the teacher by using a preset key point of a human body of the teacher, so as to determine whether the teacher currently belongs to the writing tablet state according to the at least one type of feature information and the action of the teacher.

In another embodiment, the camera 812 may be integrated into the intelligent terminal 811, for example, after acquiring a monitoring image including at least a teacher directly through the camera 812 carried by any kind of intelligent terminal 811, such as a smart camera, a drone or an intelligent robot, the monitoring image is processed through a processor of the intelligent terminal 811.

Optionally, the intelligent terminal 811 respectively counts the number of times, the single duration and the total duration of the writing of the blackboard writing of the teacher facing the students, facing away from the students, not writing the blackboard writing of the students and facing away from the students within a set time, and displays the counting result through the display screen thereof.

Optionally, the intelligent terminal 811 may be one of a mobile phone, a tablet computer, a server, and the like, which is not limited in this application.

Based on the general inventive concept, the present application further provides an intelligent terminal, please refer to fig. 9, and fig. 9 is a schematic structural diagram of an embodiment of the intelligent terminal of the present application.

The intelligent terminal 91 comprises a memory 911 and a processor 912, which are coupled to each other, wherein the memory 911 stores program data, and the processor 912 is configured to execute the program data to implement the detection method as described in any one of the above.

Based on the general inventive concept, the present application further provides a computer-readable storage medium, please refer to fig. 10, and fig. 10 is a schematic structural diagram of an embodiment of the computer-readable storage medium of the present application. In which a computer readable storage medium 101 has stored therein program data 1011, the program data 1011 being executable to implement any of the above-described detection methods.

In one embodiment, the computer-readable storage medium 101 may be a memory chip in a terminal, a hard disk, or a removable hard disk or other readable and writable storage tool such as a flash disk, an optical disk, or the like, and may also be a server or the like.

In the several embodiments provided in the present application, it should be understood that the disclosed method and apparatus may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a processor or a memory is merely a logical division, and an actual implementation may have another division, for example, a plurality of processors and memories may be combined to implement the functions or may be integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or connection may be an indirect coupling or connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

The above description is only for the purpose of illustrating embodiments of the present application and is not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application or are directly or indirectly applied to other related technical fields, are also included in the scope of the present application.

Claims

1. A human behavior detection method is characterized in that the human behavior detection comprises the following steps:

the method comprises the steps that a monitoring device obtains a monitoring image comprising a target human body, and determines a target area in the monitoring image;

positioning the target human body in the monitoring image, and determining the position information of the target human body;

judging whether the target human body is in the target area or not based on the position information;

if the target human body is in the target area, determining at least one of orientation and motion characteristic information of the target human body based on the monitoring image; and

determining the action of the target human body by using the preset key points of the target human body;

and determining the current behavior of the target human body according to the at least one characteristic information and the action of the target human body.

2. The human behavior detection method according to claim 1, wherein the step of determining at least one of orientation and motion characteristic information of the target human body based on the monitored image comprises:

and carrying out classification prediction on the orientation and/or action type of the target human body in the monitored image by utilizing a classification model of a deep learning network to obtain the orientation and/or action of the target human body.

3. The human behavior detection method according to claim 1, wherein the step of determining the action of the target human body using the preset key points of the target human body comprises:

and detecting the positions of preset key points of the target human body by using a deep learning network, and classifying the behavior types of the target human body according to the positions of the preset key points to obtain the actions of the target human body.

4. The human behavior detection method according to claim 1, wherein the step of determining the current behavior of the target human body according to the at least one feature information and the action of the target human body specifically comprises:

and determining the current behavior of the target human body according to the at least one type of characteristic information, and correcting the determined result of the current behavior of the target human body through the action of the target human body to obtain the classification result of the current behavior of the target human body.

5. The human behavior detection method according to claim 2, wherein before the step of performing classification prediction on the type of the orientation and/or the motion of the target human body in the monitored image by using a classification model of a deep learning network to obtain the orientation and/or the motion of the target human body, the method further comprises:

carrying out target human body detection on the target area to obtain a characteristic image comprising a target human body detection frame;

expanding the area of the target human body detection frame in the characteristic image according to a preset proportion coefficient so as to intercept the expanded target image in the target human body detection frame;

the step of performing classification prediction on the orientation and/or action type of the target human body in the monitored image by using a classification model of a deep learning network to obtain the orientation and/or action of the target human body comprises:

and carrying out classification prediction on the orientation and/or action type of the target human body in the target image by utilizing a classification model of a deep learning network to obtain the orientation and/or action of the target human body.

6. The human behavior detection method according to any one of claims 1 to 4, wherein the step of acquiring, by the monitoring device, the monitoring image including the target human body and determining the target region in the monitoring image specifically includes:

the monitoring device monitors a set monitoring area in real time to acquire the monitoring image including the target human body when the target human body appears in the set monitoring area;

and demarcating the target area in the monitoring image according to the activity area of the target human body.

7. The human behavior detection method according to any one of claims 1 to 4, wherein the step of locating the target human body in the monitored image and determining the position information of the target human body comprises:

and positioning the target human body in the monitoring image by utilizing the trained detection network model so as to determine the position information of the target human body.

8. The human behavior detection method according to claim 1, wherein after the step of determining the current behavior of the target human body according to the at least one feature information and the action of the target human body, the method further comprises:

and respectively counting the occurrence frequency, single duration and total duration of each current behavior of the target human body within a set time, and displaying the counting result.

9. A teacher behavior detection method is characterized by comprising the following steps:

the method comprises the steps that a monitoring device obtains a monitoring image including a teacher and determines a platform area in the monitoring image;

positioning a teacher in the monitoring image, and determining the position information of the teacher;

judging whether the teacher is in the platform area or not based on the position information;

determining, based on the monitoring image, whether the teacher is facing a student and whether to write at least one of character information of a board book if the teacher is in the podium area; and

determining the actions of the teacher by using the human body preset key points of the teacher;

and determining whether the teacher belongs to the writing board writing state currently according to the at least one characteristic information and the teacher action.

10. The teacher behavior detection method of claim 9, wherein after the step of determining whether the teacher currently belongs to a writing board status according to the at least one feature information and the teacher's actions, further comprising:

and respectively counting the occurrence frequency, single duration and total duration of the writing board of the teacher facing students, the non-writing board of the back students and the writing board of the back students in the set time, and displaying the counting result.

11. A human behavior detection system is characterized by comprising an intelligent terminal and a camera connected with the intelligent terminal;

the camera is used for acquiring a monitoring image including a target human body;

the intelligent terminal is used for receiving the monitoring image sent by the camera, determining a target area in the monitoring image, positioning the target human body in the monitoring image, determining position information of the target human body, judging whether the target human body is in the target area or not based on the position information, determining at least one feature information of orientation and action of the target human body based on the monitoring image when the target human body is determined to be in the target area, determining action of the target human body by using preset key points of the target human body, and determining current action of the target human body according to the at least one feature information and the action of the target human body.

12. A teacher behavior detection system is characterized by comprising an intelligent terminal and a camera connected with the intelligent terminal;

the camera is used for acquiring a monitoring image including a teacher;

the intelligent terminal is used for receiving the monitoring image sent by the camera, determining a platform area in the monitoring image, positioning a teacher in the monitoring image, determining the position information of the teacher, judging whether the teacher is in the platform area, determining whether the teacher faces students and writes at least one type of feature information in a writing board when determining the platform area, determining the actions of the teacher by utilizing the human body preset key points of the teacher, and determining whether the teacher belongs to the writing board state at present according to the at least one type of feature information and the actions of the teacher.

13. An intelligent terminal, characterized in that the intelligent terminal comprises a memory and a processor coupled to each other;

the memory stores program data;

the processor is adapted to execute the program data to implement the detection method according to any one of claims 1-8 or 9-10.

14. A computer-readable storage medium, characterized in that the computer-readable storage medium stores program data executable by a processor to implement the detection method according to any one of claims 1-8 or 9-10.