Method, device and equipment for identifying table-surfing behavior and storage medium

Abstract

The application discloses a method, device and equipment for recognizing a desk-crawling behavior and a storage medium. The method comprises the following steps: acquiring head and neck key points of a detection target and elbow and wrist key points of the same side; checking the detected human body key points by using a method for checking the detected human body key points, and if a correct conclusion cannot be drawn from the detected human body key points, the key points are terminated from being used for recognizing the desk-crawling behavior; acquiring shoulder key points of both shoulders of the detection target; calculating a head-shoulder relative angle between the head and the shoulder; obtaining a height position relationship between the head key points and the shoulder key points according to position information of the head key points and the shoulder key points; and when the height position relationship is within a preset height position relationship threshold range, if the head-shoulder relative angle is within a preset angle threshold range, it is determined that the detection target makes the desk-crawling behavior. The application has a lower demand for computing power, higher accuracy and wider applicable scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of video image processing, and more particularly, to a method and device for recognizing a desk-crouching behavior, an apparatus, and a storage medium. BACKGROUND

[0002] With the development of information technology in the field of education, it is becoming more and more common to observe and count the interactive behaviors of teachers and students in the classroom through automated and intelligent means, to supervise and objectively evaluate the learning state of each student, and to improve teaching management and the teaching quality of teachers in a targeted manner. The hand-raising action is the most typical behavior of students in the classroom, is the most intuitive form of students' active participation in the classroom content, and is the most direct way of interaction between teachers and students. It can be used as an effective way to evaluate the learning state of students and the teaching quality of teachers.

[0003] At present, the application logic of the common posture recognition method based on human key point detection in the teaching and research recording and broadcasting scene is generally as follows: ① a human key point detection algorithm is used to detect the skeletal key points of all candidate targets, for example, in the detection of the desk-crouching behavior, the head, shoulder joint, elbow joint, and wrist joint key points need to be detected first; ② different posture judgment methods are set according to different geometric arrangements using the detected key points; ③ or a posture classification model is used to directly classify the posture of the detected human target. An important prerequisite in the above scheme is that the detection result of the human key point detection algorithm in step ① has a high accuracy. To meet this condition, the common methods are: ① in the single-stage method, an algorithm model with a large number of parameters and a large amount of calculation is used, and generally the accuracy of such a model is high; ② in the two-stage method, a human detection algorithm is used to obtain the human bounding box in the image, and then the target in the bounding box is transmitted to a high-precision single-target human key point detector; the above two methods can suppress the false detection rate to a certain extent. However, it is obvious that these schemes have high requirements for computing power, and are not friendly to edge devices with insufficient computing power. SUMMARY

[0004] To overcome the low accuracy of the current desk-crouching detection method, the present application provides a method, device, apparatus, and storage medium for recognizing a desk-crouching behavior. The technical solutions adopted by the present application are as follows.

[0005] In a first aspect, the present application provides a method for recognizing a desk-crouching behavior, comprising:

[0006] obtaining the head key point, the neck key point, and the elbow key point and the wrist key point on the same side of the detection target;

[0007] The method for checking the detected human body key points is used to check the detected human body key points, and if a correct conclusion cannot be drawn, the current key points are terminated from being used for the prone desk behavior recognition.

[0008] Obtaining shoulder key points of the double shoulders of the detection target;

[0009] Calculating a head-shoulder relative angle between the head and the shoulders, wherein the head-shoulder relative angle is an angle formed by a line connecting the head key point and the neck key point and its extension line, and a line connecting the double shoulders and its extension line;

[0010] According to the position information of the head key point and the shoulder key point, a height position relationship of the head key point and the shoulder key point is obtained, wherein the height position relationship is used to describe a height position relationship between lines connecting the head key point and the shoulder key points of the double shoulders;

[0011] When the height position relationship is within a preset height position relationship threshold range, if the head-shoulder relative angle is within a preset angle threshold range, it is determined that the detection target performs the prone desk behavior;

[0012] The method for checking the detected human body key points comprises:

[0013] Obtaining head key points and neck key points of the detection target;

[0014] Calculating a distance from the head key point to the neck key point, denoted as a head length;

[0015] Obtaining elbow key points and wrist key points of the same side of the detection target;

[0016] Calculating a distance from the elbow key point to the wrist key point, denoted as a one-side arm length;

[0017] Based on a preset comparison model, the head length and the one-side arm length are compared to obtain a relative scale estimation value;

[0018] When the relative scale estimation value is within a preset scale threshold range, it is determined that the detected human body key points are correct.

[0019] In an embodiment, the preset comparison model in the method for checking the detected human body key points is:

[0020] wherein β is a relative scale estimation value, D H is the head length, D A is the arm length, and W and b are preset parameters.

[0021] In an embodiment, the method for checking the detected human key points further comprises:

[0022] If the head length is not within the preset head length range, it is determined that the human key points of the current detection target are not available.

[0023] If the one-side arm length is not within the preset arm length range, it is determined that the human key points of the current detection target are not available.

[0024] If it is determined that the human key points of the current detection target are not available, it is determined that the detected human key points are incorrect.

[0025] In an embodiment, the preset height position relationship threshold range is that the head key point is not higher than the line connecting the shoulder key points of the two shoulders.

[0026] In an embodiment, the head-shoulder relative angle is the acute angle formed by the line connecting the head key point and the neck key point and its extension line, and the line connecting the two shoulders and its extension line.

[0027] The preset angle threshold range is n degrees to 90 degrees, where 50≤n≤70.

[0028] In a second aspect, the present application provides a device for recognizing the behavior of lying on a table, comprising:

[0029] A first obtaining module is configured to obtain the head key point, the neck key point, and the elbow key point and the wrist key point of the same side of the detection target.

[0030] A first determining module is configured to use the device for checking the detected human key points to check the detected human key points, and if a correct conclusion about the detected human key points cannot be drawn, the behavior of lying on a table is recognized using the key points of this time.

[0031] A second obtaining module is configured to obtain the shoulder key points of the two shoulders of the detection target.

[0032] An angle calculating module is configured to calculate the head-shoulder relative angle between the head and the shoulder, where the head-shoulder relative angle is the angle formed by the line connecting the head key point and the neck key point and its extension line, and the line connecting the two shoulders and its extension line.

[0033] The second determining module is configured to obtain a height position relationship between the head key point and the shoulder key point according to position information of the head key point and the shoulder key point, wherein the height position relationship is used to describe a height position relationship between the head key point and a line connecting the two shoulder key points; when the height position relationship is within a preset height position relationship threshold range, if the head-shoulder relative angle is within a preset angle threshold range, it is determined that the detection target makes a table-crouching behavior.

[0034] The device for checking the detected human key points comprises:

[0035] The acquisition module is configured to acquire a head key point and a neck key point of the detection target, and acquire an elbow key point and a wrist key point on the same side of the detection target.

[0036] The calculation module is configured to calculate a distance between the head key point and the neck key point, denoted as a head length, and calculate a distance between the elbow key point and the wrist key point, denoted as a one-side arm length.

[0037] The comparison module is configured to compare the head length and the one-side arm length based on a preset comparison model to obtain a relative scale estimation value.

[0038] The determining module is configured to determine that the detected human key points are correct when the relative scale estimation value is within a preset scale threshold range.

[0039] In an embodiment, the preset comparison model in the device for checking the detected human key points is:

[0040] wherein β is the relative scale estimation value, D H is the head length, D A is the arm length, and W and b are preset parameters.

[0041] In an embodiment, the determining module of the device for checking the detected human key points is further configured to determine that the human key points of the current detection target are unavailable when the head length is not within a preset head length range, determine that the human key points of the current detection target are unavailable when the one-side arm length is not within a preset arm length range, and determine that the detected human key points are incorrect when it is determined that the human key points of the current detection target are unavailable.

[0042] In a third aspect, the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method of any of the above embodiments.

[0043] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein the program is executed by a processor to implement the method of any of the above embodiments.

[0044] In the present application, for the scene of classroom teaching, whether the detected human key points are correct is checked and detected by a relatively simple method through the relationship between the head length and the arm length; then, in the case that the key points are correct, the relative position and the relative angle of the height of the head and the shoulders are compared with the preset threshold range respectively, and whether the desk crouching behavior occurs is judged according to the comparison result. Compared with the judgment mode in the prior art, the present application has lower demand for computing power, higher accuracy and wider applicable scene. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a whole flow schematic diagram of the embodiment one of the present application.

[0046] Figure 2 is a flow schematic diagram of the detection method of the embodiment one of the present application.

[0047] Figure 3 is a whole structure schematic diagram of the embodiment two of the present application. DETAILED DESCRIPTION

[0048] Exemplary embodiments of the present disclosure will be described more fully hereinafter with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0049] It should be noted that the terms "first", "second", and the like in the embodiments of the present application are only to distinguish similar objects, and do not represent a specific order of the objects. Understandably, "first", "second", and the like can be interchanged in a specific order or sequence as allowed. It should be understood that the objects distinguished by "first", "second", and the like can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0050] Embodiment one

[0051] Please refer to Figure 1 , Figure 1A flowchart of a method for recognizing a desk-crawling behavior provided by Embodiment One of the present application is shown in FIG. 1. The method includes steps S110, S120, S130, S140, S150, and S160. It should be noted that steps S110, S120, S130, S140, S150, and S160 are merely reference numerals for clearly explaining the embodiments and the corresponding relationship of the drawings, and do not represent the order of the steps in the embodiments. Figure 1

[0052] In step S110, the head key point, the neck key point, and the elbow key point and the wrist key point of the same side of the detection target are obtained.

[0053] In step S120, the detected human key points are checked using a method for checking the detected human key points. If a correct conclusion cannot be drawn from the detected human key points, the recognition of the desk-crawling behavior using the key points this time is terminated.

[0054] In step S130, the shoulder key points of the shoulders of the detection target are obtained.

[0055] In step S140, the head-shoulder relative angle between the head and the shoulder is calculated. The head-shoulder relative angle is the angle formed by the intersection of the line and the extension line connecting the head key point and the neck key point and the line and the extension line connecting the shoulders.

[0056] In step S150, the height position relationship between the head key point and the shoulder key point is obtained according to the position information of the head key point and the shoulder key point. The height position relationship is used to describe the height position relationship between the line connecting the head key point and the shoulder key point of the shoulder.

[0057] In step S160, when the height position relationship is within a preset height position relationship threshold range, if the head-shoulder relative angle is within a preset angle threshold range, it is determined that the detection target has performed the desk-crawling behavior.

[0058] As shown in FIG. 2, a flowchart of a method for checking the detected human key points is shown in FIG. 2. The method includes steps S210, S220, S230, S240, S250, and S260. Figure 2 Figure 2 In step S210, the head key point and the neck key point of the detection target are obtained.

[0059] In step S220, the distance from the head key point to the neck key point is calculated, which is denoted as the head length.

[0060] In step S230, the distance from the neck key point to the head length is calculated, which is denoted as the neck length.

[0061] ​​Step S230, acquire the elbow key point and the wrist key point on the same side of the detection target;

[0062] Step S240, calculate the distance between the elbow key point and the wrist key point, denoted as a side arm length;

[0063] Step S250, based on a preset comparison model, compare the head length and the side arm length to obtain a relative scale estimation value;

[0064] Step S260, when the relative scale estimation value is within a preset scale threshold range, determine that the detected human key points are correct.

[0065] The method is suitable for devices capable of analyzing human actions, such as recording and broadcasting hosts, etc. For the convenience of explanation, the recording and broadcasting host is taken as an example to illustrate the method. The recording and broadcasting host acquires the video stream captured from the recording and broadcasting terminal, uses a human key point detection algorithm to perform full-image detection on the video frames in the video stream, and acquires all the skeletal key points of the human body in the video frames and the human body information corresponding to the key points. Specifically as follows:

[0066] (1) The human key points include: head, neck, both shoulders, both elbows, and both wrists, a total of 8 key points;

[0067] (2) In a classroom scene, the human target can generally only see the upper body, so the six key points of the lower body, i.e., both hips, both knees, and both ankles, are not considered, and only the key points of the upper body are calculated, which can reduce the computational overhead.

[0068] The common method of posture recognition based on skeletal key points often ignores an important judgment premise, that is, the accuracy of the key points detected by the human key point detection algorithm. Since the human key point detection algorithm cannot guarantee that every detected key point is correct, a judgment criterion needs to be added for preliminary screening. Steps S110 and S120 are to first judge the correctness of the key points of the detection target, and if they are not correct, the key points will not be used for desk slouching behavior detection.

[0069] Specifically, when judging the correctness of the key points of the detection target, a method for checking the detected human key points is used.

[0070] In the recording and broadcasting scene of classroom teaching and research analysis, the head is a relatively less occluded body part, and the probability of error in the human key point detection algorithm is also relatively low compared to other key points. Therefore, in steps S210 and S220, the length of the head of the detected target is suitable as the scale standard for the target, and the distance between the head key point and the neck key point is calculated to obtain D HThe length of the head of the detected target at the current scale (panoramic or close-up shot).

[0071] Compared with the head, the arm is often misestimated or low-precision estimated by the human key point detection algorithm due to the small area in the picture or easy to be blocked, which will seriously affect the accuracy of the subsequent gesture recognition based on the skeleton key points. In steps S230 and S240, it is first assumed that all human key points are accurately detected, and if the elbow and wrist key points on the same side can be detected by the human key point detection algorithm, the distance between the elbow key point and the wrist key point is calculated to obtain D A The length of the left / right arm of the detected target at the current scale.

[0072] In step S250, the arm length D A and the head length D H are compared using a preset comparison model formula to obtain the relative scale estimation value β between the head and the arm. In step S260, the relative scale estimation value β is compared with a preset scale threshold range, and when it is within the preset scale threshold range, it is determined that the detected human key points are correct.

[0073] Because the target that does not meet the conditions of steps S250 and S260 has the following situations: first, under the premise that the human key point detection algorithm correctly detects all key points, the target does not make a detectable behavior gesture; second, the human key points are misdetected, resulting in unreliable length calculation data; and third, the camera is shot at different angles, and the target's arm and other parts are too small or too large in imaging scale due to angle differences. The three situations will seriously affect the accuracy of the subsequent gesture recognition method based on key points, so only when the above conditions are met, it can be determined that the detected human key points are correct.

[0074] In an embodiment, the preset comparison model is:

[0075] wherein β is the relative scale estimation value, D H is the head length, D A is the arm length, and W and b are preset parameters.

[0076] wherein W and b can be parameters obtained after training, and using an image dataset in which human key points have been labeled, the image scale information in the training data and the position information of each labeled key point are used to train W and b.

[0077] In an embodiment, the method of checking the detected human key points further comprises steps S270, S280 and S290.

[0078] Step S270, if the head length is not within the preset head length range, it is determined that the human key point of the current detection target is unusable.

[0079] Step S280, if the one-side arm length is not within the preset arm length range, it is determined that the human key point of the current detection target is unusable.

[0080] Step S290, when it is determined that the human key point of the current detection target is unusable, it is determined that the detected human key point is incorrect.

[0081] The head length and the arm length are compared with the respective preset length threshold range, and if any of the head length and the arm length is not within the respective preset length threshold range, it is considered that the human key point of the current detection target is unusable. This is because the target that does not meet the condition has two cases: the first case is that the size of the target in the image is very small, and does not reach the minimum standard of detection and recognition; the second case is that the human key point is misdetected, resulting in unreliable data obtained by length calculation.

[0082] It should be noted that steps S270 and S280 are to determine whether the detected human key point is usable or not. If it is unusable, it is necessarily that the human key point is incorrect. Therefore, the two steps are preferably performed before steps S250 and S260. If it is determined that it is unusable, there is no need to perform the following steps.

[0083] In the method for checking the detected human key point, for the scene of classroom teaching, the relationship between the head length and the arm length is used to check whether the detected human key point is correct in a relatively simple way. Compared with the judgment method in the prior art, the method for checking the detected human key point has lower demand for computing power and is suitable for a wider range of scenes.

[0084] After the correctness of the key point of the detected target is checked through steps S110 and S120, if it is determined to be correct, the subsequent step of detecting the behavior of lying on the table is entered.

[0085] In the recording and broadcasting scene of classroom teaching and research analysis, the main reference basis for detecting whether the target makes behaviors such as lying on the table is the relative position of the head and the head and shoulders.

[0086] The relative position of the head and the shoulders includes two aspects: the first aspect is the relative height position relationship, and the second aspect is the relative angle position relationship.

[0087] The first aspect, i.e. the process of step S150 and the judgment process before step S160, compares the position information of the head key point with the position information of the shoulder, where the position information of the shoulder is the line connecting the shoulder key points of the two shoulders, and if the position of the head key point is within a preset height position relationship threshold range of the position of the shoulder key point, the target enters the next step of judgment.

[0088] It should be noted that the preset height position relationship threshold range herein can be a height position relationship threshold range obtained by a person skilled in the art according to past experience, for example, the height of the position of the head is not more than a preset threshold from the position of the two shoulders, or the position of the head is lower than the position of the two shoulders, etc.

[0089] The second aspect, i.e. the process of step S140 and the judgment process after step S160, takes the angle formed by the intersection of the line connecting the head key point and the neck key point and its extension line, and the line connecting the two shoulders and its extension line, as the head-shoulder relative angle between the head and the shoulder. On the premise of meeting the conditions of the first aspect, the head-shoulder relative angle between the head and the shoulder is compared with a preset angle threshold. If the head-shoulder relative angle meets the preset angle threshold range, it is considered that the detection target exists the behavior of lying on the table; otherwise, it is considered that there can be other identifiable behaviors, which can be added to the candidate set of other posture judgments.

[0090] It should be noted that the head-shoulder relative angle is the angle formed by the intersection of the line connecting the head key point and the neck key point and its extension line, and the line connecting the two shoulders and its extension line, which can be acute or obtuse. At this time, it does not matter whether the acute angle or the obtuse angle is taken, as long as the range of the preset angle threshold can be matched, for example, if the acute angle is taken, the range of the preset angle threshold is 60 to 90 degrees, and then if the obtuse angle is taken, the range of the preset angle threshold is 90 to 120 degrees.

[0091] In an embodiment, the preset height position relationship threshold range is that the head key point is not higher than the line connecting the shoulder key points of the two shoulders.

[0092] The embodiment is the case that the position of the head is lower than the position of the two shoulders as described above.

[0093] In an embodiment, the head-shoulder relative angle takes the acute angle formed by the intersection of the line connecting the head key point and the neck key point and its extension line, and the line connecting the two shoulders and its extension line;

[0094] The range of the preset angle threshold is n degrees to 90 degrees, where 50≤n≤70.

[0095] In the embodiment, the best range of the range of the preset angle threshold is obtained according to actual experience.

[0096] In the method, for the scene of classroom teaching, whether the detected human key points are correct is checked in a relatively simple method through the relationship between the head length and the arm length; then, in the case that the key points are correct, the relative positions and the relative angles of the heights of the head and the shoulders are compared with preset threshold ranges respectively, and whether the desk-crouching behavior occurs is judged according to the comparison results. Compared with the judgment mode in the prior art, the method has lower demand for computing power, higher accuracy, and wider applicable scene.

[0097] Embodiment Two

[0098] Corresponding to the method of embodiment one, as shown in Figure 3 The application further provides a desk-crouching behavior recognition device 3, which comprises a first acquisition module 310, a first judgment module 320, a second acquisition module 330, an angle calculation module 340, and a second judgment module 350.

[0099] The first acquisition module 310 is used for acquiring the head key point, the neck key point of the detection target, and the elbow key point and the wrist key point of the same side;

[0100] The first judgment module 320 is used for checking the detected human key points by using the device for checking the detected human key points, and if the conclusion that the detected human key points are correct cannot be drawn, the desk-crouching behavior recognition using the key points this time is terminated;

[0101] The second acquisition module 330 is used for acquiring the shoulder key point of the shoulder of the detection target;

[0102] The angle calculation module 340 is used for calculating the head-shoulder relative angle between the head and the shoulder, wherein the head-shoulder relative angle is the angle formed by the intersection of the line and the extension line of the head key point and the neck key point and the line and the extension line between the shoulders;

[0103] The second judgment module 350 is used for obtaining the height position relationship of the head key point and the shoulder key point according to the position information of the head key point and the shoulder key point, wherein the height position relationship is used for describing the height position relationship between the line of the head key point and the shoulder key point of the shoulder of the shoulder; when the height position relationship is within the preset height position relationship threshold range, if the head-shoulder relative angle is within the preset angle threshold range, it is judged that the detection target makes the desk-crouching behavior.

[0104] The device for checking the detected human key points comprises:

[0105] The acquisition module is configured to acquire head key points and neck key points of a detection target; and acquire elbow key points and wrist key points of the same side of the detection target.

[0106] The calculation module is configured to calculate a distance from the head key points to the neck key points, denoted as a head length; and calculate a distance from the elbow key points to the wrist key points, denoted as a one-side arm length.

[0107] The comparison module is configured to compare the head length and the one-side arm length based on a preset comparison model to obtain a relative scale estimation value.

[0108] The judgment module is configured to judge that the detected human key points are correct when the relative scale estimation value is within a preset scale threshold range.

[0109] In an embodiment, the preset comparison model in the device for checking the detected human key points is as follows:

[0110] wherein β is a relative scale estimation value, D H is the head length, D A is the arm length, and W and b are preset parameters.

[0111] In an embodiment, the judgment module of the device for checking the detected human key points is further configured to judge that the human key points of the current detection target are unusable when the head length is not within a preset head length range; judge that the human key points of the current detection target are unusable when the one-side arm length is not within a preset arm length range; and judge that the detected human key points are incorrect when the human key points of the current detection target are judged to be unusable.

[0112] In an embodiment, the preset height position relationship threshold range is that the head key points are not higher than a line connecting the shoulder key points of the two shoulders.

[0113] In an embodiment, the head-shoulder relative angle is an acute angle formed by a line connecting the head key points and the neck key points and its extension line, and a line connecting the two shoulders and its extension line.

[0114] The preset angle threshold range is n degrees to 90 degrees, wherein 50≤n≤70.

[0115] In the device, for the scene of classroom teaching, whether the detected human key points are correct is checked in a relatively simple method through the relationship between the head length and the arm length; then, in the case that the key points are correct, the relative positions of the height and the angle of the head and the shoulders are compared with preset threshold ranges respectively, and whether the desk crouching behavior occurs is judged according to the comparison result. Compared with the judgment mode in the prior art, the device has lower demand for computing power, higher accuracy, and wider applicable scene.

[0116] Embodiment three

[0117] The embodiment of the present application also provides a storage medium, which has computer instructions stored thereon, and the instructions are executed by a processor to realize the desk crouching behavior recognition method of any one of the above embodiments.

[0118] Those skilled in the art can understand that all or part of the steps of the above method embodiments can be completed by program instruction related hardware, and the foregoing program can be stored in a computer readable storage medium, and the program is executed to perform the steps of the above method embodiments; and the foregoing storage medium includes mobile storage equipment, random access memory (RAM), read-only memory (ROM), magnetic disc or optical disc and various program code storage media.

[0119] Alternatively, the integrated unit of the present application can be stored in a computer readable storage medium if it is realized in the form of a software function module and sold or used as an independent product. Based on this understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, terminal or network device, etc.) execute all or part of the embodiments of the present application. The foregoing storage medium includes mobile storage equipment, RAM, ROM, magnetic disc or optical disc and various program code storage media.

[0120] Corresponding to the above computer storage medium, in one embodiment, a computer device is also provided, which includes a memory, an encoder and a computer program stored in the memory and executable on the encoder, wherein the encoder executes the program to realize the method for recognizing the desk crouching behavior in any one of the above embodiments.

[0121] The computer device checks whether the detected human body key points are correct in a relatively simple method through the relationship between the head length and the arm length for the scene of classroom teaching; then, in the case that the key points are correct, the relative positions of the heights and the relative positions of the angles of the head and the shoulders are combined, and are compared with preset threshold ranges respectively, and whether the desk crouching behavior occurs is judged according to the comparison result. Compared with the judgment mode in the prior art, the computer device has lower demand for computing power, higher accuracy, and wider applicable scene.

[0122] The technical features of the above-described embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above-described embodiments are not described, but as long as the combinations of the technical features do not exist contradictory, they shall be considered within the scope of the present disclosure.

[0123] Obviously, the above-described embodiments of the present application are merely examples for clearly illustrating the present application, and are not intended to limit the implementation modes of the present application. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the implementation modes are not required or possible to be exhausted. Any modification, equivalent replacement and improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the claims of the present application.

Claims

1. A method for identifying table lying behavior, characterized in that: include: Obtain the key points of the head and neck of the detection target, as well as the key points of the elbow and wrist on the same side; The detected key points of the human body are checked using a method for checking the detected key points of the human body. If a conclusion that the detected key points of the human body are correct cannot be drawn, the use of the key points for the table lying behavior recognition is terminated. Get the shoulder key points of the detection target's shoulders; Calculate the relative head-shoulder angle between the head and shoulders, where the relative head-shoulder angle is the angle formed by the intersection of the line connecting the key points of the head and the key points of the neck, and the line extending therefrom, with the line connecting the shoulders, and the line extending therefrom; Obtaining a height position relationship between the head key point and the shoulder key point based on the position information of the head key point and the shoulder key point, wherein the height position relationship is used to describe the height position relationship between the head key point and the line connecting the shoulder key points of both shoulders; When the height position relationship is within a preset height position relationship threshold range, and if the head-shoulder relative angle is within a preset angle threshold range, it is determined that the detected target is lying on the table; The method for checking the detected key points of the human body includes: Obtain the key points of the head and neck of the detection target; Calculate the distance from the head key point to the neck key point, and record it as the head length; Obtain the elbow key points and wrist key points on the same side of the detection target; Calculate the distance from the elbow key point to the wrist key point, and record it as the length of one side of the arm; Based on a preset comparison model, the head length is compared with the arm length on one side to obtain a relative scale estimation value; When the relative scale estimation value is within a preset scale threshold range, it is determined that the detected human body key points are correct.

2. The method for identifying table lying behavior according to claim 1, characterized in that: The preset comparison model in the method for checking the detected key points of the human body is: Where β is the relative scale estimate, D H is the head length, D A is the arm length, W and b are preset parameters.

3. The method for identifying table lying behavior according to claim 1, characterized in that: The method for checking the detected key points of the human body also includes: If the head length is not within the preset head length range, it is determined that the human body key points of the current detection target are unavailable; If the arm length on one side is not within the preset arm length range, it is determined that the key point of the human body of the current detection target is unavailable; When it is determined that the human body key points of the current detection target are unavailable, it is determined that the detected human body key points are incorrect.

4. The method for identifying table lying behavior according to claim 1, characterized in that: The preset height position relationship threshold range is that the head key point is not higher than the line connecting the shoulder key points of the two shoulders.

5. The method for identifying table lying behavior according to claim 1, characterized in that: The head-shoulder relative angle is the acute angle formed by the intersection of the line connecting the key points of the head and the key points of the neck and its extension, and the line connecting the shoulders and its extension; The preset angle threshold ranges from n degrees to 90 degrees, wherein 50≤n≤70.

6. A device for identifying table lying behavior, characterized in that: include: The first acquisition module is used to obtain the key points of the head and neck of the detection target, as well as the key points of the elbow and wrist on the same side; The first judgment module is configured to use a device for checking the detected key points of the human body to check the detected key points of the human body, and if a conclusion cannot be drawn that the detected key points of the human body are correct, terminate the use of the key points for the table lying behavior recognition; The second acquisition module is used to obtain the shoulder key points of the shoulders of the detection target; An angle calculation module, used to calculate the relative head-shoulder angle between the head and shoulders, wherein the relative head-shoulder angle is the angle formed by the intersection of the line connecting the key points of the head and the key points of the neck and the line extending therefrom, and the line connecting the shoulders and the line extending therefrom; A second judgment module is configured to determine a height position relationship between the head key points and the shoulder key points based on the position information of the head key points and the shoulder key points, wherein the height position relationship describes the height position relationship between the head key points and the line connecting the shoulder key points of both shoulders; when the height position relationship is within a preset height position relationship threshold range, and if the head-shoulder relative angle is within a preset angle threshold range, then it is determined that the detected target has engaged in a behavior of lying on the table; Wherein, the device for checking the detected key points of the human body includes: The acquisition module is used to obtain the key points of the head and neck of the detection target; obtain the key points of the elbow and wrist on the same side of the detection target; A calculation module is used to calculate the distance from the head key point to the neck key point, which is recorded as the head length; and calculate the distance from the elbow key point to the wrist key point, which is recorded as the arm length on one side; a comparison module, configured to compare the head length with the arm length on one side based on a preset comparison model to obtain a relative scale estimation value; The judgment module is used to judge whether the detected human body key points are correct when the relative scale estimation value is within a preset scale threshold range.

7. The device for identifying table lying behavior according to claim 6, characterized in that: The preset comparison model in the device for checking the detected key points of the human body is: Where β is the relative scale estimate, D H is the head length, D A is the arm length, W and b are preset parameters.

8. The device for identifying table lying behavior according to claim 6, characterized in that: The judgment module of the device for checking and detecting the key points of the human body is also used to judge that the key points of the human body of the current detection target are unavailable if the head length is not within the preset head length range; if the length of the arm on one side is not within the preset arm length range, judge that the key points of the human body of the current detection target are unavailable; when it is judged that the key points of the human body of the current detection target are unavailable, judge that the detected key points of the human body are incorrect.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images