Campus safety risk early warning method based on wireless signals and related equipment

Abstract

The invention discloses a campus safety risk early warning method based on wireless signals and related equipment, the problem of visual dead angles existing in monitoring equipment is solved by comprehensively covering the wireless signals at all corners of a campus, and potential safety risks can be sensed and predicted in real time by continuously capturing and analyzing the wireless signals. The system solves the problems that a traditional video monitoring technology is lack of intelligent identification and analysis capability for student behaviors, cannot automatically identify and warn risk behaviors of students, and cannot perform whole-process monitoring, so that high-risk behaviors and action routes of the students cannot be identified, information is not fused, an emergency plan cannot be made in time, and the safety of the students is influenced. And a campus security event is further generated.

Description

Technical Field

[0001] The present invention relates to the technical field of campus security risk early warning, and specifically relates to a campus security risk early warning method and related devices based on wireless signals. Background Art

[0003] Campus security prevention and control includes the identification of low-risk behaviors, such as people fighting, people falling, people staying, entering an area, running in the corridor, throwing objects from height, etc.; the identification of high-risk behaviors, such as perimeter intrusion, people crowding, falling from height, etc. Campus security prevention and control cannot do without behavior monitoring and behavior identification. Currently, campus monitoring devices mainly rely on traditional video monitoring and lack the ability to intelligently identify students' behaviors. Security personnel need to manually watch the monitoring videos to identify students' risk behaviors, which is not only inefficient but also prone to missed or misjudged due to fatigue or negligence. Moreover, there are often visual dead angles in the layout of campus monitoring devices, resulting in some key areas or corners being unable to be effectively monitored. This increases the possibility of students having dangerous behaviors in these areas and also brings great potential safety hazards to campus security; at the same time, the existing monitoring systems cannot achieve full-process monitoring of students' behaviors, that is, from the start to the end of students' behaviors and the changes during the behavior process cannot be completely recorded and analyzed. This leads to the inability to accurately trace the action routes and behavior trajectories of students when a security incident occurs, thus bringing difficulties to incident handling. There is a lack of an effective information integration mechanism between the campus monitoring system and the emergency plan system, resulting in the inability to trigger the emergency plan in a timely manner after identifying students' risk behaviors, thus delaying the rescue and handling time.

[0004] In summary, traditional video monitoring technology mainly relies on cameras and video recording devices to monitor the situation on campus by recording and playing back videos. However, this technology lacks the ability to intelligently identify and analyze students' behaviors, cannot automatically identify and warn students of risk behaviors. At the same time, it cannot achieve full-process monitoring, resulting in the inability to identify students' high-risk behaviors and action routes, making information not integrated, unable to make an emergency plan in a timely manner, and thus leading to campus security incidents. Summary of the Invention

[0005] Based on the problems raised in the above background art, the purpose of the present invention is to provide a campus security risk early warning method and related devices based on wireless signals, which solve the problems that traditional video monitoring technology lacks the ability to intelligently identify and analyze students' behaviors, cannot automatically identify and warn students of risk behaviors, and at the same time, it cannot achieve full-process monitoring, resulting in the inability to identify students' high-risk behaviors and action routes, making information not integrated, unable to make an emergency plan in a timely manner, and thus leading to campus security incidents.

[0006] The present invention is achieved through the following technical solutions:

[0007] The first aspect of the present invention provides a campus security risk early warning method based on wireless signals, including the following steps:

[0008] Step S1: Obtain the wireless signals of the target person, perform continuous action perception on the wireless signals of the target person, and obtain behavior information;

[0009] Step S2: Based on the behavior information, perform identity authentication on the target person to obtain identity information;

[0010] Step S3: Analyze the identity information to obtain classroom location information, and determine the regular behavior route according to the classroom location information;

[0011] Step S4: Integrate the behavior information, the identity information, the classroom location information, and the regular behavior route into a target person behavior vector set;

[0012] Step S5: Construct a campus security risk analysis model, input the target person behavior vector set into the campus security risk analysis model for analysis, generate the security risk behavior of the target person, and generate a security risk early warning information according to the security risk behavior of the target person.

[0013] In the above technical solution, first, capture the wireless signals related to the target person, where the target person includes those in blind spots of monitoring and areas not covered by monitoring. Perform continuous action perception on the wireless signals of the target person, and judge the behavior information of the target person through the changes in the wireless signals, and then generate behavior information.

[0014] This behavior information can not only be used to judge the behavior actions of the target person, but also can be used for identity recognition based on the behavior actions. In this method, the behavior information is used in combination with the personnel behavior pattern to perform identity authentication on the target person to obtain the identity information about the target person. The identity information in this method refers to the relevant identity information based on human behavior characteristics.

[0015] Match and search from the historical information according to its identity information to determine the campus identity information of the target person, where the campus identity information includes information such as the grade, class, and teaching building where the target person is located. If there is no historical personnel behavior pattern, it will be determined as an outsider, which is also one of the campus risks that need attention.

[0016] Integrate behavioral information, identity information, classroom location information, and regular behavioral routes into a target person's behavioral vector set, which can be input into a security risk analysis model for security risk analysis. Among them, the security risk analysis model can process the information in the target person's behavioral vector set, input the target person's behavioral vector set into the model for analysis, generate the target person's security risk behaviors, and generate security risk warning information based on the security risk behaviors, so as to take preventive measures in a timely manner.

[0017] This method overcomes the visual dead angle problem of monitoring devices through wireless signals that cover all corners of the campus, and by continuously capturing and analyzing wireless signals, it can perceive and predict potential security risks in real time. It solves the problems of traditional video surveillance technology lacking the ability to intelligently identify and analyze students' behaviors, being unable to automatically identify and warn students' risk behaviors. At the same time, its inability to monitor the entire process leads to the failure to identify students' high-risk behaviors and action routes, resulting in information incoherence and the inability to make emergency response plans in a timely manner, thus generating campus security incidents.

[0018] In an alternative embodiment, obtain the wireless signal of the target person, and perform continuous action perception on the wireless signal of the target person, including the following steps:

[0019] Step S11: Receive the campus wireless signal, perform noise correction on the campus wireless signal, and perform target association estimation on the noise-corrected campus wireless signal to generate the wireless signal of the target person;

[0020] Step S12: Extract the signal amplitude change and signal phase change from the wireless signal of the target person, extract limb information from the signal amplitude change and signal phase change, and arrange the extracted limb information in time sequence to generate a time-sequence action signal;

[0021] Step S13: Window the time-sequence action signal to generate a windowed time-sequence action signal;

[0022] Step S14: Extract action segment features from the window of the windowed time-sequence action signal to generate an action segment feature sequence;

[0023] Step S15: Perform continuous feature capture on the action segment feature sequence to obtain an action probability;

[0024] Step S16: Integrate the action probability with the windowed time-sequence action signal to generate behavioral information.

[0025] In an alternative embodiment, extracting limb information from the signal amplitude change and signal phase change includes:

[0026] Convert the signal amplitude change and the signal phase change into the spatial domain to generate a signal feature image;

[0027] Map the limb information feature image into a three-dimensional space to generate a three-dimensional signal feature image;

[0028] Extract the bone node features from the three-dimensional signal feature image, and use the extracted bone node features to form a limb structure to generate limb information.

[0029] In an optional embodiment, identity authentication of a target person based on the behavior information includes the following steps:

[0030] Step S21: Extract walking information from the behavior information, extract gait features from the walking information to obtain a gait signal;

[0031] Step S22: Estimate the normal walking speed of the target person, calculate the current walking speed of the target person according to the gait signal, and calculate the time series difference between the current walking speed and the normal walking speed;

[0032] Step S23: Use the time series difference to correct the gait signal to obtain a corrected gait signal;

[0033] Step S24: Perform identity authentication on the corrected gait signal to obtain identity information.

[0034] In an optional embodiment, constructing a campus security risk analysis model includes:

[0035] An action route simulation layer for performing continuous time series action perception on behavior information and simulating an action route according to the continuous time series action perception;

[0036] An action attention layer for matching the action route simulation result with the normal action route;

[0037] A fusion layer for semantically fusing the behavior information according to the action route matching result;

[0038] A discrimination layer for performing category judgment based on the result of semantic fusion.

[0039] In an optional embodiment, inputting the target person behavior vector set into the campus security risk analysis model for analysis includes the following steps:

[0040] Step S51: Perform semantic enhancement on the target person behavior vector set to generate a semantic feature vector;

[0041] Step S52: Extract the behavior information semantic feature vector from the semantic feature vector, and input the behavior information semantic feature vector into the action route model layer for continuous time-series action perception to obtain continuous time-series action features;

[0042] Step S53: Determine the traveling direction, traveling speed, and traveling state of the target person according to the continuous time-series action features;

[0043] Step S54: Combine the campus map to simulate the action route for the traveling direction and generate a simulated action route;

[0044] Step S55: Input the simulated action route into the action attention layer to match the action route with the regular action route to obtain the action route similarity;

[0045] Step S56: Normalize the action route similarity into an action route probability value, and perform semantic fusion by integrating the action route probability value, the classroom location information, and the traveling state to generate a target person behavior set;

[0046] Step S57: Input the target person behavior set into the discriminant layer for category judgment to obtain the target person's safety risk behavior.

[0047] The second aspect of the present invention provides a campus safety risk warning system based on wireless signals, including:

[0048] A behavior module, configured to obtain the wireless signal of the target person, perform continuous action perception on the wireless signal of the target person to obtain behavior information;

[0049] An identity module, configured to authenticate the identity of the target person based on the behavior information to obtain identity information;

[0050] A positioning module, configured to analyze the identity information to obtain classroom location information, and determine the regular behavior route according to the classroom location information;

[0051] An integration module, configured to integrate the behavior information, the identity information, the classroom location information, and the regular behavior route into a target person behavior vector set;

[0052] A risk analysis module, configured to build a campus safety risk analysis model, input the target person behavior vector set into the campus safety risk analysis model for analysis, generate the target person's safety risk behavior, and generate a safety risk warning information according to the target person's safety risk behavior.

[0053] In an optional embodiment, the risk analysis module includes:

[0054] A semantic unit for semantically enhancing the target personnel behavior vector set to generate a semantic feature vector;

[0055] A continuous action perception unit for extracting a behavior information semantic feature vector from the semantic feature vector, inputting the behavior information semantic feature vector into the action route model layer for continuous temporal action perception, and obtaining continuous temporal action features;

[0056] A traveling unit for determining the traveling direction, traveling speed, and traveling state of the target personnel according to the continuous temporal action features;

[0057] A simulated route unit for simulating an action route for the traveling direction in combination with a campus map to generate a simulated action route;

[0058] A similarity calculation unit for inputting the simulated action route into the action attention layer to match the action route with the regular action route to obtain an action route similarity;

[0059] A semantic fusion unit for normalizing the action route similarity into an action route probability value, comprehensively performing semantic fusion on the action route probability value, the classroom location information, and the traveling state to generate a target personnel behavior set;

[0060] A category judgment unit for inputting the target personnel behavior set into the discrimination layer for category judgment to obtain target personnel safety risk behaviors.

[0061] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements a campus security risk warning method based on wireless signals.

[0062] A fourth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements a campus security risk warning method based on wireless signals.

[0063] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0064] By overcoming the visual dead angle problem of monitoring devices through wireless signals that cover all corners of the campus, and by continuously capturing and analyzing wireless signals, potential security risks can be perceived and predicted in real time. It solves the problems of the traditional video monitoring technology lacking the ability to intelligently identify and analyze students' behaviors, being unable to automatically identify and warn students' risk behaviors. At the same time, it cannot monitor the whole process, resulting in the inability to identify students' high-risk behaviors and action routes, making the information not well-connected and unable to make emergency plans in a timely manner, thus leading to campus security incidents. Brief Description of the Drawings

[0065] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings. In the drawings:

[0066] Figure 1 is a schematic flowchart of the campus security risk early warning method based on wireless signals provided in Embodiment 1 of the present invention;

[0067] Figure 2 is a schematic structural diagram of the campus security risk early warning system based on wireless signals provided in Embodiment 2 of the present invention;

[0068] Figure 3 is a schematic structural diagram of an electronic device provided in Embodiment 3 of the present invention. Detailed Embodiments

[0069] To make the purpose, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the embodiments and the drawings. The illustrative embodiments of the present invention and their descriptions are only used to explain the present invention and are not used to limit the present invention.

[0070] Embodiment 1

[0071] Figure 1 is a schematic flowchart of the campus security risk early warning method based on wireless signals provided in Embodiment 1 of the present invention. As Figure 1 shown, the campus security risk early warning method based on wireless signals includes the following steps:

[0072] Step S1: Obtain the wireless signal of the target person, and perform continuous action perception on the wireless signal of the target person to obtain behavior information;

[0073] Step S2: Based on the behavior information, perform identity authentication on the target person to obtain identity information;

[0074] Step S3: Analyze the identity information to obtain classroom location information, and determine the regular behavior route according to the classroom location information;

[0075] Step S4: Integrate the behavior information, the identity information, the classroom location information and the regular behavior route into a target person behavior vector set;

[0076] Step S5: Construct a campus security risk analysis model, input the set of target personnel behavior vectors into the campus security risk analysis model for analysis, generate target personnel security risk behaviors, and generate security risk warning information based on the target personnel security risk behaviors.

[0077] It should be noted that currently, there are often visual dead angles in the layout of campus monitoring devices, resulting in some key areas or corners being unable to be effectively monitored. There may be some extreme behaviors of students in the monitoring dead angles, leading to the occurrence of security risks, and the videos cannot be monitored. At the same time, schools usually cannot carry devices such as mobile phones, resulting in the inability to obtain students' behaviors and information from the devices. Therefore, this method proposes a campus security risk warning method based on wireless signals. The wireless signals include short-range wireless communication technologies such as Bluetooth, Wi-Fi, ZigBee, GPRS, and NFC. For example, Wi-Fi signals, which are the basic communication facilities deployed in each campus and cover all corners of the campus comprehensively. Using Wi-Fi for campus risk detection and warning overcomes the inability to monitor key areas and corners due to the dead angles in the layout of monitoring devices. At the same time, using Wi-Fi signals for behavior analysis also does not require students to carry any devices.

[0078] First, capture the wireless signals related to the target personnel, where the target personnel include those in the monitoring dead angles and the areas not covered by the monitoring. Continuously sense the actions of the wireless signals of the target personnel, and discriminate the behavior information of the target personnel through the changes in the wireless signals, and then generate the behavior information.

[0079] This behavior information can not only be used to discriminate the behavior actions of the target personnel, but also can be used for identity recognition based on the behavior actions. In this method, the behavior information is combined with the personnel behavior pattern to identify the target personnel, and the identity information about the target personnel is obtained. The identity information in this method refers to the relevant identity information based on human behavior characteristics.

[0080] Match and search from the historical information according to its identity information to determine the campus identity information of the target personnel. The campus identity information includes information such as the grade, class, and teaching building where the target personnel are located. If there is no historical personnel behavior pattern, it will be determined as an outsider, which is also one of the campus risks that need attention.

[0081] Integrate behavior information, identity information, classroom location information, and regular behavior routes into a target person behavior vector set, which can be input into a security risk analysis model for security risk analysis. The security risk analysis model can process the information in the target person behavior vector set, input the target person behavior vector set into the model for analysis, generate the security risk behaviors of the target person, and generate security risk warning information based on the security risk behaviors so as to take preventive measures in a timely manner.

[0082] This method overcomes the visual dead angle problem of monitoring devices through wireless signals covering all corners of the campus, and can continuously capture and analyze wireless signals to perceive and predict potential security risks in real time. It solves the problems of traditional video monitoring technology lacking the ability of intelligent recognition and analysis of students' behaviors, being unable to automatically identify and warn students' risk behaviors. At the same time, it cannot monitor the whole process, resulting in the inability to identify students' high-risk behaviors and action routes, making the information not communicate smoothly, unable to make emergency plans in a timely manner, and thus generating campus security incidents.

[0083] In an alternative embodiment, obtain the wireless signal of the target person, and perform continuous action perception on the wireless signal of the target person, including the following steps:

[0084] Step S11: Receive the campus wireless signal, perform noise correction on the campus wireless signal, and perform target association estimation on the noise-corrected campus wireless signal to generate the wireless signal of the target person;

[0085] Step S12: Extract the signal amplitude change and signal phase change from the wireless signal of the target person, extract limb information from the signal amplitude change and signal phase change, and arrange the extracted limb information in time sequence to generate a time-sequence action signal;

[0086] Step S13: Window the time-sequence action signal to generate a windowed time-sequence action signal;

[0087] Step S14: Extract action segment features from the window of the windowed time-sequence action signal to generate an action segment feature sequence;

[0088] Step S15: Perform continuous feature capture on the action segment feature sequence to obtain an action probability;

[0089] Step S16: Integrate the action probability with the windowed time-sequence action signal to generate behavior information.

[0090] It should be noted that usually a signal transmitting and receiving device on campus is equipped with several antennas for signal transmission and reception. As a result, numerous campus wireless signals are received, and most of them have time errors. At the same time, wireless signals are also affected by electromagnetic and other environmental and objective factors during the transmission process. Therefore, after receiving campus wireless signals, it is necessary to perform noise correction on the campus wireless signals. At the same time, through target association estimation on the campus wireless signals after noise correction, the purpose of this step is to associate wireless signals belonging to the same person and exclude wireless information belonging to different people to avoid misjudgment of the action behavior of the target person in the future.

[0091] Extract the signal amplitude and signal phase changes from the wireless signals of the target person. The signal amplitude and signal phase changes can be used to calculate the actions of the target person and reflect the behavioral characteristics such as the movement state of the target person. Integrating them in chronological order to generate the chronological action signal one is for continuous action perception of each time sequence and also for predicting the action of the next time sequence.

[0092] Since this method is for continuous action perception, therefore, after the data is time-sequenced, it needs to be further windowed. By analyzing the action signals within the window and processing the chronological action signals window by window, extract the action segment features. These features can reflect the specific behavior of the target person during this time period. Through the association relationship of the action segment features between windows, continuous action capture is performed to complete continuous action perception. Integrate the calculated action probability with the windowed chronological action signals to generate complete behavior information. This behavior information can reflect the behavior pattern and changes of the target person within a certain time period.

[0093] In this embodiment, the noise correction of the campus wireless signal includes:

[0094] Calculate the phase shift caused by the frequency interval between two consecutive carriers during the continuous propagation of the campus wireless signal, and use this phase shift to calculate the propagation time of the campus wireless signal. And based on this phase shift and propagation time, construct a propagation matrix for the campus wireless signal.

[0095] Since the campus wireless signal is a superposition of multipath propagation, therefore, expand the propagation matrix of the campus wireless signal into a one-dimensional vector, calculate the covariance matrix of this one-dimensional vector, and perform singular value decomposition on the covariance matrix. The covariance matrix is decomposed into eigenvectors with several eigenvalues. Based on the minimum description length criterion, use these several eigenvalues to estimate the number of wireless signals of the target person. Select the maximum number from the estimated numbers, and construct the noise signal vector with the eigenvectors corresponding to the remaining eigenvalues obtained by subtracting the maximum data from the several eigenvalues.

[0096] Calculate the signal candidate pairs of the wireless signal of the target person using the noise signal vector, perform a general peak search on the signal candidate pairs of the wireless signal of the target person, and obtain the wireless signal of the target person.

[0097] The most core inventive point in the present invention is to capture the continuous features of the action segment feature sequence. This is the basic and core content of the present invention because the capture of the continuous features of the action segment feature sequence is related to the identification of the identity and behavior of the target person. Only by accurately capturing the continuity of the action segment feature sequence can accurate basis be provided for subsequent identity recognition and behavior recognition.

[0098] It should be noted that continuous actions will cause continuous fluctuations in the amplitude and phase of the wireless signal of the target person. Therefore, the present invention extracts the changes in signal amplitude and signal phase to extract and continuousize the action signals. At the same time, the degree and duration of amplitude changes and phase changes caused by different limb actions are not the same, resulting in only partial data of the action in one window. Therefore, the present invention fragments the action segment features extracted from each windowed time-series action signal, and then through continuous feature capture, splices and identifies the fragmented action features to obtain the action probability required by the present invention.

[0099] Furthermore, since campus security risks not only include tracking the whereabouts of the target person but also the behavior of the target person, where the behavior of the target person includes dangerous behaviors such as fighting, etc., it is necessary to perform limb recognition on the action signals at each time series to determine the limb information contained in the action signals at each time series. On this basis, this embodiment further proposes to extract the signal amplitude change and signal phase change from the wireless signal of the target person for limb information extraction, and integrate the extracted limb information according to the time series to generate time-series action signals.

[0100] In an optional embodiment, extracting limb information from the signal amplitude change and signal phase change includes:

[0101] Convert the signal amplitude change and the signal phase change into the spatial domain to generate a signal feature image;

[0102] Map the limb information feature image into a three-dimensional space to generate a three-dimensional signal feature image;

[0103] Extract the bone node features from the three-dimensional signal feature image, and use the extracted bone node features to construct the limb structure to generate limb information.

[0104] Specifically, each limb information represents an action. The limb information is arranged in chronological order to generate chronological action information. In this embodiment, the chronological action signal is Act = {act1, act2,..., act i ,..., act n}, where act i is the action signal at the i-th chronology, and n is the number of chronologies. The chronological action signal is windowed to generate a windowed chronological action signal, where the windowed chronological action signal is Win = {win1, win2,..., win i ,..., win m}, where win i is the chronological action signal of the i-th window, and m is the number of windows, m < n. Each window includes several chronological action signals. For example, win1 = (act1, act2, act3); the number of chronological action signals included in each window is n / m.

[0105] The windowed chronological action signal is input into the feature extraction network to extract action segment features. The action segment features are sorted by time to generate an action segment feature sequence. Continuity features are extracted from two adjacent action segment features in the action segment feature sequence, and the action segment feature sequence is aggregated according to the continuity features. According to the aggregation result, the same or similar actions are merged, and the unaggregated parts are deleted to obtain a complete action segment feature sequence Wor = {wor1, wor2,..., wor i ,..., wor k}, where wor i is the i-th action segment, and k is the length of the action segment feature sequence. Probability calculations are performed on each action segment in the action segment feature sequence, and the calculation results are integrated into the corresponding windowed chronological action signals to generate behavior information.

[0106] In this embodiment, the behavior information is Act-p = {(act1, p1), (act2, p1),...,(act i , p j ),...,(act n , p k}}, where p j is the probability value of the j-th action segment.

[0107] In an alternative embodiment, identity authentication of the target person is performed based on the behavior information, including the following steps:

[0108] Step S21: Extract walking information from the behavior information, extract gait features from the walking information, and obtain a gait signal;

[0109] Step S22: Estimate the normal walking speed of the target person, calculate the current walking speed of the target person according to the gait signal, and calculate the time series difference between the current walking speed and the normal walking speed;

[0110] Step S23: Use the time series difference to correct the gait signal to obtain a corrected gait signal;

[0111] Step S24: Perform identity authentication on the corrected gait signal to obtain identity information.

[0112] Specifically, extract the walking information related to walking features from Act-p = {(act1, p1), (act2, p1),..., (act, p),..., (act, p)}. In this embodiment, the walking feature refers to the action feature related to the legs. Among them, the gait feature signal refers to the walking information with a moving trend in the walking information as the gait signal. Specifically, in this embodiment, the gait signal is calculated as follows: i t,p j ),...,(ac n t,p k ) The walking information related to the walking features is extracted from act in. In this embodiment, the walking feature refers to the action feature related to the legs. Among them, the gait feature signal refers to the walking information with a moving trend in the walking information as the gait signal. Specifically, in this embodiment, the gait signal is calculated as follows: i The walking information related to the walking features is extracted from act in Act-p = {(act1, p1), (act2, p1),..., (act, p),..., (act, p)}. In this embodiment, the walking feature refers to the action feature related to the legs. Among them, the gait feature signal refers to the walking information with a moving trend in the walking information as the gait signal. Specifically, in this embodiment, the gait signal is calculated as follows:

[0113] wal i = act_w i * p_w j

[0114] In the above formula, wal i is the gait probability of the i-th time series action signal, act_w i is the walking signal related to the walking feature in the i-th time series action signal, and p_w j is the walking probability corresponding to the walking signal related to the walking feature in the i-th time series action signal.

[0115] Set a gait probability threshold, regard the time series action signal with a gait probability greater than or equal to the gait probability threshold as the gait signal, and integrate all the gait signals according to the time series to generate a gait signal time series set.

[0116] Connect the walking features in the gait signal time series set, and calculate based on the time series of the connected walking features, the current walking speed of the target person can be obtained. In this embodiment, the average value of the walking speeds in the gait signal time series set is estimated as the normal walking speed of the target person, calculate the time series difference, and correct the gait signal to avoid the gait calculation error caused by the time series error.

[0117] Based on the corrected gait signal combined with the historical campus personnel information, identity authentication can be achieved according to the length and speed of the steps, etc., and identity information can be obtained.

[0118] In an alternative embodiment, the identity information is analyzed to obtain classroom location information, and a regular behavior route is determined according to the classroom location information, including the following steps:

[0119] Step S31: Obtain the campus identity database, compare the identity information with the campus identity database, and obtain classroom location information;

[0120] Step S32: Obtain the campus map and the current location of the target person, and determine the regular behavior route from the campus map according to the current location of the target person and the classroom location information.

[0121] In an alternative embodiment, a campus security risk analysis model is constructed, including:

[0122] An action route simulation layer for continuously perceiving sequential actions of behavior information and simulating an action route according to the continuous sequential action perception;

[0123] An action attention layer for matching the action route simulation result with the regular action route;

[0124] A fusion layer for semantically fusing the behavior information according to the action route matching result;

[0125] A discrimination layer for making a category judgment based on the result of semantic fusion.

[0126] In an alternative embodiment, inputting the target person's behavior vector set into the campus security risk analysis model for analysis includes the following steps:

[0127] Step S51: Semantically enhance the target person's behavior vector set to generate a semantic feature vector;

[0128] Step S52: Extract the behavior information semantic feature vector from the semantic feature vector, input the behavior information semantic feature vector into the action route model layer for continuous sequential action perception, and obtain continuous sequential action features;

[0129] Step S53: Determine the traveling direction, traveling speed, and traveling state of the target person according to the continuous sequential action features;

[0130] Step S54: Combine the campus map to simulate an action route for the traveling direction and generate a simulated action route;

[0131] Step S55: Input the simulated action route into the action attention layer to perform action route matching with the regular action route, and obtain the action route similarity.

[0132] Step S56: Normalize the action route similarity into an action route probability value, and perform semantic fusion on the basis of the action route probability value, the classroom location information, and the travel status to generate a set of target personnel behaviors.

[0133] Step S57: Input the set of target personnel behaviors into the discriminant layer for category judgment to obtain the safety risk behaviors of the target personnel.

[0134] It should be noted that since there are numerous and complex categories of human behaviors, it is difficult to collect training data for all categories during the training of the network model. Therefore, in the face of the problem of insufficient small-sample training data, the present invention enhances the semantics of the behaviors of the target personnel, expands the samples of the target personnel behaviors with relevant behaviors and behavior characteristics, generates semantic feature vectors, and enhances the generalization ability of the campus safety risk analysis model. In this embodiment, the behavior information semantic feature vector in the semantic feature vector is used to perceive the continuous time-series action features to predict the next behavior of the target personnel. And the travel data of the target personnel, such as the travel direction, travel speed, and travel status, are determined according to the continuous time-series action features. These three data are the basic data for judging the next dangerous behavior of the target personnel.

[0135] Mark the travel direction on the campus map, regard all the locations connected to this travel direction as the possible destinations, simulate the action route, generate several simulated action routes, and perform similarity matching between the simulated action route and the regular action route, so as to determine whether the target personnel deviates from the regular route. Because the routes of campus personnel on campus are relatively fixed, students have fixed classrooms, and faculty and staff have fixed offices. When the target personnel deviates from the regular route, there is a possibility of dangerous behavior.

[0136] The travel status refers to the action status of each part of the target personnel including the torso. By fusing the action route probability value, the classroom location information, and the travel status to jointly determine the set of target personnel behaviors, and performing action judgment by the discriminant layer, the location that the target personnel will go to and the actions to be performed can be obtained. Based on the above information, the safety risk behavior category can be judged to generate the safety risk behaviors of the target personnel.

[0137] Furthermore, generating the safety risk warning information according to the safety risk behaviors of the target personnel includes: fusing the safety risk behaviors of the target personnel with identity authentication to generate the action prediction and behavior risk including the identity of the target personnel, and transmitting the action prediction and action risk as the safety risk warning information to the security personnel for safety risk prevention.

[0138] Embodiment 2

[0139] Figure 2 FIG. is a schematic structural diagram of a campus security risk warning system based on wireless signals provided by Embodiment 2 of the present invention, as Figure 2 shown. The campus security risk warning system based on wireless signals includes:

[0140] A behavior module, configured to obtain the wireless signals of a target person, perform continuous action perception on the wireless signals of the target person, and obtain behavior information;

[0141] An identity module, configured to perform identity authentication on the target person based on the behavior information to obtain identity information;

[0142] A positioning module, configured to analyze the identity information to obtain classroom positioning information, and determine a regular behavior route according to the classroom positioning information;

[0143] An integration module, configured to integrate the behavior information, the identity information, the classroom positioning information, and the regular behavior route into a target person behavior vector set;

[0144] A risk analysis module, configured to construct a campus security risk analysis model, input the target person behavior vector set into the campus security risk analysis model for analysis, generate a target person security risk behavior, and generate a security risk warning information according to the target person security risk behavior.

[0145] In an alternative embodiment, the risk analysis module includes:

[0146] A semantic unit, configured to perform semantic enhancement on the target person behavior vector set to generate a semantic feature vector;

[0147] A continuous action perception unit, configured to extract a behavior information semantic feature vector from the semantic feature vector, input the behavior information semantic feature vector into the action route model layer for continuous time series action perception, and obtain continuous time series action features;

[0148] A travel unit, configured to determine the travel direction, travel speed, and travel state of the target person according to the continuous time series action features;

[0149] A simulated route unit, configured to simulate an action route for the travel direction in combination with a campus map to generate a simulated action route;

[0150] A similarity calculation unit, configured to input the simulated action route into the action attention layer to perform action route matching with the regular action route to obtain an action route similarity;

[0151] A semantic fusion unit is configured to normalize the similarity of the action routes into action route probability values, and perform semantic fusion by integrating the action route probability values, the classroom location information, and the traveling status to generate a set of target personnel behaviors.

[0152] A category judgment unit is configured to input the set of target personnel behaviors into the discrimination layer for category judgment to obtain target personnel safety risk behaviors.

[0153] Embodiment 3

[0154] Figure 3 The structural schematic diagram of an electronic device provided in Embodiment 3 of the present invention is as Figure 3 shown. The electronic device includes a processor 21, a memory 22, an input device 23, and an output device 24. The number of processors 21 in the computer device may be one or more. Figure 3 Here, one processor 21 is taken as an example. The processor 21, the memory 22, the input device 23, and the output device 24 in the electronic device may be connected through a bus or other means. Figure 3 Here, connection through a bus is taken as an example.

[0155] The memory 22, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules. The processor 21 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 22, that is, implements the campus safety risk warning method based on wireless signals in Embodiment 1.

[0156] The memory 22 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function. The data storage area can store data created according to the use of the terminal, etc. In addition, the memory 22 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some instances, the memory 22 may further include a memory remotely set relative to the processor 21, and these remote memories can be connected to the electronic device through a network. Examples of the above network include but are not limited to the Internet, an enterprise internal network, a local area network, a mobile communication network, and their combinations.

[0157] The input device 23 can be used to receive the id, password, etc. input by the user. The output device 24 is used to output a network configuration page.

[0158] Embodiment 4

[0159] Embodiment 4 of the present invention further provides a computer-readable storage medium. The computer-executable instructions are used to implement the campus security risk warning method based on wireless signals provided in Embodiment 1 when executed by a computer processor.

[0160] A storage medium containing computer-executable instructions provided in an embodiment of the present invention. The computer-executable instructions are not limited to the method operations provided in Embodiment 1, and can also execute relevant operations in the campus security risk warning method based on wireless signals provided in any embodiment of the present invention.

[0161] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A campus security risk early warning method based on wireless signals, characterized in that, It includes the following steps: Step S1: Obtain the wireless signal of the target person, perform continuous action perception on the wireless signal of the target person, and obtain behavior information; Step S2: Based on the behavior information, perform identity authentication on the target person to obtain identity information; Step S3: Analyze the identity information to obtain classroom location information, and determine the regular behavior route according to the classroom location information; Step S4: Integrate the behavior information, the identity information, the classroom location information, and the regular behavior route into a target person behavior vector set; Step S5: Construct a campus security risk analysis model, input the target person behavior vector set into the campus security risk analysis model for analysis, generate the security risk behavior of the target person, and generate a security risk warning message according to the security risk behavior of the target person.

2. The campus security risk early warning method based on wireless signals according to claim 1, characterized in that, Obtain the wireless signal of the target person, and perform continuous action perception on the wireless signal of the target person, including the following steps: Step S11: Receive the campus wireless signal, perform noise correction on the campus wireless signal, and perform target association estimation on the noise-corrected campus wireless signal to generate the wireless signal of the target person; Step S12: Extract the signal amplitude change and signal phase change from the wireless signal of the target person, extract limb information from the signal amplitude change and signal phase change, and arrange the extracted limb information in time sequence to generate a time-sequence action signal; Step S13: Window the time-sequence action signal to generate a windowed time-sequence action signal; Step S14: Extract action segment features from the window of the windowed time-sequence action signal to generate an action segment feature sequence; Step S15: Capture the continuity feature of the action segment feature sequence to obtain an action probability; Step S16: Integrate the action probability with the windowed time-sequence action signal to generate behavior information.

3. The campus security risk warning method based on wireless signals according to claim 2, wherein, Extract limb information from the signal amplitude change and signal phase change, including: Convert the signal amplitude change and the signal phase change to the spatial domain to generate a signal feature image; Map the limb information feature image to the three-dimensional space to generate a three-dimensional signal feature image; Extract the bone node features from the three-dimensional signal feature image, and use the extracted bone node features to construct the limb structure to generate limb information.

4. The campus security risk warning method based on wireless signals according to claim 1, wherein, Based on the behavior information, perform identity authentication on the target person, including the following steps: Step S21: Extract walking information from the behavior information, extract gait features from the walking information to obtain a gait signal; Step S22: Estimate the regular walking speed of the target person, calculate the current walking speed of the target person according to the gait signal, and calculate the time-sequence difference between the current walking speed and the regular walking speed; Step S23: Use the time-sequence difference to correct the gait signal to obtain a corrected gait signal; Step S24: Perform identity authentication on the corrected gait signal to obtain identity information.

5. The campus security risk early warning method based on wireless signals according to claim 1, wherein Construct a campus security risk analysis model, including: An action route simulation layer for performing continuous sequential action perception on behavior information and simulating an action route based on the continuous sequential action perception; An action attention layer for matching the action route of the result of the action route simulation with the regular action route; A fusion layer for semantically fusing the behavior information according to the result of the action route matching; A discrimination layer for performing category judgment based on the result of the semantic fusion.

6. The campus security risk early warning method based on wireless signals according to claim 5, wherein Inputting the set of target person behavior vectors into the campus security risk analysis model for analysis, including the following steps: Step S51, performing semantic enhancement on the set of target person behavior vectors to generate semantic feature vectors; Step S52, extracting the behavior information semantic feature vectors from the semantic feature vectors, inputting the behavior information semantic feature vectors into the action route model layer for continuous sequential action perception to obtain continuous sequential action features; Step S53, determining the traveling direction, traveling speed and traveling state of the target person according to the continuous sequential action features; Step S54, simulating an action route for the traveling direction in combination with the campus map to generate a simulated action route; Step S55, inputting the simulated action route into the action attention layer to match the action route with the regular action route to obtain an action route similarity; Step S56, normalizing the action route similarity into an action route probability value, and comprehensively performing semantic fusion on the action route probability value, the classroom location information and the traveling state to generate a set of target person behaviors; Step S57, inputting the set of target person behaviors into the discrimination layer for category judgment to obtain the security risk behaviors of the target person.

7. A campus security risk early warning system based on wireless signals, characterized in that, A campus security risk warning system based on wireless signals, including: A behavior module for acquiring the wireless signals of the target person and performing continuous action perception on the wireless signals of the target person to obtain behavior information; An identity module for authenticating the identity of the target person based on the behavior information to obtain identity information; A positioning module for analyzing the identity information to obtain classroom location information and determining a regular behavior route according to the classroom location information; An integration module for integrating the behavior information, the identity information, the classroom location information and the regular behavior route into a set of target person behavior vectors; A risk analysis module for constructing a campus security risk analysis model, inputting the set of target person behavior vectors into the campus security risk analysis model for analysis, generating the security risk behaviors of the target person, and generating security risk warning information according to the security risk behaviors of the target person.

8. The campus security risk early warning method based on wireless signals according to claim 7, wherein The risk analysis module includes: A semantic unit for performing semantic enhancement on the set of target person behavior vectors to generate semantic feature vectors; A continuous action perception unit for extracting the behavior information semantic feature vectors from the semantic feature vectors, inputting the behavior information semantic feature vectors into the action route model layer for continuous sequential action perception to obtain continuous sequential action features; A movement unit, configured to determine the movement direction, movement speed, and movement state of a target person according to the continuous sequential action features; A simulated route unit, configured to simulate an action route for the movement direction in combination with a campus map to generate a simulated action route; A similarity calculation unit, configured to input the simulated action route into the action attention layer to perform action route matching with the regular action route to obtain an action route similarity; A semantic fusion unit, configured to normalize the action route similarity into an action route probability value, and perform semantic fusion by integrating the action route probability value, the classroom location information, and the movement state to generate a target person behavior set; A category judgment unit is configured to input the target person behavior set into the discrimination layer to perform category judgment to obtain a target person safety risk behavior.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method for warning of campus safety risks based on wireless signals as described in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and when the computer program is executed by a processor, the method for warning of campus safety risks based on wireless signals as described in any one of claims 1 to 6 is implemented.

Images