Artificial intelligence safety anti-bullying system

Abstract

The invention provides an artificial intelligence safety anti-bullying system, and belongs to the technical field of safety prevention and control, and the system comprises a detection module which is used for determining personnel track data and personnel posture data; the analysis module is used for analyzing the personnel trajectory data to determine trajectory label data and analyzing the personnel attitude data to determine attitude label data; the recognition module is used for recognizing abnormal behaviors and determining abnormal behavior data and abnormal behavior labels; the determination module is used for acquiring personnel conversations in a voice acquisition area in real time, determining voice data and determining language violence data; and the execution module is used for sending out a bullying alarm, determining a bullying strategy, executing the bullying strategy and generating a bullying report. According to the method, the detection precision and intelligent level can be improved, the accuracy and timeliness of behavior violence and language violence recognition are improved, the violence early warning precision and coping capacity are improved, timely response and intervention to potential violence events are guaranteed, the risk of violence behaviors is reduced, and comprehensive prevention and control of bullying behaviors are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of security prevention and control, and particularly to an artificial intelligence security anti-bullying system. Background Art

[0002] With the increasing social attention to violent behaviors, especially bullying incidents in schools and public places, traditional monitoring means and protection measures can no longer fully meet the requirements of real-time monitoring, intelligent analysis, and precise response. Most of the existing anti-bullying technologies rely on manual intervention or simple monitoring devices, and these methods often have defects such as slow response and information lag, and the accuracy and efficiency in dealing with violent behaviors are limited.

[0003] In recent years, with the progress of artificial intelligence technology, deep learning, speech recognition, and sensor technology, intelligent protection systems have gradually become a breakthrough to solve this problem. However, the existing intelligent monitoring systems have functional limitations in the application process, unable to comprehensively integrate multiple data sources for efficient analysis and timely response, unable to accurately judge violent behaviors in complex environments, and unable to provide fast and effective response and intervention.

[0004] Therefore, the present invention provides an artificial intelligence security anti-bullying system. Summary of the Invention

[0005] The present invention provides an artificial intelligence security anti-bullying system, which respectively analyzes personnel trajectory data and personnel posture data to determine trajectory label data and posture label data, identifies abnormal behaviors according to the trajectory label data and the posture label data to determine behavior abnormal data and behavior abnormal labels, determines language violence data according to the real-time collected voice data, issues a bullying alarm according to the behavior abnormal data, the behavior abnormal labels, and the language violence data and determines a bullying strategy, executes the bullying strategy and generates a bullying report. It can improve the detection accuracy and intelligent level, improve the accuracy and timeliness of the recognition of behavior violence and language violence, enhance the accuracy and response ability of violence early warning, ensure the immediate response and intervention to potential violent events, reduce the risk of violent behaviors occurring, and enhance the comprehensive prevention and control of bullying behaviors.

[0006] The present invention provides an artificial intelligence security anti-bullying system, including:

[0007] A detection module: detecting the activities of personnel in the trajectory monitoring area in real time to determine personnel trajectory data, and detecting the postures of personnel in the posture monitoring area in real time to determine personnel posture data;

[0008] An analysis module: analyzing the personnel trajectory data to determine trajectory label data, and analyzing the personnel posture data to determine posture label data;

[0009] Recognition module: Based on trajectory label data and posture label data, recognize abnormal behaviors, and determine behavior abnormal data and behavior abnormal labels;

[0010] Determination module: Collect the conversations of people in the voice collection area in real time, determine voice data, and determine language violence data based on the voice data;

[0011] Execution module: Send a bullying alarm based on behavior abnormal data, behavior abnormal labels, and language violence data, determine a bullying strategy, execute the bullying strategy, and generate a bullying report.

[0012] According to the artificial intelligence security anti-bullying system provided by the present invention, the artificial intelligence security anti-bullying system is integrated on an anti-bullying product, and the anti-bullying product is installed in any area with bullying risks.

[0013] According to the artificial intelligence security anti-bullying system provided by the present invention, the detection module includes:

[0014] Detection range unit: Obtain the preset trajectory detection range of the personnel movement trajectory radar of the artificial intelligence security anti-bullying system, and at the same time, obtain the preset posture detection range of the personnel posture radar of the artificial intelligence security anti-bullying system;

[0015] Installation position unit: Based on the scene anti-bullying requirements, preset trajectory detection range, and preset posture detection range of the bullying risk area, determine the installation position of the anti-bullying product;

[0016] Monitoring area unit: Based on the installation position of the anti-bullying product, determine the trajectory monitoring area and the posture monitoring area;

[0017] Sub-personnel trajectory data unit: Based on the sub-personnel trajectory data of each person in the trajectory monitoring area at the installation position of the anti-bullying product within the trajectory detection time period detected by the personnel movement trajectory radar in real time;

[0018] Personnel trajectory data unit: Determine the personnel trajectory data based on the sub-personnel trajectory data of all personnel in the trajectory monitoring area;

[0019] Sub-personnel posture data unit: Based on the sub-personnel posture data of each person in the posture monitoring area at the installation position of the anti-bullying product at the current time detected by the personnel posture radar in real time;

[0020] Personnel posture data unit: Determine the personnel posture data based on the sub-personnel posture data of all personnel in the posture monitoring area.

[0021] According to the artificial intelligence security anti-bullying system provided by the present invention, the analysis module includes:

[0022] Trajectory Feature Vector Unit: Preprocesses the personnel trajectory data, and determines the trajectory feature vector of each person at each time point within the trajectory detection time period in the trajectory monitoring area based on the sub-personnel trajectory data of each person in the preprocessed personnel trajectory data;

[0023] Tr ij =(x ij ,y ij ,v ij ,a ij ,κ ij ,μd ij ,σd ij ,dc ij );

[0024] Among them, Tr ij represents the trajectory feature vector of the i-th person in the trajectory monitoring area at the j-th time point within the trajectory detection time period, and x ij , y ij , v ij , a ij , κ ij , μd ij , σd ij , dc ij respectively represent the abscissa, ordinate, speed, acceleration, trajectory curvature, average value of personnel distance, standard deviation of personnel distance, and rate of change of distance of the i-th person in the trajectory monitoring area at the j-th time point within the trajectory detection time period;

[0025] Trajectory Label Unit: Determines the trajectory label of each person based on the trajectory feature vector of each person at each time point within the trajectory detection time period in the trajectory monitoring area;

[0026] Trajectory Label Data Unit: Determines the trajectory label data based on the trajectory labels of all personnel in the trajectory monitoring area.

[0027] According to the artificial intelligence security anti-bullying system provided by the present invention, the trajectory label unit includes:

[0028]

[0029] Among them, LTA i represents the trajectory label of the i-th person in the trajectory monitoring area, TN represents the normal trajectory evaluation threshold, TA represents the abnormal trajectory evaluation threshold, A i represents the trajectory evaluation value of the i-th person in the trajectory monitoring area, m i represents the weight of the i-th person in the trajectory monitoring area, j1 represents the start time point of the trajectory detection time period, jN represents the end time point of the trajectory detection time period, λ arepresents the acceleration adjustment coefficient, r represents the curvature sensitivity coefficient, λ d represents the distance change rate adjustment coefficient, E(x ij ,y ij ) represents the environmental potential energy function between the abscissa, ordinate of the i-th person in the trajectory monitoring area at the j-th time point during the trajectory detection time period and the obstacles in the trajectory monitoring area, represents the Laplace operator, λ E represents the environmental adjustment coefficient, Ad ij represents the distance evaluation value of the i-th person in the trajectory monitoring area, N1 represents the number of people in the trajectory monitoring area, represents the distance between the i-th person and the k-th person in the trajectory monitoring area at the j-th time point during the trajectory detection time period, e represents the base of the natural logarithm, and π represents the pi.

[0030] According to the artificial intelligence security anti-bullying system provided by the present invention, the analysis module further includes:

[0031] Posture feature vector unit: preprocess the personnel posture data, extract features from the sub-personnel posture data of each person in the preprocessed personnel posture data, and determine the posture feature vector of each person;

[0032] Attack feature vector and defense feature vector unit: extract features from the attack postures of each attack label in the posture template library to determine the attack feature vector, and extract features from the defense postures of each defense label in the posture template library to determine the defense feature vector;

[0033] Posture label unit: determine the posture label of each person based on the posture feature vector of the sub-personnel posture data of each person in the personnel posture data, the attack feature vectors of all attack postures, and the defense feature vectors of all defense postures;

[0034]

[0035] Among them, LAA p represents the posture label of the p-th person in the posture monitoring area, attack posture a represents that the posture label of the person is an attack posture with the attack label a, defense posture b represents that the posture label of the person is a defense posture with the defense label b, v p represents the posture feature vector of the sub-personnel posture data of the p-th person in the posture monitoring area, vA a represents the attack feature vector of the attack posture with the attack label a in the posture template library, vD bDenote the defense feature vector of the defense posture with defense label b in the posture template library. STA represents the attack similarity threshold, STD represents the defense similarity threshold, and s(v p , vA a ) represents the similarity function between the posture feature vector of the sub-person posture data of the p-th person in the posture monitoring area and the attack feature vector of the attack posture with attack label a in the posture template library. s(v p , vD b represents the similarity function between the posture feature vector of the sub-person posture data of the p-th person in the posture monitoring area and the defense feature vector of the defense posture with defense label b in the posture template library;

[0036] Posture label data unit: Determine the posture label data based on the posture labels of all persons in the posture monitoring area.

[0037] According to the artificial intelligence security anti-bullying system provided by the present invention, the recognition module includes:

[0038] High probability violence unit: If there is any person's trajectory label as an abnormal trajectory in the trajectory label data and there is any person's posture label as an attack posture in the posture label data, determine the behavior abnormal data based on the sub-person trajectory data of all persons with trajectory labels as abnormal trajectories in the trajectory label data and the sub-person posture data of all persons with posture labels as attack postures in the posture label data, and determine the behavior abnormal label as high probability violence;

[0039] First possible violence unit: If there is any person's trajectory label as a suspicious trajectory in the trajectory label data and there is any person's posture label as an attack posture in the posture label data, determine the behavior abnormal data based on the sub-person trajectory data of all persons with trajectory labels as suspicious trajectories in the trajectory label data and the sub-person posture data of all persons with posture labels as attack postures in the posture label data, and determine the behavior abnormal label as possible violence;

[0040] Second possible violence unit: If there is any person's trajectory label as an abnormal trajectory in the trajectory label data and there is any person's posture label as a defense posture in the posture label data, determine the behavior abnormal data based on the sub-person trajectory data of all persons with trajectory labels as abnormal trajectories in the trajectory label data and the sub-person posture data of all persons with posture labels as defense postures in the posture label data, and determine the behavior abnormal label as potential violence;

[0041] Normal unit: Otherwise, determine that the behavior abnormal data is empty and determine the behavior abnormal label as normal.

[0042] According to the artificial intelligence security anti-bullying system provided by the present invention, the determination module includes:

[0043] Voice collection unit: Determine voice data based on all personnel conversations within the voice collection time period collected in real time by a hidden microphone; Recognition unit: Convert the voice data to determine text data; Analyze the text data based on a big data language model to identify multiple sensitive words;

[0044] Verbal violence data unit: Determine verbal violence data based on all sensitive words identified by the big data language model.

[0045] Compared with the prior art, the beneficial effects of this application are as follows:

[0046] Analyze personnel trajectory data and personnel posture data respectively to determine trajectory label data and posture label data. According to the trajectory label data and the posture label data, identify abnormal behaviors to determine behavior anomaly data and behavior anomaly labels. Determine verbal violence data based on the voice data collected in real time. Issue a bullying alert according to the behavior anomaly data, behavior anomaly labels, and verbal violence data and determine a bullying strategy. Execute the bullying strategy and generate a bullying report. It can improve the accuracy and intelligence level of detection, improve the accuracy and timeliness of the identification of physical violence and verbal violence, enhance the accuracy of violence warning and response capabilities, ensure instant response and intervention for potential violent incidents, reduce the risk of violent behaviors occurring, and enhance the comprehensive prevention and control of bullying behaviors. Brief Description of the Drawings

[0047] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0048] Figure 1 It is a schematic structural diagram of an artificial intelligence security anti-bullying system provided by an embodiment of the present invention. Detailed Embodiments

[0049] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention with reference to the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0050] Embodiment 1:

[0051] The embodiment of the present invention provides an artificial intelligence security anti-bullying system, as Figure 1 shown, including:

[0052] Detection module: Detect the activities of personnel in the trajectory monitoring area in real time to determine personnel trajectory data, and detect the postures of personnel in the posture monitoring area in real time to determine personnel posture data;

[0053] Analysis module: Analyze the personnel trajectory data to determine trajectory label data, and analyze the personnel posture data to determine posture label data;

[0054] Recognition module: Based on the trajectory label data and the posture label data, recognize abnormal behaviors, and determine behavior anomaly data and behavior anomaly labels;

[0055] Determination module: Collect the conversations of personnel in the voice collection area in real time to determine voice data, and determine language violence data based on the voice data;

[0056] Execution module: Issue a bullying alarm based on the behavior anomaly data, behavior anomaly labels, and language violence data, determine a bullying strategy, execute the bullying strategy, and generate a bullying report.

[0057] In this embodiment, the activities of personnel in the trajectory monitoring area and the posture monitoring area are monitored in real time. Through the integrated sensors (personnel movement trajectory radar and personnel posture radar), the system can track the position changes of each personnel within the trajectory monitoring area and the trajectory detection time period to generate personnel trajectory data; at the same time, within the posture monitoring area, the system monitors the posture movements of each personnel to generate personnel posture data.

[0058] In this embodiment, based on the personnel trajectory data obtained by the detection module, the system performs data processing and analysis to generate trajectory label data to determine the behavior of personnel. At the same time, the personnel posture data is analyzed to generate posture label data.

[0059] In this embodiment, according to the trajectory label data and the posture label data, the system performs behavior recognition to determine behavior anomaly data and behavior anomaly labels. The behavior anomaly labels include highly likely violence, likely violence, potential violence, and normal.

[0060] In this embodiment, the voice data in the voice collection area is collected in real time, and the system collects the conversation content through a hidden microphone. According to the analysis of the big data language model, it is identified whether there is language violence data (such as insults, threatening words).

[0061] In this embodiment, when the system detects behavior anomaly data or language violence data, the system automatically issues a bullying alarm and determines the corresponding bullying strategy. For example, calling the police, taking security measures, etc. At the same time, the system will generate a bullying report to record the details of the violent incident and the response measures.

[0062] In this embodiment, when keywords such as "help", "fight", and "abuse" are recognized, voice intervention can be performed, alarm information can be uploaded, and vibration reminder can be given at the same time. Relevant personnel can communicate with the scene and stop the situation by real-time shouting. At the same time, when bullying or other dangerous situations occur in the dormitory, emergency alarm can be made through voice.

[0063] In this embodiment, the artificial intelligence security anti-bullying system further includes: an environment module: collecting environmental information in the environmental monitoring area in real time to determine environmental data, analyzing the environmental data to identify smoke leakage and giving environmental warnings, and associating potential fire risks.

[0064] The beneficial effects of the above technical solutions: Analyze personnel trajectory data and personnel posture data respectively to determine trajectory label data and posture label data. According to the trajectory label data and the posture label data, identify abnormal behaviors to determine behavior abnormal data and behavior abnormal labels. Determine language violence data according to the real-time collected voice data. Send a bullying alarm according to the behavior abnormal data, behavior abnormal labels and language violence data and determine a bullying strategy, execute the bullying strategy and generate a bullying report. It can improve the accuracy and intelligence level of detection, improve the accuracy and timeliness of behavior violence and language violence recognition, enhance the accuracy and response ability of violence warning, ensure the instant response and intervention to potential violence events, reduce the risk of violence behavior occurrence, and enhance the comprehensive prevention and control of bullying behavior.

[0065] Embodiment 2:

[0066] The embodiment of the present invention provides an artificial intelligence security anti-bullying system. The artificial intelligence security anti-bullying system is integrated on an anti-bullying product, and the anti-bullying product is installed in any area with bullying hidden dangers.

[0067] In this embodiment, the anti-bullying product can be installed in any place considered to be a high-risk area for bullying, such as school corridors, classrooms, public transportation, sports fields, etc. By covering these areas, the anti-bullying product monitors possible violent or bullying behaviors all day long.

[0068] In this embodiment, the area with bullying hidden dangers refers to the area where bullying behaviors may occur, usually places with dense population and weak supervision.

[0069] The beneficial effects of the above technical solutions: It can achieve real-time monitoring and instant response, automatically identify abnormal behaviors and take countermeasures, and reduce human negligence.

[0070] Embodiment 3:

[0071] The embodiment of the present invention provides an artificial intelligence security anti-bullying system, a detection module, including:

[0072] Detection range unit: Obtain the preset trajectory detection range of the personnel movement trajectory radar of the artificial intelligence security anti-bullying system. Meanwhile, obtain the preset posture detection range of the personnel posture radar of the artificial intelligence security anti-bullying system;

[0073] Installation location unit: Based on the scene anti-bullying requirements of the bullying risk area, the preset trajectory detection range, and the preset posture detection range, determine the installation location of the anti-bullying product;

[0074] Monitoring area unit: Based on the installation location of the anti-bullying product, determine the trajectory monitoring area and the posture monitoring area;

[0075] Sub-personnel trajectory data unit: Based on the sub-personnel trajectory data of each person within the trajectory monitoring area at the installation location of the anti-bullying product during the trajectory detection time period detected in real-time by the personnel movement trajectory radar;

[0076] Personnel trajectory data unit: Determine the personnel trajectory data based on the sub-personnel trajectory data of all personnel within the trajectory monitoring area;

[0077] Sub-personnel posture data unit: Based on the sub-personnel posture data of each person within the posture monitoring area at the installation location of the anti-bullying product at the current time detected in real-time by the personnel posture radar;

[0078] Personnel posture data unit: Determine the personnel posture data based on the sub-personnel posture data of all personnel within the posture monitoring area.

[0079] In this embodiment, within the preset trajectory detection range of the personnel movement trajectory radar, the radar can capture the movement of personnel.

[0080] In this embodiment, the preset posture detection range of the personnel posture radar determines the monitoring area of the posture recognition radar, and can obtain the postures and movements of personnel in real-time.

[0081] In this embodiment, according to the specific requirements of the bullying risk area, combined with the trajectory detection range and the posture detection range, determine the installation location of the anti-bullying product to ensure coverage of key areas, and install at least one or more anti-bullying products in the bullying risk area.

[0082] In this embodiment, according to the installation location of the anti-bullying product, the system determines the trajectory monitoring area and the posture monitoring area. These areas are the actual monitoring scopes and can comprehensively track the actions and postures of personnel.

[0083] In this embodiment, through the personnel movement trajectory radar, collect the sub-personnel trajectory data of each person entering the trajectory monitoring area, including time, location, etc.

[0084] In this embodiment, the trajectory data of all personnel in the entire monitoring area is obtained by summarizing the sub-personnel trajectory data of all personnel.

[0085] In this embodiment, through a personnel attitude radar, the sub-personnel attitude data of each person entering the area at each current time in the attitude monitoring area is obtained.

[0086] In this embodiment, the sub-personnel attitude data of all personnel entering the attitude monitoring area is summarized to form complete personnel attitude data.

[0087] The beneficial effects of the above technical solutions: Detect the activities of personnel in the trajectory monitoring area in real time to determine the personnel trajectory data, and detect the personnel attitudes in the attitude monitoring area in real time to determine the personnel attitude data, which can achieve highly accurate detection of violent behaviors and provide high-quality data basis for determining the trajectory label data and attitude label data.

[0088] Embodiment 4:

[0089] The embodiment of the present invention provides an artificial intelligence security anti-bullying system, an analysis module, including:

[0090] Trajectory feature vector unit: Preprocess the personnel trajectory data, and based on the sub-personnel trajectory data of each person in the preprocessed personnel trajectory data, determine the trajectory feature vector of each person in the trajectory monitoring area at each time point within the trajectory detection time period;

[0091] Tr ij =(x ij ,y ij ,v ij ,a ij ,κ ij ,μd ij ,σd ij ,dc ij );

[0092] Among them, Tr ij represents the trajectory feature vector of the i-th person in the trajectory monitoring area at the j-th time point within the trajectory detection time period, and x ij , y ij , v ij , a ij , κ ij , μd ij , σd ij , dc ij respectively represent the abscissa, ordinate, speed, acceleration, trajectory curvature, average value of personnel distance, standard deviation of personnel distance, and rate of change of distance of the i-th person in the trajectory monitoring area at the j-th time point within the trajectory detection time period;

[0093] Trajectory Label Unit: Determine the trajectory label of each person based on the trajectory feature vectors of each person at each time point within the trajectory monitoring period in the trajectory monitoring area.

[0094] Trajectory Label Data Unit: Determine the trajectory label data based on the trajectory labels of all persons in the trajectory monitoring area.

[0095] In this embodiment, preprocess the personnel trajectory data, and through removing noise or redundant data, sort out more accurate trajectory data. Based on the preprocessed data, for the sub-personnel trajectory data of each individual (i.e., the action trajectory of each individual entering the monitoring area), generate trajectory feature vectors. The trajectory feature vectors include the specific position, speed, acceleration, trajectory curvature, average value of the distances of the personnel, standard deviation of the distances of the personnel, and rate of change of distance of the personnel at each time point.

[0096] In this embodiment, based on the trajectory feature vectors, at each time point, the system analyzes the movement patterns of the personnel and generates a trajectory label for each person. The trajectory labels are used to classify the behaviors of the personnel to quickly identify abnormal behaviors.

[0097] In this embodiment, generate the trajectory label data by summarizing the trajectory labels of all persons in the trajectory monitoring area.

[0098] Beneficial effects of the above technical solution: Analyzing the personnel trajectory data to determine the trajectory label data can improve the intelligence and accuracy of detection, provide strong data support for realizing timely early warning and rapid response for determining violent data, and ensure the effective prevention and control of potential bullying behaviors.

[0099] Embodiment 5:

[0100] The embodiment of the present invention provides an artificial intelligence security anti-bullying system, and the trajectory label unit includes:

[0101]

[0102] Among them, LTA i represents the trajectory label of the i-th person in the trajectory monitoring area, TN represents the normal trajectory evaluation threshold, TA represents the abnormal trajectory evaluation threshold, A i represents the trajectory evaluation value of the i-th person in the trajectory monitoring area, m i represents the weight of the i-th person in the trajectory monitoring area, j1 represents the start time point of the trajectory monitoring period, jN represents the end time point of the trajectory monitoring period, λ a represents the acceleration adjustment coefficient, r represents the curvature sensitivity coefficient, λ d represents the rate of change of distance adjustment coefficient, E(x ij ,y ijdenotes the environmental potential energy function between the abscissa and ordinate of the \(i\)-th person at the \(j\)-th time point within the trajectory detection time period and the obstacles within the trajectory monitoring area. denotes the Laplace operator, \(\lambda\) E denotes the environmental regulation coefficient, \(Ad\) ij denotes the distance evaluation value of the \(i\)-th person within the trajectory monitoring area, and \(N1\) denotes the number of people within the trajectory monitoring area. denotes the distance between the \(i\)-th person and the \(k\)-th person at the \(j\)-th time point within the trajectory detection time period, \(e\) denotes the base of the natural logarithm, and \(\pi\) denotes the ratio of the circumference of a circle to its diameter.

[0103] In this embodiment, means that the distance between the \(i\)-th person and the \(k\)-th person is more likely to be around \(\mu_d\) ij and as the distance deviates from the average value increases, the probability drops rapidly. That is to say, if the distance between two people is greater, the probability of bullying behavior occurring is smaller.

[0104] In this embodiment, denotes the fractional derivative of the curvature \(\kappa\) ij i.e., the semi-derivative, which captures the non-linear dynamic characteristics of the curvature changing over time. Through the semi-derivative, the complexity of the local changes of the trajectory can be better described.

[0105] Beneficial effects of the above technical solution: Based on the trajectory feature vectors of each person within the trajectory monitoring area at each time point within the trajectory detection time period, determining the trajectory labels of each person can provide a data basis for determining the trajectory label data, improve the intelligence and accuracy of detection, and achieve effective prevention and control of potential bullying behavior.

[0106] Embodiment 6:

[0107] The embodiment of the present invention provides an artificial intelligence security anti-bullying system, and the analysis module further includes:

[0108] Posture feature vector unit: Preprocess the personnel posture data, extract features from the sub-personnel posture data of each person in the preprocessed personnel posture data, and determine the posture feature vector of each person;

[0109] Attack feature vector and defense feature vector unit: Extract features from the attack postures of each attack label in the posture template library to determine the attack feature vectors, and extract features from the defense postures of each defense label in the posture template library to determine the defense feature vectors;

[0110] Posture Label Unit: Determine the posture label of each person based on the posture feature vector of the sub-person posture data of each person in the personnel posture data, the attack feature vector of all attack postures, and the defense feature vector of all defense postures;

[0111]

[0112] Among them, LAA p represents the posture label of the p-th person in the posture monitoring area, and the attack posture a represents that the posture label of the person is an attack posture with an attack label of a, and the defense posture b represents that the posture label of the person is a defense posture with a defense label of b, v p represents the posture feature vector of the sub-person posture data of the p-th person in the posture monitoring area, vA a represents the attack feature vector of the attack posture with an attack label of a in the posture template library, vD b represents the defense feature vector of the defense posture with a defense label of b in the posture template library, STA represents the attack similarity threshold, STD represents the defense similarity threshold, s(v p , vA a ) represents the similarity function between the posture feature vector of the sub-person posture data of the p-th person in the posture monitoring area and the attack feature vector of the attack posture with an attack label of a in the posture template library, s(v p , vD b represents the similarity function between the posture feature vector of the sub-person posture data of the p-th person in the posture monitoring area and the defense feature vector of the defense posture with a defense label of b in the posture template library;

[0113] Posture Label Data Unit: Determine the posture label data based on the posture labels of all the personnel in the posture monitoring area.

[0114] In this embodiment, the personnel posture data is preprocessed, including removing noise, smoothing the data, etc., to ensure the quality of the data, and key features such as joint angles, limb lengths, body center of gravity positions, etc. are extracted from the sub-person posture data of each person entering the monitoring area, thereby generating posture feature vectors. Each feature vector represents the action state of the corresponding person at the current moment, for example: standing, defending, punching, etc.

[0115] In this embodiment, the construction process of the posture template library can be as follows: Data collection: With the help of high-precision motion capture devices (such as optical motion capture systems, inertial motion capture systems, etc.), professional actors or volunteers are asked to simulate common attack postures, such as raising the arm to attack, punching, kicking, etc. Multiple collections are carried out under different lighting conditions, angles, and motion amplitudes to obtain rich and diverse posture data; Data preprocessing: The collected original posture data is cleaned to remove noise and outliers. A data smoothing algorithm (such as Kalman filtering) is used to process the data to make the posture data more continuous and stable. At the same time, the data is normalized to eliminate the scale differences brought by different acquisition devices and environmental factors; 3D model creation: Based on the preprocessed posture data, 3D modeling software (such as Maya, Blender, etc.) is used to create 3D models of each attack posture. Key feature points and joint information are added to each model for subsequent matching calculations; Template library storage: The created 3D models are stored in the template library, and an attack label is added to each template.

[0116] In this embodiment, the attack posture features corresponding to each attack label, such as punching, pushing, etc., are extracted from the posture template library to generate attack feature vectors. Similarly, the defense postures of each defense label (such as defending, dodging, etc.) are feature-extracted to generate defense feature vectors.

[0117] In this embodiment, based on the posture feature vectors of each person, comparisons are made with all the attack feature vectors and defense feature vectors to determine the posture labels of each person. These labels help the system identify whether the current behavior of the person is an attack, defense, or other normal postures, so as to judge in real time whether there is violent behavior.

[0118] In this embodiment, the posture labels of all the people in the posture monitoring area are summarized to generate posture label data.

[0119] The beneficial effects of the above technical solution: By analyzing the personnel posture data to determine the posture label data, real-time recognition of violent postures and defense postures can be achieved, improving the accuracy and intelligence level of detection.

[0120] Embodiment 7:

[0121] The embodiment of the present invention provides an artificial intelligence security anti-bullying system, and an identification module, including:

[0122] High Probability Violence Unit: If there is any person's trajectory label in the trajectory label data as an abnormal trajectory, and there is any person's posture label in the posture label data as an attack posture, determine the behavior abnormal data based on the sub-person trajectory data of all persons with abnormal trajectory labels in the trajectory label data and the sub-person posture data of all persons with attack posture labels in the posture label data, and determine the behavior abnormal label as high probability violence;

[0123] First Probability Violence Unit: If there is any person's trajectory label in the trajectory label data as a suspicious trajectory, and there is any person's posture label in the posture label data as an attack posture, determine the behavior abnormal data based on the sub-person trajectory data of all persons with suspicious trajectory labels in the trajectory label data and the sub-person posture data of all persons with attack posture labels in the posture label data, and determine the behavior abnormal label as probable violence;

[0124] Second Probability Violence Unit: If there is any person's trajectory label in the trajectory label data as an abnormal trajectory, and there is any person's posture label in the posture label data as a defensive posture, determine the behavior abnormal data based on the sub-person trajectory data of all persons with abnormal trajectory labels in the trajectory label data and the sub-person posture data of all persons with defensive posture labels in the posture label data, and determine the behavior abnormal label as potential violence;

[0125] Normal Unit: Otherwise, determine the behavior abnormal data as empty and determine the behavior abnormal label as normal.

[0126] In this embodiment, when there is any person's trajectory label in the trajectory label data as an abnormal trajectory, and there is any person's posture label in the posture label data as an attack posture, the system will combine the sub-person trajectory data of all persons with abnormal trajectories in the trajectory label data and the sub-person posture data of all persons with attack postures in the posture label data to generate behavior abnormal data, and label the behavior violence as high probability violence.

[0127] In this embodiment, if a person's trajectory label in the trajectory label data is a suspicious trajectory, and there is any person's posture label in the posture label data as an attack posture, the system will combine these data to generate behavior abnormal data and label it as probable violence. The probable violence label indicates a possible violent behavior, but it is not as clear as the high probability violence label.

[0128] In this embodiment, if the trajectory label of a certain person in the trajectory label data is an abnormal trajectory, and there is a label of a defensive posture in the posture label data, the system will generate a potential violence label. This means that although the person's behavior shows a defensive posture, it may still indicate the potential occurrence of a violent conflict (defense is usually a reaction to an attack).

[0129] In this embodiment, if none of the above conditions are met, the system determines that the abnormal behavior data is empty and gives a normal label, indicating that no signs of violent behavior are detected.

[0130] Beneficial effects of the above technical solution: Based on the trajectory label data and the posture label data, identify abnormal behaviors to determine violent data and abnormal behavior labels, which can timely identify potential violent behaviors, improve the accuracy and response ability of violent early warnings, and provide a more predictive and effective bullying prevention strategy for security prevention and control.

[0131] Embodiment 8:

[0132] The embodiment of the present invention provides an artificial intelligence security anti-bullying system, a determination module, including:

[0133] Voice acquisition unit: Determine voice data based on all personnel conversations within the voice acquisition time period collected in real time by a hidden microphone;

[0134] Recognition unit: Convert the voice data to determine text data; Analyze the text data based on a big data language model to identify multiple sensitive words;

[0135] Verbal violence data unit: Determine verbal violence data based on all sensitive words identified by the big data language model. In this embodiment, a hidden microphone is used to collect voice data in real time, continuously collecting the conversation content of all personnel in the area within the voice acquisition time period. The hidden microphone has the characteristics of concealment and high fidelity, ensuring that it will not interfere with the normal behavior of personnel or attract attention, and can accurately collect the sound information in the area without disturbing the environment.

[0136] In this embodiment, after obtaining the voice data, first convert and process the collected voice content into text. Then, in combination with advanced big data language models, these big data models can identify signs of verbal violence, such as violent words like insults, threats, and abuse, by learning a large amount of text data.

[0137] Beneficial effects of the above technical solution: Real-time collection of personnel conversations in the voice acquisition area to determine voice data, and determination of verbal violence data based on the voice data, which can efficiently monitor and identify verbal violence, improve the detection accuracy and timeliness of verbal violence, and enhance the comprehensive prevention and control ability of bullying behavior.

[0138] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative work.

[0139] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. The artificial intelligence security anti-bullying system is characterized by: include: Detection module: detects the activities of people in the trajectory monitoring area in real time to determine the trajectory data of people, and detects the posture of people in the posture monitoring area in real time to determine the posture data of people; Analysis module: analyzes personnel trajectory data to determine trajectory label data, and analyzes personnel posture data to determine posture label data; Identification module: Identify abnormal behaviors based on trajectory label data and posture label data, and determine abnormal behavior data and abnormal behavior labels; Determination module: collects conversations between people in the voice collection area in real time, determines voice data, and determines language violence data based on the voice data; Execution module: issues bullying alerts and determines bullying strategies based on abnormal behavior data, abnormal behavior labels, and verbal violence data, executes bullying strategies, and generates bullying reports.

2. The artificial intelligence safety anti-bullying system according to claim 1 is characterized in that: The artificial intelligence security anti-bullying system is integrated into the anti-bullying products, and the anti-bullying products are installed in any areas with potential bullying risks.

3. The artificial intelligence safety anti-bullying system according to claim 2 is characterized in that: Detection module, including: Detection range unit: obtains the preset trajectory detection range of the personnel movement trajectory radar of the artificial intelligence security anti-bullying system, and at the same time, obtains the preset posture detection range of the personnel posture radar of the artificial intelligence security anti-bullying system; Installation location unit: Determines the installation location of the anti-bullying product based on the anti-bullying needs of the scenario in the potential bullying area, the preset trajectory detection range, and the preset posture detection range; Monitoring area unit: determines the trajectory monitoring area and posture monitoring area based on the installation location of the anti-bullying product; Sub-personnel trajectory data unit: sub-personnel trajectory data of each person in the trajectory monitoring area where the anti-bullying product is installed within the trajectory detection time period detected in real time by the person movement trajectory radar; Personnel trajectory data unit: determines personnel trajectory data based on sub-personnel trajectory data of all personnel in the trajectory monitoring area; Sub-personnel posture data unit: sub-personnel posture data of each person in the posture monitoring area where the anti-bullying product is installed at the current time detected in real time by the personnel posture radar; Personnel posture data unit: determines personnel posture data based on sub-personnel posture data of all personnel in the posture monitoring area.

4. The artificial intelligence security anti-bullying system according to claim 1 is characterized in that: Analysis modules, including: Trajectory feature vector unit: pre-processes the personnel trajectory data, and determines the trajectory feature vector of each person in the trajectory monitoring area at each time point in the trajectory detection time period based on the sub-personnel trajectory data of each person in the pre-processed personnel trajectory data; Tr ij =(x ij ,y ij ,v ij ,a ij ,k ij ,μd ij ,σd ij ,dc ij ); Among them, Tr ij represents the trajectory feature vector of the i-th person in the trajectory monitoring area at the j-th time point in the trajectory detection time period, x ij ,y ij 、v ij 、a ij , κ ij , μd ij ,σd ij 、dc ij Respectively represent the abscissa, ordinate, speed, acceleration, trajectory curvature, average value of personnel distance, standard deviation of personnel distance, and distance change rate of the i-th person in the trajectory monitoring area at the j-th time point within the trajectory detection time period; Trajectory label unit: determines the trajectory label of each person based on the trajectory feature vector of each person in the trajectory monitoring area at each time point within the trajectory detection time period; Trajectory label data unit: determines trajectory label data based on the trajectory labels of all persons in the trajectory monitoring area.

5. The artificial intelligence safety anti-bullying system according to claim 4 is characterized in that: Track labeling unit, including: Among them, LTA i represents the trajectory label of the i-th person in the trajectory monitoring area, TN represents the normal trajectory assessment threshold, TA represents the abnormal trajectory assessment threshold, and A i represents the trajectory evaluation value of the i-th person in the trajectory monitoring area, m i represents the weight of the ith person in the trajectory monitoring area, j1 represents the starting time point of the trajectory detection time period, jN represents the ending time point of the trajectory detection time period, and λ a represents the acceleration adjustment coefficient, r represents the curvature sensitivity coefficient, λ d Represents the distance change rate adjustment coefficient, E(x ij ,y ij ) represents the environmental potential energy function between the horizontal coordinate and vertical coordinate of the i-th person in the trajectory monitoring area at the j-th time point in the trajectory detection time period and the obstacles in the trajectory monitoring area, represents the Laplace operator, λ E represents the environmental adjustment coefficient, Ad ij represents the distance evaluation value of the i-th person in the trajectory monitoring area, N1 represents the number of people in the trajectory monitoring area, It represents the distance between the i-th person and the k-th person in the trajectory monitoring area at the j-th time point within the trajectory detection time period, e represents the base of the natural logarithm, and π represents pi.

6. The artificial intelligence safety anti-bullying system according to claim 5 is characterized in that: The analysis module also includes: Posture feature vector unit: preprocessing personnel posture data, extracting features from sub-personnel posture data of each person in the preprocessed personnel posture data, and determining the posture feature vector of each person; Attack feature vector and defense feature vector unit: extract features from the attack posture of each attack tag in the posture template library to determine the attack feature vector, extract features from the defense posture of each defense tag in the posture template library to determine the defense feature vector; Posture labeling unit: determining a posture label of each person based on a posture feature vector of the sub-personnel posture data of each person in the personnel posture data, an attack feature vector of all attack postures, and a defense feature vector of all defense postures; Among them, LAA p Indicates the posture label of the pth person in the posture monitoring area, attack posture a Indicates that the person's posture label is attack label a, the attack posture is defense posture b Indicates that the posture label of the person is the defensive posture with the defense label b, v p The posture feature vector representing the posture data of the p-th person in the posture monitoring area, vA a represents the attack feature vector of the attack posture with attack label a in the posture template library, vD b represents the defense feature vector of the defense posture with the defense label b in the posture template library, STA represents the attack similarity threshold, STD represents the defense similarity threshold, s(v p ,vA a ) represents the similarity function between the posture feature vector of the posture data of the p-th person in the posture monitoring area and the attack feature vector of the attack posture with the attack label a in the posture template library, s(v p ,vD b A similarity function representing a posture feature vector of the posture data of a sub-personnel of a p-th person in the posture monitoring area and a defense feature vector of a defense posture with a defense label of b in the posture template library; Posture label data unit: determines posture label data based on the posture labels of all persons in the posture monitoring area.

7. The artificial intelligence safety anti-bullying system according to claim 6, characterized in that: Identification module, including: High-probability violence unit: If any person's trajectory label is an abnormal trajectory in the trajectory label data, and any person's posture label is an attack posture in the posture label data, the behavior abnormality data is determined based on the sub-personnel trajectory data of all persons with abnormal trajectory labels in the trajectory label data and the sub-personnel posture data of all persons with attack posture labels in the posture label data, and the behavior abnormality label is determined as high-probability violence; The first possible violence unit: if any person's trajectory label is a suspicious trajectory in the trajectory label data, and any person's posture label is an attacking posture in the posture label data, the abnormal behavior data is determined based on the sub-personnel trajectory data of all persons with suspicious trajectory labels in the trajectory label data and the sub-personnel posture data of all persons with attacking posture labels in the posture label data, and the abnormal behavior label is determined as possible violence; Second potential violence unit: If any person's trajectory label is an abnormal trajectory in the trajectory label data, and any person's posture label is a defensive posture in the posture label data, the abnormal behavior data is determined based on the sub-personnel trajectory data of all persons with abnormal trajectory labels in the trajectory label data and the sub-personnel posture data of all persons with defensive posture labels in the posture label data, and the abnormal behavior label is determined as potential violence; Normal unit: Otherwise, the behavior anomaly data is determined to be empty, and the behavior anomaly label is determined to be normal.

8. The artificial intelligence safety anti-bullying system according to claim 1, characterized in that: Identify modules, including: Voice collection unit: determines voice data based on all conversations between people within the voice collection time period collected in real time by the hidden microphone; recognition unit: converts voice data to determine text data; analyzes text data based on the big data language model to identify multiple sensitive words; Language violence data unit: determines language violence data based on all sensitive words identified by the big data language model.

Images