Community monitoring system

Abstract

The invention discloses a community monitoring system. According to the invention, the behavior identification module can identify various abnormal behaviors in real time, such as fighting, suspicious wandering, crowd gathering and the like, and immediately triggers an alarm to notify a manager to deal with the abnormal behaviors, so that the occurrence of safety accidents is effectively prevented. And the behavior modeling module can identify potential criminal behaviors by establishing normal and abnormal behavior models, so that the crime rate is further reduced. And meanwhile, the behavior trajectory analysis module can track personnel trajectories in real time and quickly position personnel positions in an emergency, so that the emergency response capability is improved, and the safety of residents is ensured. And the behavior analysis module can effectively improve the management efficiency of the cell. Through automatic behavior analysis, the behavior analysis module can reduce the manual workload, and managers can concentrate on key tasks, thereby improving the management efficiency. A behavior analysis report is provided for managers, the managers are helped to know the safety condition of the cell, and a more effective safety management strategy is formulated.

Description

Technical Field

[0001] The present invention belongs to the technical field of residential area security management, and in particular relates to a residential area monitoring system. Background Art

[0002] Community security management refers to the effective maintenance and protection of residents' living order, property safety, and public safety within residential communities through a series of measures and systems. This includes access control systems, video surveillance, security patrols, fire safety, and emergency response plans. Community security management aims to create a safe, comfortable, and harmonious living environment and safeguard the legitimate rights and interests of residents. Community security monitoring utilizes modern technology to monitor and record residential communities in real time to prevent and deter criminal activity and protect the personal and property safety of residents. This system typically includes closed-circuit television (CCTV), electronic patrol systems, and intrusion alarm systems. CCTV monitors public areas around the clock through cameras installed in strategic locations within the community; electronic patrol systems ensure that security personnel complete patrols within specified timeframes; and intrusion alarm systems promptly issue alerts upon detecting unauthorized intrusions. The efficient operation of community security monitoring systems not only improves community security management and provides residents with a more secure living environment, but also provides important clues and evidence for public security agencies in combating crime.

[0003] However, traditional video surveillance systems lack intelligent analysis capabilities and cannot automatically identify abnormal behavior. They rely on manual review of recorded footage, which is time-consuming and labor-intensive. Furthermore, traditional systems lack monitoring of residential environmental parameters, making it impossible to promptly identify environmental issues that could impact resident health. Summary of the Invention

[0004] The purpose of the present invention is to provide a cell monitoring system in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: a community monitoring system, the system comprising: a face recognition access control integrated machine module, a monitoring camera module, a behavior analysis module, a community monitoring platform, and an auxiliary equipment module;

[0006] The behavior analysis module is internally provided with a behavior recognition module, a behavior modeling module and a behavior trajectory analysis module;

[0007] The auxiliary equipment module is internally provided with an environmental monitoring sensor module, an information publishing module and a mobile application module;

[0008] The face recognition access control integrated machine module is connected to the community monitoring platform via a wired or wireless network.

[0009] The monitoring camera module is also connected to the platform via the network to upload video recordings and snapshots.

[0010] The behavior analysis module is connected to the platform via the network, uploads the behavior analysis results, and accepts the platform's rule configuration and analysis results viewing.

[0011] The cell monitoring platform connects various modules through the network, collects and manages all data, and performs data analysis and alarm processing;

[0012] The auxiliary equipment module includes an environment monitoring sensor module, an information release module and a mobile application module connected to the platform through a network.

[0013] In a preferred embodiment, the face recognition and access control integrated machine module adopts the DS-K1 T342M series face recognition and access control integrated machine.

[0014] In a preferred embodiment, the surveillance camera module adopts a DS-2CD2245C(D)V6-L series surveillance camera.

[0015] In a preferred embodiment, the YOLOv4 algorithm of the behavior recognition module detects the location of people in the community and consists of five parts, namely:

[0016] A. Coordinate loss (center coordinate):

[0017] For each bounding box in each grid, if the bounding box is responsible for predicting the object (i.e., I_ij_obj is 1), the squared difference between the predicted center coordinates (x_i, y_i) and the true center coordinates (_hat{x}_i, _hat{y}_i) is calculated and these squared differences are summed.

[0018] This part of the loss is multiplied by the coordinate loss weight λ_coord.

[0019] B. Coordinate loss (width and height):

[0020] For the bounding box responsible for predicting the target, the squared difference between the predicted width and height (wi,hi) and the true width and height (hat{w}_i,hat{h}_i) is calculated and summed. This part of the loss is also multiplied by the coordinate loss weight λ_coord.

[0021] C. Confidence loss (with target):

[0022] For the bounding box responsible for predicting the target, calculate the squared difference between the predicted confidence C_i and the true confidence _hat{C}_i and sum these squared differences.

[0023] D. Confidence loss (no target):

[0024] For bounding boxes that do not contain an object (i.e., I_ij_noobj is 1), calculate the squared difference between the predicted confidence C_i and the true confidence _hat{C}_i and sum these squared differences.

[0025] This part of the loss is multiplied by the no-object bounding box loss weight λ_noobj.

[0026] E. Class prediction loss:

[0027] For the grid containing the target, calculate the predicted class probability distribution and the true category probability distribution The squared difference is calculated and summed over all categories.

[0028] Adding these five losses together gives the total loss function value. By minimizing this loss function value, the YOLOv4 model can be trained to improve the accuracy of object detection.

[0029] In a preferred embodiment, the calculation formula of the One-ClassSVM algorithm of the behavior modeling module is:

[0030]

[0031] subjectto:w T φ(x i )-ρ≤1-ξ i ,ξ i ≥0,i=1,...,n

[0032] Where w represents the normal vector of the separating hyperplane.

[0033] ξi represents the slack variable.

[0034] ρ represents the offset.

[0035] ν represents the parameter that controls the proportion of outliers.

[0036] n represents the number of samples.

[0037] φ(xi) represents the function that maps samples to high-dimensional space.

[0038] In a preferred embodiment, the density calculation formula of the behavior trajectory analysis module is:

[0039] Density(p)=|N_Eps(p)|

[0040] Where Density(p) represents the density of point p, N_Eps(p) represents the set of trajectory points in the Eps neighborhood of point p, and |N_Eps(p)| represents the number of trajectory points in the set.

[0041] The core point judgment condition formula is:

[0042] if Density(p)>=MinPts then

[0043] p is a core point

[0044] Among them, MinPts represents the minimum number of samples.

[0045] In a preferred embodiment, the cell monitoring platform includes: a server, client software, and network equipment.

[0046] In a preferred embodiment, the environmental monitoring sensor module includes: a sensor module, a data collector module, and an early warning system module.

[0047] In a preferred embodiment, the information publishing module includes: an information publishing platform, a publishing device, and an interactive function module.

[0048] In a preferred embodiment, the mobile application module includes a mobile application module, a data encryption transmission module, and a user management module.

[0049] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0050] 1. In the present invention, the behavior analysis module can significantly improve the security of the community. The behavior recognition module can identify various abnormal behaviors in real time, such as fighting, suspicious wandering, crowds, etc., and immediately trigger an alarm to notify management personnel to handle it, thereby effectively preventing the occurrence of safety accidents. The behavior modeling module can identify potential criminal behaviors by establishing normal and abnormal behavior models, further reducing the crime rate. The behavior trajectory analysis module can track the trajectory of personnel in real time and quickly locate the position of personnel in an emergency, improving emergency response capabilities and ensuring the safety of residents. The behavior analysis module can effectively improve the management efficiency of the community. Through automated behavior analysis, the behavior analysis module can reduce manual workload, allowing managers to focus more on key tasks, thereby improving management efficiency. The behavior analysis module can present the analysis results in the form of charts, heat maps, etc., which makes it convenient for managers to intuitively understand the safety status of the community and make decisions. The behavior analysis module can also provide management personnel with behavior analysis reports to help them understand the safety status of the community and formulate more effective safety management strategies.

[0051] 2. In this invention, the behavior analysis module can also enhance residents' sense of security and promote community harmony. By promptly detecting and addressing abnormal behavior, the behavior analysis module can provide residents with greater peace of mind and strengthen their confidence in community safety. Furthermore, the behavior analysis module can make its analysis results public to residents, strengthening their trust in community management and promoting community harmony.

[0052] 3. Finally, in the present invention, the behavior analysis module can also optimize the resource allocation of the community. By analyzing the activity trajectories of personnel, the behavior analysis module can help managers optimize patrol routes and improve patrol efficiency. At the same time, the behavior analysis module can help managers understand the security status of the community, rationally allocate police force, and improve the efficiency of police force utilization. The behavior analysis module of this system achieves comprehensive monitoring and analysis of community personnel behavior through the internal behavior recognition module, behavior modeling module, and behavior trajectory analysis module, thereby improving the safety and management efficiency of the community, enhancing residents' sense of security, optimizing resource allocation, and promoting community harmony. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a block diagram of the overall system of the present invention;

[0054] Figure 2 This is a system block diagram of the behavior analysis module in the present invention;

[0055] Figure 3 This is a system block diagram of the auxiliary equipment module in the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0057] Example:

[0058] Reference Figure 1-3 , a community monitoring system, the system includes: a face recognition access control integrated machine module, a monitoring camera module, a behavior analysis module, a community monitoring platform, and an auxiliary equipment module;

[0059] The behavior analysis module is internally configured with a behavior recognition module, a behavior modeling module, and a behavior trajectory analysis module;

[0060] The auxiliary equipment module is internally provided with an environmental monitoring sensor module, an information release module and a mobile application module;

[0061] The facial recognition access control integrated machine module is connected to the community monitoring platform via a wired or wireless network, uploads data such as entry and exit records and captured photos, and accepts configuration and management from the platform.

[0062] The surveillance camera module is also connected to the platform through the network, uploading video recordings and captured photos, and accepting parameter settings and video monitoring from the platform.

[0063] As an extended function of the surveillance camera module, the behavior analysis module is connected to the platform through the network, uploads the behavior analysis results, and accepts the platform's rule configuration and analysis results viewing.

[0064] As the core of the system, the community monitoring platform connects various modules through the network, collects and manages all data, performs data analysis and alarm processing, and feeds back the processing results to other modules, such as sending alarm information, publishing information, etc.

[0065] Auxiliary equipment modules such as environmental monitoring sensor modules, information release modules and mobile application modules are also connected to the platform through the network to upload monitoring data, release information and receive alarm information, and accept the configuration and management of the platform.

[0066] The facial recognition access control module is a core component of the residential surveillance system, responsible for the security management of the residential entrances and exits. This module uses the DS-K1 T342M series facial recognition access control module, which has the following features:

[0067] High-performance face recognition: Based on deep learning algorithm, it has fast recognition speed and high accuracy, supports 1:N face recognition, recognition speed is less than 0.2 seconds, and recognition accuracy is greater than 99%.

[0068] Multiple authentication methods: supports face recognition, card swiping (Mifare card / IC card, mobile phone NFC card, CPU card serial number, ID card serial number), password, fingerprint, QR code and other authentication methods, and supports multiple combined authentication to meet the needs of places with different security levels.

[0069] Video intercom function: supports video intercom with Yunmo, Internet APP, 4200 client, indoor unit, and management machine, making it convenient for visitors and residents to communicate.

[0070] Video preview function: supports remote video preview in the management center, allowing managers to view entrance and exit conditions in real time.

[0071] Mask detection function: supports mask detection mode, can be configured with reminder mode or mandatory mask wearing mode, associated with access control, to ensure the safety of community epidemic prevention and control.

[0072] Flexible configuration: The recognition interface is configurable, supporting the configuration of whether button icons such as "call", "QR code", and "password" are displayed; the authentication result display is configurable, supporting the configuration of whether information such as "photo", "name", and "work number" are displayed, and supports the desensitized display of name and work number; the authentication result voice is customizable, supporting the configuration of voice broadcast of authentication success and authentication failure, and supporting the superimposed broadcast of name.

[0073] Multiple working modes: support standby advertising mode, simple mode, theme mode, which can be selected according to actual needs.

[0074] Safe and reliable: supports external security modules to prevent the host from being maliciously damaged; supports card reader mode and can be used as a card reader; supports access control plan template management and supports normally open and normally closed time period management; supports multiple authentications to improve security; supports blacklist verification and identification of blacklisted personnel; supports alarm functions such as anti-tampering alarm, door opening alarm by external force, duress card and duress password alarm, blacklist alarm, etc.; supports event upload, and uploads authentication results and linked snapshot photos to the platform in real time; supports stand-alone use and can be managed locally; supports WEB management and can perform personnel management, parameter configuration, event query, system maintenance and other operations.

[0075] The surveillance camera module is responsible for security monitoring inside and around the community, real-time monitoring of the community environment, and recording video footage. This module uses the DS-2CD2245C(D)V6-L series surveillance camera and has the following features:

[0076] High-definition image quality: The maximum resolution can reach 2560×1440@25fps, providing clear video images.

[0077] Strong adaptability: Supports SmartIR to prevent infrared overexposure at night; supports backlight compensation, strong light suppression, 3D digital noise reduction, digital wide dynamic, and adapts to different usage environments.

[0078] Multiple functions: Supports ROI region of interest enhanced coding to improve video encoding efficiency; supports open network video interface, ISAPI, SDK, GB28181 protocol, and supports access to the EZVIZ platform; built-in microphone supports audio capture; supports intelligent fill light, supports white light / infrared dual fill light, infrared light can reach up to 30m, white light can reach up to 30m; meets IP66 dust and waterproof design, and has high reliability.

[0079] Multiple configurations: Supports multiple lens focal lengths, allowing you to choose the appropriate lens based on actual needs; supports multiple video compression standards and bit rate control methods; supports audio compression standards and bit rates; supports multiple network protocols and interface protocols; supports multiple image settings and functions; supports multiple power supply methods and certifications.

[0080] The behavior recognition module uses the YOLOv4 algorithm to detect the location of residents in the community, processes the video stream transmitted by the video surveillance module in real time, and uses image recognition and computer vision technology to analyze the behavior of residents in the community in the video, identifying specific behaviors such as illegal operation, not wearing a helmet, and entering prohibited areas;

[0081] The behavior recognition module is used in the following ways:

[0082] Input: sequence of video frames.

[0083] Processing: Image preprocessing: denoising, enhancement.

[0084] Target detection: Use the YOLOv4 algorithm to detect the location of people in the community.

[0085] Behavior classification: Based on the detected location of people in the community, a CNN+LSTM network is used to classify behaviors.

[0086] Output: Identified behavior category and location information.

[0087] The YOLOv4 algorithm detects the location of people in a community and consists of five parts:

[0088] A. Coordinate loss (center coordinate):

[0089] For each bounding box in each grid, if the bounding box is responsible for predicting the object (i.e., I_ij_obj is 1), the squared difference between the predicted center coordinates (x_i, y_i) and the true center coordinates (_hat{x}_i, _hat{y}_i) is calculated and these squared differences are summed.

[0090] This part of the loss is multiplied by the coordinate loss weight λ_coord.

[0091] B. Coordinate loss (width and height):

[0092] For the bounding box responsible for predicting the target, the squared difference between the predicted width and height (wi,hi) and the true width and height (hat{w}_i,hat{h}_i) is calculated and summed. This part of the loss is also multiplied by the coordinate loss weight λ_coord.

[0093] C. Confidence loss (with target):

[0094] For the bounding box responsible for predicting the target, calculate the squared difference between the predicted confidence C_i and the true confidence _hat{C}_i and sum these squared differences.

[0095] D. Confidence loss (no target):

[0096] For bounding boxes that do not contain an object (i.e., I_ij_noobj is 1), calculate the squared difference between the predicted confidence C_i and the true confidence _hat{C}_i and sum these squared differences.

[0097] This part of the loss is multiplied by the no-object bounding box loss weight λ_noobj.

[0098] E. Class prediction loss:

[0099] For the grid containing the target, calculate the predicted class probability distribution and the true category probability distribution The squared difference is calculated and summed over all categories.

[0100] Adding these five losses together gives the total loss function value. By minimizing this loss function value, the YOLOv4 model can be trained to improve the accuracy of object detection.

[0101] The behavior modeling module uses the One-ClassSVM algorithm to establish normal and abnormal behavior models based on historical cell monitoring data, providing a comparison basis for behavior identification.

[0102] The behavior modeling module specifically includes:

[0103] Input: Historical behavioral cell monitoring data.

[0104] deal with:

[0105] Community monitoring data annotation: Behavior annotation of historical community monitoring data.

[0106] Feature extraction: Use autoencoders to extract behavioral features.

[0107] Model training: Use One-ClassSVM to train the normal behavior model.

[0108] Output: Normal behavior model.

[0109] The calculation formula of the One-ClassSVM algorithm is:

[0110]

[0111] subjectto:w T φ(x i )-ρ≤1-ξ i ,ξ i ≥0,i=1,...,n

[0112] Where w represents the normal vector of the separating hyperplane.

[0113] ξi represents the slack variable.

[0114] ρ represents the offset.

[0115] ν represents the parameter that controls the proportion of outliers.

[0116] n represents the number of samples.

[0117] φ(xi) represents the function that maps samples to high-dimensional space.

[0118] The behavior trajectory analysis module uses the density-based trajectory clustering algorithm (DBSCAN) algorithm to construct trajectories, integrate behavior recognition, analyze trajectories, and output results based on the collected location information;

[0119] The specific steps include:

[0120] a. Data preprocessing:

[0121] Data cleaning: remove outliers and noise data to ensure data quality.

[0122] Data standardization: standardize the location coordinates and unify the data scale.

[0123] b. Trajectory construction:

[0124] Data sorting: Sort location information by timestamp.

[0125] Trajectory representation: The sorted position information is represented as a trajectory, which is a series of points or path segments.

[0126] c. Behavior recognition fusion:

[0127] Behavior recognition: The behavior type of each trajectory point or path segment is identified through a behavior recognition module (such as a pedestrian behavior recognition algorithm based on deep learning).

[0128] Data fusion: The identified behavior type is combined with the trajectory data to label the behavior type for each trajectory point or path segment.

[0129] d. Trajectory cluster analysis (DBSCAN algorithm):

[0130] Parameter setting: Set the parameters of the DBSCAN algorithm, including the neighborhood radius (Eps) and the minimum number of samples (MinPts).

[0131] Density calculation: For each trajectory point, calculate the number of trajectory points in its Eps neighborhood.

[0132] Core point identification: If the number of trajectory points in the Eps neighborhood of a trajectory point is greater than or equal to MinPts, the trajectory point is marked as a core point.

[0133] Cluster expansion: Starting from the core point, adjacent core points are connected through density reachability to form clusters. Noise point processing: Trajectory points that do not belong to any cluster are marked as noise points.

[0134] e. Trajectory analysis:

[0135] Abnormal trajectory identification: Analyze clustering results to identify isolated points or small clusters, which may indicate abnormal trajectories or abnormal behaviors.

[0136] Frequent activity area identification: Analyze large clusters and identify frequent activity areas.

[0137] f. Result output:

[0138] Trajectory visualization: Display the trajectory and clustering results in the form of graphs.

[0139] Abnormal behavior report: Generate abnormal behavior reports, including abnormal tracks, frequent activity areas and other information.

[0140] The density calculation formula is:

[0141] Density(p)=|N_Eps(p)|

[0142] Where Density(p) represents the density of point p, N_Eps(p) represents the set of trajectory points in the Eps neighborhood of point p, and |N_Eps(p)| represents the number of trajectory points in the set.

[0143] The core point judgment condition formula is:

[0144] if Density(p)>=MinPts then

[0145] p is a core point

[0146] Among them, MinPts represents the minimum number of samples.

[0147] The residential surveillance platform is responsible for collecting and managing data from facial recognition access control devices, surveillance cameras, behavior analysis modules, and other devices, and performs operations such as data analysis, alarm processing, and video playback. The platform consists of the following components:

[0148] Server: stores data such as face database, card information, fingerprint information, event records, video recordings, etc. The server needs to have sufficient storage space and computing power to ensure data security and reliability.

[0149] Client software: Provides a user interface for personnel management, equipment management, event query, video playback, and other operations. The client software needs to be user-friendly and easy to use for management personnel.

[0150] Network equipment: Routers, switches, etc., responsible for network connections and data transmission. Network equipment needs to be stable and reliable to ensure data transmission efficiency.

[0151] The environmental monitoring sensor module monitors environmental data within the community, such as temperature, humidity, air quality, etc., and issues timely warnings to protect the health of residents. This module consists of the following parts:

[0152] Sensors: Sensors that measure parameters such as temperature, humidity, and air quality, such as temperature and humidity sensors, air quality sensors, etc. These sensors can be installed in different locations within the community, such as green belts, underground garages, residential buildings, etc., to comprehensively monitor the community environment.

[0153] Data collector: transmits sensor data to the monitoring platform, such as wireless data collector, wired data collector, etc. The data collector can collect sensor data at regular intervals and transmit it to the monitoring platform for analysis and processing.

[0154] Early warning system: Set early warning thresholds based on monitoring data and promptly notify management personnel or residents, such as SMS warnings, APP push notifications, etc. When the monitoring data exceeds the early warning threshold, the early warning system will automatically send an early warning message to remind relevant personnel to take measures.

[0155] The information release module is responsible for publishing information at the entrances and exits of the community and in public areas, such as weather forecasts, community activities, notices, etc., to facilitate residents to obtain information. This module consists of the following parts:

[0156] Information publishing platform: used to edit and manage published information, such as PC information publishing platform, mobile information publishing platform, etc. The information publishing platform can facilitate managers to publish various information and manage them.

[0157] Publishing equipment: used to display published information, such as LED display screens, touch screens, community bulletin boards, etc. Publishing equipment can be installed in different locations within the community, such as entrances and exits, public areas, etc., to facilitate residents to view information.

[0158] Interactive functions: support residents to query information, leave messages, etc., such as touch screen interactive query, community forum, etc. Interactive functions can enhance resident participation and improve community management efficiency.

[0159] The mobile application module allows residents and property management personnel to view surveillance images, receive alarm information, and conduct remote management at any time. This module consists of the following parts:

[0160] Mobile application: Provides functions such as video preview, event query, alarm push, device management, information release, and community interaction. Residents and property management personnel can use the mobile application to view surveillance footage at any time to understand the community's security status; receive alarm information and handle emergencies promptly; manage devices, such as checking device status and setting device parameters; release information, such as announcements and community event information; and interact with the community, such as leaving messages and participating in community activities.

[0161] Data encryption transmission: Ensure data transmission security and prevent information leakage. The mobile application uses data encryption transmission technology to protect user information security.

[0162] User management: Set permissions for different users to prevent unauthorized operations. Mobile applications can set permissions for different users, such as administrators, operators, and ordinary users, to prevent unauthorized operations.

[0163] From the above, it can be seen that the behavior analysis module in the present invention can significantly improve the safety of residential communities. The behavior recognition module can identify various abnormal behaviors in real time, such as fighting, suspicious wandering, and crowds, and immediately trigger an alarm to notify management personnel to take action, thereby effectively preventing the occurrence of safety accidents. The behavior modeling module, by establishing models of normal and abnormal behavior, can identify potential criminal behavior and further reduce the crime rate. The behavior trajectory analysis module can track the trajectory of individuals in real time and quickly locate their location in emergency situations, improving emergency response capabilities and ensuring the safety of residents. The behavior analysis module can effectively improve the management efficiency of residential communities. Through automated behavior analysis, the behavior analysis module can reduce manual workload, allowing management personnel to focus more on key tasks, thereby improving management efficiency. The behavior analysis module can present analysis results in the form of charts, heat maps, etc., allowing management personnel to intuitively understand the community security situation and make decisions. The behavior analysis module can also provide management personnel with behavior analysis reports to help them understand the community security situation and formulate more effective security management strategies.

[0164] In this invention, the behavior analysis module can also enhance residents' sense of security and promote community harmony. By promptly detecting and addressing abnormal behavior, the behavior analysis module can provide residents with greater peace of mind and strengthen their confidence in the safety of the community. Furthermore, the behavior analysis module can make the analysis results public to residents, strengthening their trust in community management and promoting community harmony.

[0165] Finally, in the present invention, the behavior analysis module can also optimize the resource allocation of the community. By analyzing the activity trajectories of personnel, the behavior analysis module can help managers optimize patrol routes and improve patrol efficiency. At the same time, the behavior analysis module can help managers understand the security status of the community, rationally allocate police force, and improve the efficiency of police force utilization. Through the internal behavior recognition module, behavior modeling module, and behavior trajectory analysis module, the behavior analysis module of this system achieves comprehensive monitoring and analysis of community personnel behavior, improving the safety and management efficiency of the community, enhancing residents' sense of security, optimizing resource allocation, and promoting community harmony.

[0166] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements that are inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0167] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A cell monitoring system, characterized by: The system includes: a face recognition access control integrated machine module, a surveillance camera module, a behavior analysis module, a community monitoring platform, and an auxiliary equipment module; The behavior analysis module is internally provided with a behavior recognition module, a behavior modeling module and a behavior trajectory analysis module; The auxiliary equipment module is internally provided with an environmental monitoring sensor module, an information publishing module and a mobile application module; The face recognition access control integrated machine module is connected to the community monitoring platform via a wired or wireless network; The surveillance camera module is also connected to the platform via the network to upload video recordings and snapshots; The behavior analysis module is connected to the platform via the network, uploads the behavior analysis results, and accepts the platform's rule configuration and analysis results viewing; The cell monitoring platform connects various modules through the network, collects and manages all data, and performs data analysis and alarm processing; The auxiliary equipment module includes an environment monitoring sensor module, an information release module and a mobile application module connected to the platform through a network.

2. A cell monitoring system according to claim 1, characterized in that: The face recognition access control integrated machine module adopts the DS-K1T342M series face recognition access control integrated machine.

3. A cell monitoring system according to claim 1, characterized in that: The monitoring camera module adopts the DS-2CD2245C(D)V6-L series monitoring camera.

4. A cell monitoring system according to claim 1, characterized in that: The YOLOv4 algorithm of the behavior recognition module detects the location of people in the community and consists of five parts: A. Coordinate loss (center coordinate): For each bounding box in each grid, if the bounding box is responsible for predicting the target (i.e., I_ij_obj is 1), calculate the squared difference between the predicted center coordinates (x_i, y_i) and the true center coordinates (_hat{x}_i, _hat{y}_i), and sum these squared differences; This part of the loss is multiplied by the coordinate loss weight λ_coord; B. Coordinate loss (width and height): For the bounding box responsible for predicting the target, calculate the square difference between the predicted width and height (w_i, h_i) and the true width and height (_hat{w}_i, _hat{h}_i), and sum these square differences; this part of the loss is also multiplied by the coordinate loss weight λ_coord; C. Confidence loss (with target): For the bounding box responsible for predicting the target, calculate the squared difference between the predicted confidence C_i and the true confidence _hat{C}_i, and sum these squared differences; D. Confidence loss (no target): For the bounding box that does not contain the target (i.e., I_ij_noobj is 1), calculate the square difference between the predicted confidence C_i and the true confidence _hat{C}_i, and sum these square differences; This part of the loss is multiplied by the no-object bounding box loss weight λ_noobj; E. Class prediction loss: For the grid containing the target, calculate the predicted class probability distribution and the true category probability distribution _hat{p}_i The squared difference is calculated and summed over all categories; Add up these five parts of loss to get the total loss function value; by minimizing this loss function value, the YOLOv4 model can be trained to improve the accuracy of target detection.

5. A cell monitoring system according to claim 1, characterized in that: The calculation formula of the One-ClassSVM algorithm of the behavior modeling module is: subjectto:w T φ(x i )-ρ≤1-ξ i ,ξ i ≥0,i=1,...,n Where w represents the normal vector of the separating hyperplane; ξi represents the slack variable; ρ represents the offset; ν represents the parameter that controls the proportion of outliers; n represents the sample size; φ(xi) represents the function that maps samples to high-dimensional space.

6. A cell monitoring system according to claim 1, characterized in that: The density calculation formula of the behavior trajectory analysis module is: Density(p)=|N_Eps(p)| Where Density(p) represents the density of point p, N_Eps(p) represents the set of trajectory points in the Eps neighborhood of point p, and |N_Eps(p)| represents the number of trajectory points in the set; The core point judgment condition formula is: if Density(p)>=MinPts then p is a core point Among them, MinPts represents the minimum number of samples.

7. A cell monitoring system according to claim 1, characterized in that: The cell monitoring platform includes: server, client software, and network equipment.

8. A cell monitoring system according to claim 1, characterized in that: The environmental monitoring sensor module includes: a sensor module, a data collector module, and an early warning system module.

9. A cell monitoring system according to claim 1, characterized in that: The information publishing module includes: an information publishing platform, a publishing device, and an interactive function module.

10. The cell monitoring system according to claim 1, wherein: The mobile application module includes a mobile application module, a data encryption transmission module, and a user management module.

Images