A method, system and device for predicting high-fall risk on campus based on visual monitoring

By setting up a risk assessment table and a three-dimensional joint node diagram extraction model on campus, combining student posture and regional information, the problem of inaccurate prediction of high-fall risk in the existing technology is solved, and a more accurate and effective high-fall risk assessment is achieved.

CN118736499BActive Publication Date: 2024-11-15CHENGDU XUNDAO TECH CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411230401.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-04
Publication Date
2024-11-15
Estimated Expiration
2044-09-04

AI Technical Summary

Technical Problem

The existing technology has failed to predict whether students' behavior will cause high falls, and has not adjusted the student behavior accordingly in combination with different levels of danger in dangerous areas, resulting in the risk prediction in some areas being too sensitive or slow, affecting the accuracy of high falls risk. .

Method used

The public areas of the campus are divided by setting up hazardous area division rules, and a risk assessment table corresponding to the risk weight is set up in each area. The risk weight is determined based on the height of the nearest side wall and the straight line distance from the nearest side wall, and the behavior video is extracted. The three-dimensional node diagram in the image is input into the pose feature extraction model, and the pose feature information is output, and the multi-dimensional evaluation and prediction are performed based on student pose information and regional position information.

Benefits of technology

It improves the accuracy and effectiveness of high-fall risk prediction, reduces the impact of judgment results, and ensures accurate assessment in areas of different risk levels.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118736499B_ABST
    Figure CN118736499B_ABST
Patent Text Reader

Abstract

The present invention discloses a campus fall risk prediction method, system and equipment based on visual monitoring, which specifically relates to the field of campus safety protection technology. The technical highlights are as follows: the method comprises: dividing the campus public area into regions according to a preset danger zone division rule to obtain a plurality of danger zones with different risk weights, and each danger zone is provided with a risk assessment table corresponding to the risk weight; acquiring a behavior video image, extracting a three-dimensional joint point graph containing a plurality of posture feature key points from the behavior video image, inputting the three-dimensional joint point graph into a preset posture feature extraction model to output a plurality of posture feature information; determining the danger zone where the student is located based on the position coordinate information, using the risk assessment table of the danger zone and the plurality of posture feature information to perform one-to-one judgment to obtain a plurality of risk coefficients, using the plurality of risk coefficients to compare with a preset risk threshold value, and obtaining a fall risk prediction result by comparison.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of campus safety protection, and in particular to a method, system and equipment for predicting the risk of falling from height on campus based on visual monitoring. Background Art

[0002] With the rapid development of campus information management construction, intelligent management of the campus is the future development direction. Effectively solving campus safety issues and maintaining normal education and life order in schools are the ardent expectations and yearnings of the whole society to create a safe campus environment and realize safe education.

[0003] In recent years, there have been many incidents of students falling from heights on campus. These incidents include accidents. For example, when students are conducting outdoor activities during recess on balconies, corridors or other public areas, their activities are too large or their actions are too dangerous, exceeding the protective height of the side walls of the balconies, corridors or other public areas, resulting in accidents. For these accidents, it is difficult to achieve early warning with the existing technology, and the only way is to adopt a "one-size-fits-all" approach to limit the students' range of activities to protect them. This will cause a sense of oppression to the students, thereby affecting their mental health.

[0004] A Chinese patent with publication number CN111967400A discloses an artificial intelligence-based campus safety monitoring and early warning management system, including a campus area division module, a GPS positioning module, a student activity image acquisition module, an image preprocessing module, a student dangerous behavior information database, a behavior comparison and identification module, a duration statistics module, a risk coefficient statistics module, a general monitoring platform and an early warning module. The present invention divides the outdoor area in the campus and collects student activity images in each sub-area. At the same time, the student behavior in the collected student activity images is compared to identify whether it is a dangerous behavior. The general monitoring platform arranges relevant personnel to go to the scene to handle the identified dangerous behavior, and conducts graded early warning to avoid processing delays, thereby realizing intelligent monitoring and management of campus safety, effectively making up for the drawbacks existing in the current campus video surveillance and inspection system, greatly reducing the probability of dangerous student behavior, and thus ensuring the safety of students.

[0005] Although the above-mentioned technical scheme utilizes the collected student behaviors, compares them with the dangerous behaviors in the student dangerous behavior information database to identify and determine whether they are dangerous behaviors, so as to issue an early warning of the danger; however, the above-mentioned technical scheme does not make risk predictions on whether the student behaviors will cause falls from heights, and is unable to make risk prediction assessments on students falling from heights. At the same time, the judgment and assessment of student behaviors are not adjusted accordingly based on the different levels of danger in dangerous areas, resulting in the danger prediction being too sensitive in some areas, affecting students' normal activities, and the danger prediction being too slow in other areas, failing to accurately predict the risk of falling from heights, resulting in a lack of accuracy in the judgment of the risk of falling from heights, affecting the prediction results.

[0006] Therefore, the present invention aims to provide a method, system and device for predicting the risk of falling from height on campus based on visual monitoring, so as to solve the above-mentioned related problems. Summary of the invention

[0007] The technical problem to be solved by the present invention is that the prior art does not predict whether student behavior will cause a fall from a height, and does not make corresponding adjustments to the judgment and evaluation of student behavior in combination with the different degrees of danger in dangerous areas, resulting in overly sensitive danger predictions in some areas, affecting students' normal activities, and overly slow danger predictions in other areas, which cannot accurately predict the risk of a fall from a height, resulting in a lack of accuracy in the judgment of the risk of a fall from a height, affecting the prediction results. The purpose is to provide a method, system and device for predicting the risk of a fall from a height on campus based on visual monitoring, which divides the public areas of the campus into dangerous areas by establishing dangerous area division rules, and establishes a risk assessment table corresponding to the risk weight in each dangerous area, thereby constructing a risk assessment table for each dangerous area of ​​the public areas of the campus. Different prediction and evaluation systems are used to increase the accuracy of evaluation and judgment; at the same time, the risk weight of the dangerous area is obtained by combining the height of the nearest side wall and the straight-line distance to the nearest side wall, so as to construct a risk assessment table to judge the risk of falling from a high position, thereby increasing the effectiveness of risk prediction; by extracting a three-dimensional joint point map containing multiple posture feature key points from the behavioral video image, and then inputting the three-dimensional joint point map into a preset posture feature extraction model to output multiple posture feature information, so as to combine the student posture information and regional location information, and combine the dangerous areas with different degrees of danger and the dangerous behaviors of students in dangerous areas with different degrees of danger, so as to evaluate and predict whether the students are in danger of falling from a high position from multiple dimensions, further increase the accuracy of evaluation and prediction, and reduce the impact on the judgment results.

[0008] The present invention is achieved through the following technical solutions:

[0009] A campus height fall risk prediction method based on visual monitoring, the method comprising:

[0010] The campus public area is divided into regions according to the preset dangerous area division rules to obtain multiple dangerous areas with different risk weights, and each dangerous area is set up with a risk assessment table corresponding to the risk weight;

[0011] Acquire a behavior video image, extract a three-dimensional joint point graph containing multiple posture feature key points from the behavior video image, and input the three-dimensional joint point graph into a preset posture feature extraction model to output multiple posture feature information, wherein the multiple posture feature information includes human body posture angle, height from the ground, human body height data, and position coordinate information;

[0012] The dangerous area where the student is located is determined based on the position coordinate information, and a risk assessment table of the dangerous area and multiple posture feature information are used to make one-to-one judgments to obtain multiple risk coefficients. The multiple risk coefficients are compared with the preset risk thresholds to obtain a height fall risk prediction result.

[0013] Furthermore, the campus public area is divided into zones according to the preset dangerous area division rules, specifically:

[0014] Obtain the height of each dangerous side wall in the campus public area, and determine a first division coefficient that is inversely proportional to the side wall height;

[0015] Based on the straight-line distance from each dangerous sidewall, a second division coefficient is determined which is inversely proportional to the straight-line distance to the sidewall;

[0016] The risk weight is determined by using the sum of the first division coefficient and the second division coefficient, and the risk weight is compared with a preset division threshold to divide the campus public area into regions, thereby obtaining a plurality of dangerous areas with different risk weights.

[0017] Furthermore, each danger zone is established with a corresponding risk assessment table, and each risk assessment table includes: risk weight and assessment control items of the first risk coefficient, at least one human height assessment threshold and assessment control items of the second risk coefficient, at least one ground height assessment threshold and assessment control items of the third risk coefficient, and at least one human posture angle assessment threshold and assessment control items of the fourth risk coefficient.

[0018] Furthermore, a behavior video image is obtained, and a three-dimensional joint point graph containing multiple posture feature key points is extracted from the behavior video image, specifically:

[0019] Obtain a video image of the student's behavior in the danger zone, and input the behavior video image into a preset first neural network model to output a three-dimensional joint point diagram containing multiple posture feature key points of the student, wherein the multiple posture feature key points include head key points, shoulder key points, elbow key points, hip key points and ankle key points, and each posture feature key point is connected by an edge.

[0020] Furthermore, the preset first neural network model adopts an improved YOLO pose model, wherein the improved YOLO pose model includes replacing the convolution block in the Head module of the existing YOLO pose model with a GSConv convolution block; and replacing the Upsample upsampling module in the existing YOLO pose model with a CARAFE upsampling module.

[0021] Furthermore, the preset posture feature extraction model adopts a convolutional neural network model.

[0022] Furthermore, the dangerous area where the student is located is determined based on the position coordinate information, and a risk assessment table of the dangerous area and multiple posture feature information are used to make one-to-one judgments to obtain multiple risk coefficients. The multiple risk coefficients are compared with the preset risk thresholds to obtain the high-fall risk prediction results, which are specifically:

[0023] Based on the location coordinate information, the straight-line distance between the student and each dangerous side wall is obtained, and the straight-line distance to the nearest dangerous side wall is extracted to determine the dangerous area where the student is located, and the risk weight of the dangerous area is obtained;

[0024] Using the risk assessment table of the dangerous area, based on the risk weight, human body height data, height from the ground, and human body posture angle, determine a first risk factor, a second risk factor, a third risk factor, and a fourth risk factor;

[0025] The first risk coefficient, the second risk coefficient, the third risk coefficient and the fourth risk coefficient are summed, and the summation result is compared with a preset risk threshold to obtain a height fall risk prediction result.

[0026] The present invention also provides a campus height fall risk prediction system based on visual monitoring, which is used in any one of the campus height fall risk prediction methods based on visual monitoring described above, and the system includes:

[0027] The area division module is used to divide the campus public area into areas according to the preset dangerous area division rules to obtain multiple dangerous areas with different risk weights, and each dangerous area is equipped with a risk assessment table corresponding to the risk weight;

[0028] A feature extraction module is used to obtain a behavior video image, extract a three-dimensional joint point graph containing multiple posture feature key points from the behavior video image, and input the three-dimensional joint point graph into a preset posture feature extraction model to output multiple posture feature information, wherein the multiple posture feature information includes human body posture angle, height from the ground, human body height data and position coordinate information;

[0029] The risk prediction module is used to determine the dangerous area where the student is located based on the location coordinate information, use the risk assessment table of the dangerous area and multiple posture feature information to make one-to-one judgments to obtain multiple risk coefficients, and use the multiple risk coefficients to compare with the preset risk thresholds to obtain the height fall risk prediction result.

[0030] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0031] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any one of the methods described above are implemented.

[0032] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0033] 1. In the present invention, the public areas of the campus are divided into dangerous areas by establishing dangerous area division rules, and a risk assessment table corresponding to the risk weight is established in each dangerous area, so as to construct a different prediction and evaluation system for each high-fall risk area in the public areas of the campus, thereby increasing the accuracy of the evaluation and judgment; at the same time, the risk weight of the dangerous area is obtained by combining the height of the nearest side wall and the straight-line distance to the nearest side wall, so as to construct a risk assessment table to judge the high-fall risk of the location, thereby increasing the effectiveness of risk prediction; by extracting a three-dimensional joint point map containing multiple posture feature key points from the behavioral video image, and then inputting the three-dimensional joint point map into a preset posture feature extraction model to output multiple posture feature information, so as to combine the student posture information and the regional position information, and combine the dangerous areas with different degrees of danger and the dangerous behaviors of the students in the dangerous areas with different degrees of danger, so as to evaluate and predict whether the students are at risk of falling from a high place from multiple dimensions, further increasing the accuracy of the evaluation and prediction, and reducing the impact on the judgment result.

[0034] 2. In the present invention, the convolution block in the Head network module of the existing YOLO pose model is replaced with a GSConv convolution block to perform upsampling and downsampling, reduce the number of model parameters and the amount of calculation, and ensure the sampling effect and training accuracy; the Upsample upsampling module in the existing YOLO pose model is replaced with a CARAFE upsampling module, and the adaptability of the CARAFE module is used to greatly reduce the redundant calculation amount. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0036] Figure 1 1 is a method flow chart of a campus high fall risk prediction method based on visual monitoring in this embodiment;

[0037] Figure 2 It is a three-dimensional joint point diagram in a campus high fall risk prediction method based on visual monitoring in this embodiment;

[0038] Figure 3 This is a structural diagram of an improved YOLOpose model in a campus fall risk prediction method based on visual monitoring in this embodiment;

[0039] Figure 4 Schematic diagram of a module of a campus fall risk prediction system based on visual monitoring in this embodiment;

[0040] Figure 5 It is a structural schematic diagram of a computer device in this embodiment. DETAILED DESCRIPTION

[0041] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0042] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0043] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0044] As described in the background, the prior art does not predict the risk of whether students' behavior will cause falls from a height, and cannot predict the danger of students falling from a height. At the same time, the judgment and evaluation of students' behavior are not adjusted accordingly based on the different danger levels in dangerous areas. As a result, the danger prediction in some areas is too sensitive, affecting students' normal activities and behaviors, and the danger prediction in other areas is too slow, which cannot achieve the purpose of predicting falls from a height, resulting in a lack of accuracy in the judgment of falls risk prediction and affecting the prediction results.

[0045] Therefore, the present invention aims to provide a campus fall risk prediction method, system and equipment based on visual monitoring, which divides the campus public area into dangerous areas by establishing dangerous area division rules, and establishes a risk assessment table corresponding to the risk weight in each dangerous area, so as to construct a different prediction and evaluation system for each high-fall risk area in the campus public area, thereby increasing the accuracy of the evaluation and judgment; at the same time, the risk weight of the dangerous area is obtained by combining the height of the nearest side wall and the straight-line distance to the nearest side wall, so as to construct a risk assessment table to judge the high-fall risk of the location, thereby increasing the effectiveness of risk prediction; by extracting a three-dimensional joint point map containing multiple posture feature key points from the behavioral video image, and then inputting the three-dimensional joint point map into a preset posture feature extraction model to output multiple posture feature information, so as to combine the student posture information and the regional position information, and combine the dangerous areas with different degrees of danger and the dangerous behaviors of the students in the dangerous areas with different degrees of danger, so as to evaluate and predict whether the students are at risk of falling from a height from multiple dimensions, further increase the accuracy of the evaluation and prediction, and reduce the impact on the judgment result.

[0046] Example 1

[0047] See also Figure 1As shown, this embodiment provides a method for predicting the risk of falling from height on campus based on visual monitoring, the method comprising:

[0048] S1: The campus public area is divided into regions according to the preset dangerous area division rules to obtain multiple dangerous areas with different risk weights, and each dangerous area is set up with a risk assessment table corresponding to the risk weight;

[0049] It should be noted that, in this embodiment, the campus public area refers to a space where students can move freely, including corridors, balconies, playgrounds, and open rooftops. In other embodiments, it may also include other spaces where students can move freely, and no excessive restrictions are made here; a dangerous side wall refers to a building with a certain height and an upper part suspended in the air, with one side being an open platform where students can walk, and the other side being an external suspended space of the building with a high risk of falling, including balcony enclosure side walls, corridor enclosure side walls, and rooftop rooftop side walls. In other embodiments, it may also include other dangerous side walls, and no excessive restrictions are made here;

[0050] At the same time, in this embodiment, the risk weight is used to characterize the degree of danger of the area; the area is divided according to the preset danger area division rules based on the dangerous side walls of each area of ​​the campus, so that each public area obtains a risk weight corresponding to the degree of danger based on its own side wall height and the straight-line distance from the side wall, and then a corresponding risk assessment table is established for each dangerous area, so as to prevent students from falling from a height without restricting their activities.

[0051] S2: Obtaining a behavior video image, extracting a three-dimensional joint point graph containing multiple posture feature key points from the behavior video image, and inputting the three-dimensional joint point graph into a preset posture feature extraction model to output multiple posture feature information, wherein the multiple posture feature information includes human body posture angle, height from the ground, human body height data, and position coordinate information;

[0052] It should be noted that, in this embodiment, the behavior video image is obtained by various surveillance cameras installed in the public area of ​​the campus, and the video frame at the current moment, that is, the behavior video image at the current moment, is extracted from the surveillance video. In other embodiments, other image acquisition methods may also be used, such as setting up cameras, cameras or monitors, which will not be described in detail here;

[0053] At the same time, in this embodiment, see Figure 2 As shown, the multiple posture feature key points include a head key point 1, a shoulder key point 2, an elbow key point 3, a hip key point 4 and an ankle key point 5, and each posture feature key point is connected by an edge. In other embodiments, the multiple posture feature key points may also include a palm key point 6 and a knee key point 7, and no excessive restrictions are made here;

[0054] At the same time, in this embodiment, the human body posture angle refers to the angle of the side between the shoulder key point and the hip key point and the side between the ankle key point and the hip key point, with the hip key point as the center; the height above the ground refers to the height of the ankle key point from the ground; the human body height data refers to the height of the head key point from the ground; the position coordinate information refers to the position parameters within the public area of ​​the campus.

[0055] S3: Determine the dangerous area where the student is located based on the position coordinate information, use the risk assessment table of the dangerous area and multiple posture feature information to make one-to-one judgments to obtain multiple risk coefficients, and compare the multiple risk coefficients with the preset risk thresholds to obtain a high-fall risk prediction result.

[0056] Specifically, in the present embodiment, the public areas of the campus are divided into dangerous areas by establishing dangerous area division rules, and a risk assessment table corresponding to the risk weight is established in each dangerous area, so as to construct a different prediction and assessment system for each high-fall hazard area in the public areas of the campus, thereby increasing the accuracy of the assessment and judgment; at the same time, by establishing a risk assessment table based on the risk weight, the hazard factor of the location can be judged based on the height of the nearest side wall and the straight-line distance to the nearest side wall; by extracting a three-dimensional joint point map containing multiple posture feature key points from the behavioral video image, and then inputting the three-dimensional joint point map into a preset posture feature extraction model to output multiple posture feature information, the student posture information and the regional location information can be combined to evaluate and predict whether the student is in danger of falling from a height from multiple dimensions, further increasing the accuracy of the assessment and prediction, and reducing the impact on the judgment result.

[0057] Further, referring to Table 1 and Table 2, the campus public area is divided into regions according to the preset dangerous area division rules, specifically:

[0058] Obtain the height of each dangerous side wall in the campus public area, and determine a first division coefficient that is inversely proportional to the side wall height;

[0059] Based on the straight-line distance from each dangerous sidewall, a second division coefficient is determined which is inversely proportional to the straight-line distance to the sidewall;

[0060] It should be noted that, in this embodiment, the higher the side wall height, the smaller the first division coefficient, and the lower the side wall height, the larger the first division coefficient; the farther the straight-line distance from the side wall, the smaller the second division coefficient, and the closer the straight-line distance from the side wall, the larger the second division coefficient.

[0061] The risk weight is determined by using the sum of the first division coefficient and the second division coefficient, and the risk weight is compared with a preset division threshold to divide the campus public area into regions, thereby obtaining a plurality of dangerous areas with different risk weights.

[0062] Table 1 Comparison table of the first division coefficient and the second division coefficient

[0063]

[0064] Table 2 Comparison of risk weights and hazardous areas

[0065]

[0066] As shown in Table 1 and Table 2 above, in this embodiment, in the preset dangerous area division rule, the side wall height and the side wall straight-line distance are both set to six ranges, and each range corresponds to a first division coefficient or a second division coefficient. When the obtained side wall height of the dangerous side wall is 0.5m, the first division coefficient is 6; when the straight-line distance from the side wall is 0.8m, the second division coefficient is 5;

[0067] Therefore, a dangerous area with a straight-line distance of 0.4m to the nearest side wall and a height of 0.5m, and another dangerous area with a straight-line distance of 0.9m to the nearest side wall and a height of 0.2m, have the same risk weight. Therefore, the risk assessment tables of the two dangerous areas are consistent. The risk weight of the dangerous area can be obtained by combining the height of the nearest side wall and the straight-line distance to the nearest side wall, thereby constructing a risk assessment table to determine the risk of falling from height at that location.

[0068] Furthermore, each danger zone is established with a corresponding risk assessment table, and each risk assessment table includes: risk weight and assessment control items of the first risk coefficient, at least one human height assessment threshold and assessment control items of the second risk coefficient, at least one ground height assessment threshold and assessment control items of the third risk coefficient, and at least one human posture angle assessment threshold and assessment control items of the fourth risk coefficient.

[0069] Specifically, in this embodiment, taking the fifth risk area with a risk weight of 12 as an example, its corresponding risk assessment table is shown in Table 3 below:

[0070] Table 3 Risk assessment table for the fifth hazardous area

[0071]

[0072] Furthermore, a behavior video image is obtained, and a three-dimensional joint point graph containing multiple posture feature key points is extracted from the behavior video image, specifically:

[0073] Obtain behavioral video images of students in a dangerous area, and input the behavioral video images into a preset first neural network model to output a three-dimensional joint point map containing multiple posture feature key points of the students, wherein the multiple posture feature key points include head key points, shoulder key points, elbow key points, hip key points and ankle key points, and each posture feature key point is connected by an edge; the preset first neural network model adopts an improved YOLO pose model, wherein the improved YOLO pose model includes replacing the convolution block in the Head module of the existing YOLO pose model with a GSConv convolution block; and replacing the Upsample upsampling module in the existing YOLO pose model with a CARAFE upsampling module.

[0074] Specifically, in this embodiment, the preset first neural network model adopts an improved YOLO pose model. In other embodiments, other neural network models that can extract a three-dimensional joint point graph containing multiple posture feature key points of the student can also be used; the improved YOLO pose model includes replacing the convolution block in the Head module of the existing YOLO pose model with a GSConv convolution block; replacing the Upsample upsampling module in the existing YOLO pose model with a CARAFE upsampling module, specifically:

[0075] The existing YOLO pose model consists of four parts, including the Input input, Backbone network module, Neck network module and Head network module; the convolution block in the Head network module is replaced with the GSConv convolution block to perform upsampling and downsampling, reduce the number of model parameters and calculations, and ensure sampling effect and training accuracy; the Upsample upsampling module in the existing YOLOpose model is replaced with the CARAFE upsampling module, and the adaptability of the CARAFE module is used to greatly reduce the redundant calculation. The structure diagram of the improved YOLO pose model can be found in Figure 3 shown.

[0076] Furthermore, the preset posture feature extraction model adopts a convolutional neural network model.

[0077] Specifically, in this embodiment, the convolutional neural network model is composed of 8 convolutional layers, 5 pooling layers and 2 fully connected layers, including a softmax function. The network sets the size of the convolution kernel to 3×3×3 and the step size to 1×1×1 to perform convolution calculation on the input three-dimensional joint point map. After the calculation is completed, the feature image is pooled, and the size of the pooling kernel is 2×2×2 and the step size is 2×2×2. A large number of three-dimensional joint point maps with feature information annotations are obtained and input into the convolutional neural network model for training. The learning rate of the training is set to 0.001, and the training is stopped when the loss function is minimized to obtain the best posture feature extraction model that can extract multiple posture feature information.

[0078] Furthermore, the dangerous area where the student is located is determined based on the position coordinate information, and a risk assessment table of the dangerous area and multiple posture feature information are used to make one-to-one judgments to obtain multiple risk coefficients. The multiple risk coefficients are compared with the preset risk thresholds to obtain the high-fall risk prediction results, which are specifically:

[0079] Based on the location coordinate information, the straight-line distance between the student and each dangerous side wall is obtained, and the straight-line distance to the nearest dangerous side wall is extracted to determine the dangerous area where the student is located, and the risk weight of the dangerous area is obtained;

[0080] Using the risk assessment table of the dangerous area, based on the risk weight, human body height data, height from the ground, and human body posture angle, determine a first risk factor, a second risk factor, a third risk factor, and a fourth risk factor;

[0081] The first risk coefficient, the second risk coefficient, the third risk coefficient and the fourth risk coefficient are summed, and the summation result is compared with a preset risk threshold to obtain a height fall risk prediction result.

[0082] Specifically, in this embodiment, the first risk coefficient, the second risk coefficient, the third risk coefficient and the fourth risk coefficient are obtained through the risk assessment table, and then the four risk coefficients are added to obtain the high fall risk coefficient of the student's current behavior in the dangerous area. The summation result is compared with the preset risk threshold to obtain the high fall risk prediction result. The preset risk threshold is specifically shown in Table 4 below:

[0083] Table 4 Risk threshold comparison table

[0084]

[0085] In this embodiment, after obtaining the high-fall risk prediction result, the corresponding warning signal can be selected to be output to the campus security system for risk reminder; in other embodiments, other risk reminder methods can also be used, such as broadcasting reminders through a broadcasting system, etc.

[0086] Example 2

[0087] See also Figure 4 As shown, the present invention also provides a campus high-fall risk prediction system based on visual monitoring, which is used in any one of the campus high-fall risk prediction methods based on visual monitoring described above, and the system includes:

[0088] The area division module 100 is used to divide the campus public area into areas according to the preset dangerous area division rules to obtain multiple dangerous areas with different risk weights, and each dangerous area is set up with a risk assessment table corresponding to the risk weight;

[0089] The feature extraction module 200 is used to obtain a behavior video image, extract a three-dimensional joint point graph containing multiple posture feature key points from the behavior video image, and input the three-dimensional joint point graph into a preset posture feature extraction model to output multiple posture feature information, wherein the multiple posture feature information includes human body posture angle, height from the ground, human body height data and position coordinate information;

[0090] The risk prediction module 300 is used to determine the dangerous area where the student is located based on the position coordinate information, use the risk assessment table of the dangerous area and multiple posture feature information to make one-to-one judgments to obtain multiple risk coefficients, and use the multiple risk coefficients to compare with preset risk thresholds to obtain a high-fall risk prediction result.

[0091] It should be noted that the modules in the system of Example 2 correspond to the steps in the method of Example 1. The steps in the method of Example 1 have been described in detail in Example 1, and the contents of the modules in the system will not be described in detail in this Example 2.

[0092] Example 3

[0093] See also Figure 5 As shown, this embodiment further provides a computer device, including a system memory 1005 and a processor 1001, wherein the system memory 1005 stores a computer program, and the processor 1001 implements the steps of any of the above methods when executing the computer program.

[0094] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, the processor 1001 implements the functions of each module / unit in the above system / device embodiments when executing the computer program.

[0095] Specifically, in this embodiment, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the system memory 1005, and are executed by the processor 1001 to complete the present application. One or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0096] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art will appreciate that this does not constitute a limitation on the terminal device, and may include more or less components than shown in the figure, or combine certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0097] The processor 1001 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0098] The system memory 1005 may be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device. The system memory 1005 may also be a storage device 1004 of the terminal device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the terminal device. Further, the system memory 1005 may also include both the internal storage unit of the terminal device and the storage device 1004. The system memory 1005 is used to store computer programs and other programs and data required by the terminal device. The system memory 1005 may also be used to temporarily store data that has been output or is to be output.

[0099] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0100] Example 4

[0101] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above methods are implemented.

[0102] Among them, the computer readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, system or device, or any combination of the above. More specific examples of computer readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above or any other form of computer readable storage media of numerical value in the art.

[0103] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In an embodiment of the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system, or device.

Claims

1. A method for predicting high-fall risk on campus based on visual monitoring, It is characterized in that Methods include: The campus public area is divided into regions according to the preset dangerous area division rules to obtain multiple dangerous areas with different risk weights, and each dangerous area is set up with a risk assessment table corresponding to the risk weight, specifically: Obtain the height of each dangerous side wall in the campus public area, and determine a first division coefficient that is inversely proportional to the side wall height; Based on the straight-line distance from each dangerous sidewall, a second division coefficient is determined which is inversely proportional to the straight-line distance to the sidewall; The risk weight is determined by using the sum of the first division coefficient and the second division coefficient, and the risk weight is compared with a preset division threshold to divide the campus public area into regions to obtain a plurality of dangerous areas with different risk weights; Each risk assessment table includes: a risk weight and an assessment control item of a first risk coefficient, at least one human height assessment threshold and an assessment control item of a second risk coefficient, at least one ground height assessment threshold and an assessment control item of a third risk coefficient, and at least one human posture angle assessment threshold and an assessment control item of a fourth risk coefficient; Acquire a behavior video image, extract a three-dimensional joint point graph containing multiple posture feature key points from the behavior video image, and input the three-dimensional joint point graph into a preset posture feature extraction model to output multiple posture feature information, wherein the multiple posture feature information includes human body posture angle, height from the ground, human body height data, and position coordinate information; Based on the location coordinate information, the dangerous area where the student is located is determined, and the risk assessment table of the dangerous area and multiple posture feature information are used to make one-to-one judgments to obtain multiple risk coefficients. The multiple risk coefficients are compared with the preset risk thresholds to obtain the high-fall risk prediction results, which are specifically: Based on the location coordinate information, the straight-line distance between the student and each dangerous side wall is obtained, and the straight-line distance to the nearest dangerous side wall is extracted to determine the dangerous area where the student is located, and the risk weight of the dangerous area is obtained; Using the risk assessment table of the dangerous area, based on the risk weight, human body height data, height from the ground, and human body posture angle, determine a first risk factor, a second risk factor, a third risk factor, and a fourth risk factor; The first risk coefficient, the second risk coefficient, the third risk coefficient and the fourth risk coefficient are summed, and the summation result is compared with a preset risk threshold to obtain a height fall risk prediction result.

2. According to claim 1, a method for predicting the risk of falling from height on campus based on visual monitoring, It is characterized in that Obtain a behavior video image, and extract a three-dimensional joint point graph containing multiple posture feature key points from the behavior video image, specifically: Obtain a video image of the student's behavior in the danger zone, and input the behavior video image into a preset first neural network model to output a three-dimensional joint point diagram containing multiple posture feature key points of the student, wherein the multiple posture feature key points include head key points, shoulder key points, elbow key points, hip key points and ankle key points, and each posture feature key point is connected by an edge.

3. According to claim 2, a method for predicting the risk of falling from height on campus based on visual monitoring, It is characterized in that The preset first neural network model adopts an improved YOLO pose model, wherein the improved YOLO pose model includes replacing the convolution block in the Head module of the existing YOLO pose model with a GSConv convolution block; and replacing the Upsample upsampling module in the existing YOLO pose model with a CARAFE upsampling module.

4. According to the method for predicting the risk of falling from height on campus based on visual monitoring according to claim 1, It is characterized in that The preset posture feature extraction model adopts a convolutional neural network model.

5. A campus fall risk prediction system based on visual monitoring, It is characterized in that The system is used in a campus fall risk prediction method based on visual monitoring as described in any one of claims 1 to 4, and the system includes: The area division module is used to divide the campus public area into areas according to the preset dangerous area division rules to obtain multiple dangerous areas with different risk weights, and each dangerous area is equipped with a risk assessment table corresponding to the risk weight; A feature extraction module is used to obtain a behavior video image, extract a three-dimensional joint point graph containing multiple posture feature key points from the behavior video image, and input the three-dimensional joint point graph into a preset posture feature extraction model to output multiple posture feature information, wherein the multiple posture feature information includes human body posture angle, height from the ground, human body height data and position coordinate information; The risk prediction module is used to determine the dangerous area where the student is located based on the location coordinate information, use the risk assessment table of the dangerous area and multiple posture feature information to make one-to-one judgments to obtain multiple risk coefficients, and use the multiple risk coefficients to compare with the preset risk thresholds to obtain the height fall risk prediction result.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Citation Information

Patent Citations

  • Campus safety monitoring and early warning management system based on artificial intelligence

    CN111967400A

  • Smart campus management method and system based on big data

    CN117172989A

  • Real-time attitude monitoring intelligent safety belt system

    CN117746574A

  • Sudden accident detection method based on artificial intelligence technology

    CN117854007A

  • System and method for event prevention and prediction

    US20210035437A1