Park visitor safety early warning method and system based on digital twinning and medium
By constructing a digital twin-based park visitor security early warning system, and utilizing visitor data analysis and suspicious behavior identification models, the system solves the problem of timely early warning of abnormal behavior in park visitor management, thereby improving park security and the efficiency of security decision-making.
Patent Information
- Application Number
- CN202411082894.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-11-12
AI Technical Summary
Existing park visitor management methods are unable to identify and warn of suspicious individuals’ abnormal behavior in a timely manner, resulting in insufficient timeliness and accuracy of security decisions.
A digital twin-based approach to visitor security early warning is adopted. By constructing a people recognition model and a behavior recognition model, and combining the park's video image data, the number of visitors and their behavior are analyzed in real time, a suspicious identification index is calculated, an abnormal behavior warning is triggered, and the visitor distribution and warning status are displayed on the digital twin platform.
It enables accurate identification and predictive analysis of abnormal visitor behavior, allowing for timely emergency measures to ensure park safety, reduce labor costs, and improve security decision-making efficiency.
Smart Images

Figure CN118942159A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of park security monitoring technology, and in particular to a park visitor safety early warning method, system and medium based on digital twins. Background Art
[0002] In the prior art, the visitor management of the park is usually carried out by the park guard registering the visitor's entry information, and then performing extensive management through security patrols. It is almost entirely dependent on manual labor, which is inefficient and unable to verify the true identity of the visitor. It is impossible to prevent sales personnel from deliberately entering the park to promote their products, and it is impossible to track the behavior of each visitor after entering the park in real time. It is impossible to detect the abnormal behavior of suspicious persons in time and cannot provide advance warning, which seriously affects the timeliness and accuracy of the park's security decision-making.
[0003] A Chinese patent with publication number CN116822664A discloses a visitor reservation management method and device, the method including visitor information collection, collecting personal information input by visitors; reservation information collection, collecting reservation information input by visitors; visitor information management, managing visitors' personal information, including recording, storing and encrypting; reservation information management, managing visitor reservation information, including setting, querying and modifying; personal information verification, verifying and authenticating visitors' personal information; intelligent identification and tracking, collecting and identifying visitors appearing in a set area; however, the Chinese patent with publication number CN116822664A does not involve specific methods for visitor identification and behavior warning. Summary of the invention
[0004] The present invention solves the problem that the existing park visitor management method is unable to provide timely warning for abnormal behavior of suspicious persons, and proposes a park visitor safety warning method, system and medium based on digital twins, which can accurately identify and predict abnormal behavior of visitors, so as to provide timely warning and take emergency measures to ensure the safety inside the park.
[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A digital twin-based park visitor safety early warning method includes the following steps: S1, constructing a crowd recognition model based on the video image data in the park, and determining the number of people in the recognition area based on the crowd recognition model; S2, constructing a behavior recognition model based on the video image data containing the visitor target, and obtaining the behavior analysis based on the visitor according to the behavior recognition model; S3: Determine whether to trigger a headcount warning based on the headcount situation, calculate the suspicious identification index based on the behavior analysis situation, and determine whether to trigger an abnormal behavior warning; S4, build a digital twin platform based on the park to display the distribution and early warning status of park visitors.
[0006] In this technical solution, firstly, the video image data in the park is obtained to construct a crowd recognition model, and then the number of people in each recognition area is obtained. Then, a behavior recognition model is constructed based on the relevant video image data to obtain the behavior analysis of visitors and predict the visitor behavior in future time periods. According to the above-mentioned behavior analysis and the number of people, it is determined whether to trigger an early warning so as to issue a timely early warning and take emergency measures. Finally, a digital twin platform is constructed based on the relevant data of the park to display the visitor situation in the park and provide data reference and decision-making assistance to park managers.
[0007] The present invention is further configured as follows: The method of constructing a human identification model based on the video image data in the park includes: S11, decomposing the video data of the same scene frame by frame to obtain frame images; S12, dividing the frame image into a plurality of grids, and monitoring the rectangular area containing the human body in the grids; S13, determining the grid to which the visitor belongs according to the center position coordinates of the rectangular area; S14, counting the visitors in the video data of all scenes to establish a crowd recognition model.
[0008] In this technical solution, we first obtain video images of the park, decompose and process the video data of the same scene to form several frame images, and determine the number and location of visitors by judging in which grid the rectangular area containing the human body is located. Finally, we construct a number recognition model that can detect the number of visitors.
[0009] The present invention is further configured as follows: The method of constructing a behavior recognition model based on video image data containing visitor targets includes: S21, decomposing the video image data containing the visitor target by frame, and marking and distinguishing different visitor targets respectively; S22, extracting visitor target features, analyzing the movement trajectory of the visitor target between each frame, determining the continuous trajectory information of the visitor target, and building a behavior recognition model; S23, predicting the body movements and movement positions of the visitor target according to the continuous trajectory information.
[0010] In this technical solution, the position and size of the visitor target in each frame of the image are determined and the visitor targets are marked and distinguished to avoid confusion of corresponding behavioral actions when there are multiple visitors in the image; then the visitor target features are extracted, and the motion trajectory of the visitor target between each frame is analyzed. The characteristic mobility and approximation of each visitor target are used to establish the continuous trajectory information of each visitor target, and the action and position of the visitor target in the next frame are predicted based on the behavioral trajectory of each visitor target.
[0011] The present invention is further configured as follows: The step S23 comprises: Based on the continuous trajectory information, the body movements and movement distance of the visitor target are identified, and the behavioral tendencies and target positions of the visitor target are determined in combination with the park entity characteristics around the visitor target, and then the body movements and movement positions of the visitor target in the next frame or the next time period are predicted.
[0012] In this technical solution, on the basis of building a behavior recognition model and obtaining the behavior analysis based on the visitor, the behavior tendency and target location of the visitor are obtained according to the actual body movements and movement distance of the visitor target, and combined with the surrounding park entity characteristics. Finally, the dynamics of the visitor target in the future period are predicted to achieve advance preparation and early warning.
[0013] The present invention is further configured as follows: The calculation of the suspicious identification index based on the behavior analysis to determine whether to trigger an abnormal behavior warning includes: Determine the types of abnormal behaviors of visitors, prioritize them, and then assign corresponding weight factors to them. Calculate the suspicious identification index based on the types of abnormal behaviors of visitors and the corresponding weight factors. If the calculated result is greater than the threshold index, an abnormal behavior warning is triggered.
[0014] In this technical solution, when the suspicious level index is higher than the threshold index, in addition to triggering a corresponding warning, the visitor will be automatically added to the blacklist and prohibited from entering the park next time.
[0015] The present invention is further configured as follows: The calculation formula of the suspicious identification index is: X=Math.floor(a*C+b*P+c*Z+d*F+e*L+f*W / (a+b+c+d+e+f)) Where X is the calculated suspicious identification index, C, P, Z, F, L, and W are the quantitative values of illegally entering restricted areas, wandering, looking around, climbing over walls, face occlusion, and tailing, respectively, a, b, c, d, e, and f are weight factors, and Math.floor() means rounding down decimals.
[0016] In this technical solution, the suspicious identification index is determined by the influence of multiple key types of visitor abnormal behaviors, which can provide early warning trigger conditions more accurately to a certain extent.
[0017] The present invention is further configured as follows: The construction of the park-based digital twin platform includes: Obtain relevant data on the park’s building space, cameras, and access control locations. The data format includes architectural CAD drawings and on-site photos. Based on the collected data, perform three-dimensional visualization modeling of the park and establish a virtual park digital twin corresponding to the physical entity.
[0018] In this technical solution, the collected data is more complete, including park building space information, cameras, access control location information and other related data. In addition to building CAD drawings and photos taken on site, the data format also includes map data and coordinates.
[0019] The present invention is further configured as follows: Before step S3, the following steps are also included: The data containing the number of people and the behavior analysis are encrypted separately, and after the encrypted data is sent to the destination, the data is restored and analyzed. The encryption algorithm uses the SHA algorithm or the SM3 algorithm.
[0020] In this technical solution, before performing early warning judgment on the data in step S3, the data containing the number of people and behavior analysis is encrypted to ensure the data security between the identification end and the destination end, prevent malicious data theft and destruction, and significantly increase the park's protection of visitor information.
[0021] A digital twin-based park visitor safety warning system adopts the above-mentioned digital twin-based park visitor safety warning method, including: The crowd recognition module uses the constructed crowd recognition model to analyze and output the number of people in the area; Behavior recognition module, which obtains the behavior analysis of the corresponding visitor target based on the behavior recognition model; The early warning module determines whether to trigger an early warning based on the number of people and behavior analysis; The three-dimensional display module shows the distribution of park visitors through a virtual digital twin of the park.
[0022] The system of the present technical solution mainly includes a number recognition module, a behavior recognition module, an early warning module and a three-dimensional display module. The steps of the number recognition module and the behavior recognition module can be performed simultaneously or one after the other. After the number recognition module and the behavior recognition module have completed the corresponding recognition process, they both enter the early warning module to determine whether the early warning is triggered. Finally, the corresponding visitor situation and the early warning situation can be displayed in a three-dimensional visual manner to ensure the safety inside the park.
[0023] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, executes the steps of a campus visitor safety warning method based on digital twins as described above.
[0024] In this technical solution, a computer program of a computer calibration storage medium is used to execute the above-mentioned digital twin-based park visitor safety warning method.
[0025] The present invention can bring the following beneficial effects: The present invention relates to a park visitor safety early warning method based on digital twins, which can accurately identify and predict abnormal behaviors of visitors so as to provide timely early warnings and take emergency measures to ensure the safety within the park. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 This is a flow chart of Example 1, a method for early warning safety of campus visitors based on digital twins.
[0027] Figure 2 This is a flow chart of Example 2, a method for early warning safety of campus visitors based on digital twins. DETAILED DESCRIPTION
[0028] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0029] Example 1 This embodiment proposes a campus visitor safety warning method based on digital twins. Figure 1 , which mainly includes the following steps.
[0030] Step S1, first, construct a corresponding number recognition model based on the video image data collected in the park. After the number recognition model is constructed, determine the number of people in the recognition area.
[0031] Constructing a corresponding human identification model based on the video image data collected in the park includes the following sub-steps.
[0032] Step S11, first, obtain video data of the same scene, and decompose the video data by frame to obtain a plurality of frame images; that is, decompose the video data into frame-by-frame images.
[0033] Step S12, then, the frame image is segmented to form a plurality of grids; in this embodiment, a total of A*A grids are formed, and it is monitored whether there is a rectangular area containing a human body in each grid and the number of rectangular areas is counted.
[0034] Step S13, determining whether the visitor belongs to this grid based on the center position coordinates of the rectangular area containing the human body; specifically, determining whether the center position of the rectangular area is within the grid. If the center position of the rectangular area is within the grid, the visitor target corresponding to the rectangular area is within the grid. If the center position of the rectangular area is not within the grid, continue to determine whether the rectangular area is within another grid until the grid to which it belongs is determined.
[0035] Step S14, statistics are obtained on the visitor attribution in the video data of all scenes, and finally a corresponding crowd recognition model is established. According to the crowd recognition model, the crowd situation of a certain scene or all scenes in the park can be obtained.
[0036] The above-mentioned crowd recognition model can display the number of people in various scenes in the park in real time, and can be combined with the digital twin platform for visual display.
[0037] In the above technical solution, video images of the park are first obtained, and the video data of the same scene is decomposed and processed to form several frame images. The number and position of visitors are determined by judging in which grid the rectangular area containing the human body is located, and finally a number recognition model capable of detecting the number of visitors is constructed.
[0038] After decomposing the video data of the same scene by frames, irrelevant features of the frame image and the background can be blurred to further accurately detect the number of visitors.
[0039] As the specific methods of irrelevant features and background blur are existing technologies, they will not be elaborated here.
[0040] Continue to refer Figure 1 After determining that the identification of the number of people is completed, proceed to step S2.
[0041] Step S2, constructing a behavior recognition model based on the video image data containing the visitor target, and obtaining a behavior analysis situation based on the visitor according to the behavior recognition model.
[0042] The above-mentioned construction of a behavior recognition model based on video image data containing visitor targets mainly includes the following sub-steps.
[0043] Step S21, screen and obtain video image data containing visitor targets, and decompose the video image data by frame. Similarly, decompose and obtain a number of frame images, on this basis, mark and distinguish different visitor targets on the frame images.
[0044] That is, each visitor target is set with a corresponding unique identification id, and the same visitor has the same id to facilitate behavioral action recognition.
[0045] Step S22, extracting the visitor target features in the frame images, analyzing the movement trajectory of the visitor target between each frame according to the visitor target features, so as to determine the continuous trajectory information about the visitor target, and then constructing a behavior recognition model.
[0046] Step S23: predicting the body movements and movement positions of the visitor target according to the continuous trajectory information.
[0047] In the above technical solution, the position and size of the visitor target in each frame of the image are determined and the visitor targets are marked and distinguished to avoid confusion of corresponding behavioral actions when there are multiple visitors in the image; then the visitor target features are extracted, and the motion trajectory of the visitor target between each frame is analyzed. The characteristic mobility and approximation of each visitor target are used to establish the continuous trajectory information of each visitor target, and the action and position of the visitor target in the next frame are predicted based on the behavioral trajectory of each visitor target.
[0048] The above process mainly includes the following processes: Target recognition: Decompose the video into frames of images, and use a deep learning-based target tracking algorithm to track the target's body movements in each frame to obtain the position and size of each visitor target. More specifically, each visitor target is assigned a corresponding unique identification ID, and the same visitor has the same ID to avoid confusion between the actions of multiple people in the image.
[0049] Feature extraction: Analyze each detected target through a deep learning network and extract the visitor target features in the image. Subsequent behavior tracking needs to be based on this result.
[0050] Human target tracking: Analyze the motion trajectory of the visitor target between each frame image, use the characteristic motion and approximation of each visitor target, establish a mapping relationship between each visitor target, and realize human target tracking.
[0051] Data Mapping: Track target visitors in all frames, solve data mapping association issues, and confirm which visitor targets belong to the same track.
[0052] Behavior prediction: Continuously analyze the content of adjacent frames, identify coherent actions, and predict the target's body movements and possible movement positions in the next frame based on the behavior trajectory of each visitor target.
[0053] For the above-mentioned step S23, more specifically, it identifies and obtains the body movements and moving distances of the visitor target based on the continuous trajectory information obtained above, and at the same time, combines the park entity characteristics around the visitor target to determine the visitor target's behavioral tendencies and target position, and then predicts the visitor target's body movements and moving positions in the next frame or the next time period.
[0054] In the above technical solution, on the basis of constructing the behavior recognition model and obtaining the behavior analysis based on the visitor, the behavior tendency and target location of the visitor are obtained according to the actual body movements and movement distance of the visitor target, and combined with the surrounding park entity characteristics. Finally, the dynamics of the visitor target in the future period are predicted to achieve advance preparation and early warning.
[0055] As an implementation method, a motion prediction model is constructed to predict body movements and moving positions. More specifically, the body movements and moving distances of the visitor target are first obtained, and the body movement information and moving distance information are preprocessed and used as training sets and verification sets of the original neural network model. The original neural network model is continuously trained to generate various parameters of the model, and the verification set is used to continuously verify the various parameters of the model to ensure that the generated motion prediction model has good prediction capabilities.
[0056] Step S3, judging whether to trigger a headcount warning based on the headcount situation, calculating a suspicious identification index based on the behavior analysis situation, and judging whether to trigger an abnormal behavior warning.
[0057] For the judgment of the number of people warning, the number of visitors in the current area is mainly obtained through the number of people recognition model, and whether to trigger the warning is determined according to the visitor number threshold of the current area, that is, the number of visitors in the current area is compared with the visitor number threshold of the current area. If the number of visitors in the current area is greater than the visitor number threshold of the current area, the crowd gathering alarm is triggered; if the number of visitors in the current area is less than or equal to the visitor number threshold of the current area, the crowd gathering alarm will not be triggered.
[0058] The process of calculating the suspicious identification index based on the behavior analysis and determining whether to trigger an abnormal behavior warning specifically includes the following steps.
[0059] First, determine the types of visitor abnormal behaviors and prioritize them.
[0060] Subsequently, corresponding weight factors are assigned respectively, and the suspicious identification index is calculated according to the types of abnormal behaviors of visitors and the corresponding weight factors.
[0061] A comparison is performed based on the above-mentioned suspicious identification index; if the calculated result is greater than the threshold index, an abnormal behavior warning is triggered; if the calculated result is less than the threshold index, no warning is triggered.
[0062] In the above technical solution, when the suspicious level index is higher than the threshold index, in addition to triggering a corresponding warning, the visitor will be automatically added to the blacklist and prohibited from entering the park next time.
[0063] The suspicious identification index is calculated specifically according to the following calculation formula.
[0064] X=Math.floor(a*C+b*P+c*Z+d*F+e*L+f*W / (a+b+c+d+e+f)) In the above formula, X is the above-mentioned suspicious identification index, C, P, Z, F, L and W represent the quantitative values of illegal entry into restricted areas, wandering, looking around, climbing over the fence, face occlusion and tailing, respectively, a, b, c, d, e and f are all weight factors, more specifically, a is the weight factor of illegal entry into restricted areas, b is the weight factor of wandering, c is the weight factor of looking around, d is the weight factor of climbing over the fence, e is the weight factor of face occlusion, and f is the weight factor of tailing. Math.floor() specifically means rounding down decimals.
[0065] In the above technical solution, the suspicious identification index is determined by the influence of multiple key types of visitor abnormal behaviors, which can provide early warning trigger conditions more accurately to a certain extent.
[0066] Step S4, finally, based on the associated data of the park, a digital twin platform based on the park is constructed to display the distribution and warning status of park visitors.
[0067] For building a park-based digital twin platform in step S4, it mainly includes the following steps.
[0068] Collect and obtain relevant data on the park’s building space, cameras, and access control locations. The relevant data formats mainly include architectural CAD drawings and on-site photos. Based on the collected data, three-dimensional visual modeling of the park is carried out, and then a virtual park digital twin corresponding to the physical entity of the park is established.
[0069] In the above technical solution, the collected data is more complete, including park building space information, cameras, access control location information and other related data. In addition to architectural CAD drawings and photos taken on site, the data format also includes map data and coordinates.
[0070] The above-mentioned architectural CAD drawings can also be replaced by three-dimensional drawings, and the photos taken on site can also be replaced by corresponding video data. Of course, other useful data sources can also be used to replace the above-mentioned architectural CAD drawings and photos taken on site, which will not be elaborated here.
[0071] In the above embodiment, the video image data in the park is first obtained to construct a number recognition model, and then the number of people in each recognition area is obtained. Subsequently, a behavior recognition model is constructed based on the relevant video image data to obtain a behavior analysis of the visitors and predict the visitor behavior in the future time period. Based on the above behavior analysis and number of people, it is determined whether to trigger an early warning so as to issue a timely early warning and take emergency measures. Finally, a digital twin platform is constructed based on the relevant data of the park to display the visitor situation in the park and provide data reference and decision-making assistance to park managers.
[0072] This embodiment can bring the following technical effects: Based on the existing video security system in the park, through machine learning, various abnormal behaviors of visitors can be automatically identified, automatic warning can be achieved, and labor costs can be reduced. It can prevent suspicious persons from entering the park at the entrance, track the behavior of visitors who have entered the park in real time, and promptly discover their abnormal behaviors and potential safety hazards. Visual warnings can be made in the digital twin platform at the first time, and the historical whereabouts of the visitor can be clearly drawn. The location of the incident can be located with one click, and the surrounding camera monitoring images can be retrieved in a linked manner. It can more intuitively grasp the safety status of park visitors and execute corresponding emergency response measures, helping park security personnel to make correct and timely response measures and eliminate various types of visitor behavior safety hazards.
[0073] Example 2 refer to Figure 2 , this embodiment includes the following steps.
[0074] Step S1, constructing a population recognition model based on the video image data in the park, and determining the number of people in the recognition area based on the population recognition model.
[0075] Step S2, constructing a behavior recognition model based on the video image data containing the visitor target, and obtaining a behavior analysis situation based on the visitor according to the behavior recognition model; Different from the first embodiment, step S1 and step S2 can be performed simultaneously without any restriction on the order.
[0076] Example 3 On the basis of Example 1, before performing step S3, the following steps are further performed.
[0077] The headcount data from step S1 and the behavior analysis data from step S2 are obtained respectively, and the headcount data and the behavior analysis data are encrypted respectively. After the encryption is completed, the encrypted data is sent from the identification end to the destination end. After being sent to the destination end, the received data is decrypted and restored respectively to finally obtain the original data.
[0078] Generally, the encryption algorithm uses the SHA algorithm or the SM3 algorithm.
[0079] In this embodiment, the headcount data from step S1 and the behavior analysis data from step S2 can be encrypted separately. For example, the headcount data can be encrypted using the SHA algorithm, and the behavior analysis data can be encrypted using the SM3 algorithm.
[0080] Of course, the same data encryption processing method can also be used for the headcount data from step S1 and the behavior analysis data from step S2. For example, both the headcount data and the analysis data use the SHA algorithm or the SM3 algorithm.
[0081] Regarding the transmission method of the encrypted data from the identification end to the destination end, the above-mentioned headcount data and analysis data may be transmitted in parallel or through a single transmission port.
[0082] In the above technical solution, before performing early warning judgment on the data in step S3, the data containing the number of people and behavior analysis is encrypted to ensure the data security between the identification end and the destination end, prevent malicious data theft and destruction, and significantly increase the park's protection of visitor information.
[0083] Example 4 Based on Example 1, this example proposes a campus visitor safety warning system based on digital twins, which mainly includes a number recognition module, a behavior recognition module, a warning module and a three-dimensional display module.
[0084] The number recognition module is connected to the behavior recognition module, the behavior recognition module is connected to the early warning module, and the early warning module is connected to the three-dimensional display module. Alternatively, the number recognition module and the behavior recognition module are connected to the early warning module at the same time.
[0085] The population recognition module mainly executes the process of step S1, and can obtain the number of people in the corresponding area according to the established population recognition model.
[0086] The behavior recognition module mainly executes the process of step S2, and can output the behavior analysis of the corresponding visitor target according to the behavior recognition model.
[0087] The early warning module can make early warning triggering judgments based on the number of people or the behavior analysis. For more detailed information, please refer to the content of step S3 of Example 1.
[0088] For the three-dimensional display module, it can display the distribution of park visitors through a virtual digital twin of the park.
[0089] The system of the present technical solution mainly includes a number recognition module, a behavior recognition module, an early warning module and a three-dimensional display module. The steps of the number recognition module and the behavior recognition module can be performed simultaneously or one after the other. After the number recognition module and the behavior recognition module have completed the corresponding recognition process, they both enter the early warning module to determine whether the early warning is triggered. Finally, the corresponding visitor situation and the early warning situation can be displayed in a three-dimensional visual manner to ensure the safety inside the park.
[0090] In the three-dimensional display module, when in daily mode, the three-dimensional picture of the entire park is displayed on the digital twin platform, and according to the number of visitors captured by the camera, the same number of virtual visitors are displayed in the corresponding area of the digital twin platform, intuitively presenting the current visitor distribution in the park. At the same time, it can switch to the visitor heat map mode. The redder the color of the heat map, the more concentrated the people, which makes it easier for administrators to know more intuitively which area the visitors are more concentrated in.
[0091] For visitor data and behaviors in areas not covered by cameras, such as conference rooms, behavior recognition models and behavior predictions are used to identify visitors. For example, although there are no cameras installed in conference rooms, the images captured by cameras in the aisles can be used to continuously analyze the contents of adjacent frames, identify visitors' continuous movements, and analyze the behavioral trajectory of each visitor target. The target's body movements and possible movement directions in the next frame are predicted, and then the number of visitors entering and leaving each conference room can be predicted based on the location coordinates and directions of the conference rooms in the digital twin platform.
[0092] When in alarm mode, when the early warning module triggers the alarm, the digital twin platform immediately locates the alarm location and displays the alarm icon and suspicious level index above the head of the virtual person of the visitor with abnormal behavior. Combined with the time point when the visitor appears in different camera images, the three-dimensional space path analysis algorithm is used to draw out its recent historical trajectory route. At the same time, the device IoT module is linked to call up surrounding cameras to view the video of the incident point, so as to fully grasp the ins and outs of the alarm event, improve the speed and efficiency of security decision-making, and prevent potential security threats as early as possible.
[0093] Example 5 This embodiment also proposes a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a campus visitor safety warning method based on digital twins in the above-mentioned embodiment 1 are executed.
[0094] 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 the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.
Claims
1. A digital twin-based park visitor safety warning method, It is characterized in that The following steps are involved: S1, constructing a crowd recognition model based on the video image data in the park, and determining the number of people in the recognition area based on the crowd recognition model; S2, constructing a behavior recognition model based on the video image data containing the visitor target, and obtaining the behavior analysis based on the visitor according to the behavior recognition model; S3: Determine whether to trigger a headcount warning based on the headcount situation, calculate the suspicious identification index based on the behavior analysis situation, and determine whether to trigger an abnormal behavior warning; S4, build a digital twin platform based on the park to display the distribution and early warning status of park visitors.
2. According to the digital twin-based park visitor safety early warning method according to claim 1, It is characterized in that The method of constructing a human identification model based on the video image data in the park includes: S11, decomposing the video data of the same scene frame by frame to obtain frame images; S12, dividing the frame image into a plurality of grids, and monitoring the rectangular area containing the human body in the grids; S13, determining the grid to which the visitor belongs according to the center position coordinates of the rectangular area; S14, counting the visitors in the video data of all scenes to establish a crowd recognition model.
3. According to the digital twin-based park visitor safety early warning method of claim 1, It is characterized in that The method of constructing a behavior recognition model based on video image data containing visitor targets includes: S21, decomposing the video image data containing the visitor target by frame, and marking and distinguishing different visitor targets respectively; S22, extracting visitor target features, analyzing the movement trajectory of the visitor target between each frame, determining the continuous trajectory information of the visitor target, and building a behavior recognition model; S23, predicting the body movements and movement positions of the visitor target according to the continuous trajectory information.
4. According to claim 3, a digital twin-based park visitor safety early warning method, It is characterized in that The step S23 comprises: Based on the continuous trajectory information, the body movements and movement distance of the visitor target are identified, and the behavioral tendencies and target positions of the visitor target are determined in combination with the park entity characteristics around the visitor target, and then the body movements and movement positions of the visitor target in the next frame or the next time period are predicted.
5. According to claim 1, a digital twin-based park visitor safety early warning method, It is characterized in that The calculation of the suspicious identification index based on the behavior analysis to determine whether to trigger an abnormal behavior warning includes: Determine the types of abnormal behaviors of visitors, prioritize them, and then assign corresponding weight factors to them. Calculate the suspicious identification index based on the types of abnormal behaviors of visitors and the corresponding weight factors. If the calculated result is greater than the threshold index, an abnormal behavior warning is triggered.
6. According to claim 5, a digital twin-based park visitor safety early warning method, It is characterized in that The calculation formula of the suspicious identification index is: X=Math.floor(a*C+b*P+c*Z+d*F+e*L+f*W / (a+b+c+d+e+f)) Where X is the calculated suspicious identification index, C, P, Z, F, L, and W are the quantitative values of illegally entering restricted areas, wandering, looking around, climbing over walls, face occlusion, and tailing, respectively, a, b, c, d, e, and f are weight factors, and Math.floor() means rounding down decimals.
7. According to claim 1, a digital twin-based park visitor safety early warning method, It is characterized in that The construction of the park-based digital twin platform includes: Obtain relevant data on the park’s building space, cameras, and access control locations. The data format includes architectural CAD drawings and on-site photos. Based on the collected data, perform three-dimensional visual modeling of the park and establish a virtual park digital twin corresponding to the physical entity.
8. A park visitor safety warning method based on digital twins according to any one of claims 1 to 7, It is characterized in that Before step S3, the following steps are also included: The data containing the number of people and the behavior analysis are encrypted separately, and after the encrypted data is sent to the destination, the data is restored and analyzed. The encryption algorithm uses the SHA algorithm or the SM3 algorithm.
9. A digital twin-based park visitor safety warning system, using a digital twin-based park visitor safety warning method according to any one of claims 1 to 8, It is characterized in that include: The crowd recognition module uses the constructed crowd recognition model to analyze and output the number of people in the area; Behavior recognition module, which obtains the behavior analysis of the corresponding visitor target based on the behavior recognition model; The early warning module determines whether to trigger an early warning based on the number of people and behavior analysis; The three-dimensional display module shows the distribution of park visitors through a virtual digital twin of the park.
10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the campus visitor safety warning method based on digital twins as described in any one of claims 1 to 8 are executed.
Citation Information
Cited By
Park security and protection method and system based on image recognition
CN119888620A
Park visual safety positioning real-time tracking method fusing multiple sensors
CN120008588A
A real-time tracking method for campus visual security positioning based on multi-sensor integration
CN120008588B
Hydropower station personnel safety early warning method and system based on digital twinning
CN120472247A
Bid evaluation base total-factor real-time linkage supervision method and system
CN120806858A
