Intelligent community access control security management method and system

Abstract

The invention relates to the technical field of intelligent access control, and relates to an intelligent community access control security management method and system, the method comprises area supervision and resident security: area supervision comprises: collecting community environment data and event rating; inputting the community environment data and the event rating into a pre-constructed security level prediction model, and outputting a security level; the resident security protection comprises the following steps: collecting historical access data and daily behavior characteristics of residents; training a matching quantity prediction model by using the historical access data and the daily behavior characteristics; collecting a current access position and a current time point; inputting the current access position and the current time point into a matching quantity prediction model, and outputting a feature matching quantity and N; activating a face card reader according to the numerical value N; and adjusting the feature matching quantity of the face card reader according to the feature matching quantity output by the matching quantity prediction model. According to the invention, the identity of a resident can be quickly verified when the resident approaches the entrance guard, so that the smoothness and convenience of passing are greatly improved.

Description

Technical Field

[0001] The present application relates to the field of smart access control technology, and in particular to a smart community access control security management method and system. Background Art

[0002] In modern cities, the security management of smart communities is particularly important, especially in terms of ensuring the safety of residents and improving traffic efficiency. For example, the Chinese patent with publication number CN116597546A discloses a community security integration system and method based on artificial intelligence. The invention relates to the field of community security technology, including a data transmission switch and an information collection system, a monitoring system, an identity recognition system and a personnel gate. The information collection system is used to collect the identity information of community personnel, the monitoring system is used to monitor the entry and exit of personnel in the access zone, and the identity recognition system is used to identify the identity information of personnel entering the access zone. The information collection system, the monitoring system and the identity recognition system transmit the information to the central management platform through the data transmission switch. The community security integration system and method based on artificial intelligence can improve the security effect of the community access control system. When a stranger follows the person in front to enter the community, an early warning signal can be issued in time to remind the guard, so that the guard can register and check the stranger, avoiding the stranger from entering the community with a card. However, traditional security systems often have problems such as slow recognition speed and long response time, which often cause inconvenience to residents when entering and leaving the community, especially during peak hours or in emergency situations, which can easily cause safety hazards. Summary of the invention

[0003] In order to solve the above problems, an embodiment of the present invention provides a smart community access control and security management method, which includes regional supervision and resident security: Regional supervision includes: Collect community environment data and event ratings; Input community environment data and event ratings into the pre-built security level prediction model to output the security level; Resident security includes: Collect residents’ historical entry and exit data and daily behavior characteristics; Use historical entry and exit data and daily behavior characteristics to train a matching quantity prediction model; Collect the current entry and exit location and current time point; Input the current entry and exit position and the current time point into the matching quantity prediction model, and output the feature matching quantity and N; Activate the face reader according to the value N; Adjust the feature matching amount of the face reader according to the feature matching amount output by the matching amount prediction model; Collect facial feature images from the face card reader and surveillance images; The facial feature images and surveillance images are input into the pre-built human recognition model to output the alarm strategy; the alarm strategy includes not outputting the alarm and outputting the alarm. If the alarm strategy is to output the alarm, the alarm will be pushed to the community security and the access control will be automatically unlocked.

[0004] Furthermore, the training method of the security level prediction model includes: All community environment data and event ratings are used as inputs of a security level prediction model, which takes the security level corresponding to each set of community environment data and event rating predictions as output, the actual security level corresponding to each set of community environment data and event ratings as a prediction target, and minimizing the sum of the first prediction accuracies of all predicted security levels as a training target; the security level prediction model is trained until the sum of the first prediction accuracies reaches convergence, and the training is stopped; the security level prediction model is a decision tree model.

[0005] Furthermore, the training method of the matching quantity prediction model includes: All historical entry and exit data and daily behavior characteristics are used as inputs of a matching quantity prediction model, the matching quantity prediction model takes the feature matching quantity and N corresponding to the prediction of each group of historical entry and exit data and daily behavior characteristics as outputs, takes the actual feature matching quantity and N corresponding to each group of historical entry and exit data and daily behavior characteristics as prediction targets, and takes minimizing the sum of the second prediction accuracies of all predicted feature matching quantities and N as a training target; the matching quantity prediction model is trained until the sum of the second prediction accuracies reaches convergence, and the training is stopped; the matching quantity prediction model is a long short-term memory network model.

[0006] Furthermore, the method for activating the face card reader includes: The recognition range of the face reader is divided into M block areas of equal area, where N is less than or equal to M, and N≠0, M≠0, and one block area is one unit. When the sensor of the face reader detects that the resident walks to the Mth unit, the face reader is pre-activated. When the sensor of the face reader detects that the resident walks to the Nth unit, the face reader is activated.

[0007] Furthermore, the training method of the human body recognition model includes: All facial feature images and surveillance images are used as inputs of a human body recognition model; the human body recognition model uses the alarm strategy corresponding to the prediction of each group of facial feature images and surveillance images as output, uses the actual alarm strategy corresponding to each group of facial feature images and surveillance images as the prediction target, and uses minimizing the sum of the third prediction accuracies of all predicted alarm strategies as the training target; the human body recognition model is trained until the sum of the third prediction accuracies reaches convergence and the training is stopped; the human body recognition model is a convolutional neural network model.

[0008] An intelligent community access control and security management system includes a regional supervision system and a resident security system: The regional regulatory system includes: A first collection module, the first collection module is used to collect community environment data and event ratings; A security rating module, which is used to input community environment data and event ratings into a pre-built security level prediction model and output a security level; Resident security systems include: A second collection module, the second collection module is used to collect historical entry and exit data and daily behavior characteristics of residents; A model training module, wherein the model training module uses historical input and output data and daily behavior characteristics to train a matching quantity prediction model; A third collection module, the third collection module is used to collect the current entry and exit position and the current time point; A non-synaesthesia preparation module, wherein the non-synaesthesia preparation module is used to input the current entry and exit position and the current time point into the matching amount prediction model, and output the feature matching amount and N; An activation recognition module is used to activate the face card reader according to the value N; A matching amount prediction module, the matching amount prediction module is used to adjust the feature matching amount of the face card reader according to the feature matching amount output by the matching amount prediction model; A fourth collection module, the fourth collection module is used to collect facial feature pictures and surveillance pictures in the face card reader; Emergency alarm module, which is used to input facial feature images and surveillance images into a pre-built human recognition model and output an alarm strategy; the alarm strategy includes not outputting an alarm and outputting an alarm. If the alarm strategy is to output an alarm, the alarm is pushed to community security and the access control is automatically unlocked.

[0009] The technical effects and advantages of the smart community access control and security management method and system provided by the present invention are as follows: The present invention combines an intelligent recognition system with a prediction model, so that residents can quickly verify their identities when approaching the access control without stopping, which greatly improves the smoothness and convenience of passage and avoids the waiting and congestion common in traditional security measures. The present invention introduces multi-dimensional analysis of environmental data in the area (such as weather, noise, light, etc.) and event ratings, and can dynamically adjust the security level to ensure that the most appropriate security protection can be provided in different situations. For example, the security level can be appropriately reduced on sunny days, and automatically increased during special events or in high-noise environments, responding to external changes in real time. At the same time, the clarification of the security level can also make the verification method more intelligent and save the smoothness of passage. Through face recognition technology and Combined with surveillance images, the system can accurately determine the safety status of residents, and promptly issue an alarm when an abnormal situation occurs, automatically unlock the access control, ensure the rapid entry of rescue personnel, and avoid delays caused by manual operation; machine learning models (such as decision trees, long short-term memory networks, and convolutional neural networks) are constantly adapted and optimized, and can predict traffic characteristics and security needs more accurately based on the accumulated resident entry and exit data and behavioral characteristics, thereby improving the overall intelligence level of the system; by setting the activation logic of the face card reader, the device is only activated in specific areas within the recognition range, avoiding unnecessary energy consumption, effectively extending the service life of the equipment, and ensuring the response speed of the system under high traffic conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a flow chart of a smart community access control and security management method in Embodiment 1; Figure 2 This is a schematic diagram of the activation method of the face card reader in Example 1; Figure 3 This is a schematic diagram of the scene faced when the face card reader in Example 1 is activated; Figure 4 This is a connection diagram of an intelligent community access control and security management system in Example 2. DETAILED DESCRIPTION

[0011] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0012] Embodiment 1: See also Figure 1 As shown, this embodiment discloses a smart community access control and security management method, including regional supervision and resident security: Regional supervision includes: Collect community environment data and event ratings; Input community environment data and event ratings into the pre-built security level prediction model to output the security level; Resident security includes: Collect residents’ historical entry and exit data and daily behavior characteristics; Use historical entry and exit data and daily behavior characteristics to train a matching quantity prediction model; Collect the current entry and exit location and current time point; Input the current entry and exit position and the current time point into the matching quantity prediction model, and output the feature matching quantity and N; Activate the face reader according to the value N; Adjust the feature matching amount of the face reader according to the feature matching amount output by the matching amount prediction model; Collect facial feature images from the face card reader and surveillance images; The facial feature images and surveillance images are input into the pre-built human recognition model to output the alarm strategy; the alarm strategy includes not outputting the alarm and outputting the alarm. If the alarm strategy is to output the alarm, the alarm will be pushed to the community security and the access control will be automatically unlocked.

[0013] Community environment data refers to sensor data from the community, such as weather, noise level, and light intensity. On sunny days, the security level can be reduced accordingly, while on foggy days, the security level needs to be increased accordingly. When the noise level is huge, for example, if there is a construction unit working next to the community, the security level needs to be increased accordingly to avoid ignoring harmful events due to noise. Light intensity represents visibility. Night vision devices are generally not installed in normal communities or hospitals. Therefore, if the light intensity is weak, the possibility of harmful events being ignored will increase, so the security level needs to be increased. Event ratings are calculated using a weighted average method, where the weights come from community activities and security events. Community activities include nighttime activities, holidays, and major events. When there are loopholes in the community's internal security, the event rating needs to be increased to prevent similar incidents from occurring.

[0014] The security level is the result of the model output, which is used to indicate the security demand level of the current community. The security level is usually divided into several levels (such as low, medium, and high). The system will adjust the access control security strategy according to the security level (such as increasing monitoring, automatic locking, and increasing verification methods, etc.).

[0015] The training method of the security level prediction model includes: All community environment data and event ratings are used as the input of the security level prediction model. The security level prediction model takes the security level corresponding to each set of community environment data and event rating as output, the actual security level corresponding to each set of community environment data and event rating as the prediction target, and minimizing the sum of the first prediction accuracies of all predicted security levels as the training target.

[0016] Among them, the calculation formula for the first prediction accuracy is: ,in, is the number of each group of community environment data and event ratings, is the first prediction accuracy, For the The predicted security level corresponding to the community environment data and event rating, For the The actual security level corresponding to the community environment data and event rating is grouped; the security level prediction model is trained until the sum of the first prediction accuracies reaches convergence and the training is stopped; the security level prediction model is a decision tree model.

[0017] The historical entry and exit data of residents are used to analyze the residents' common entry and exit time points and locations, including the current entry and exit location and the current time point, to establish the residents' behavior patterns. The daily behavior characteristics are the residents' entry and exit frequency and entry and exit time periods, which are used to improve the details of the residents' behavior patterns. The combination of historical entry and exit data and daily behavior characteristics is used to find out the residents' physical time periods for passage, providing a data basis for the subsequent implementation of contactless passage.

[0018] The training method of the matching volume prediction model includes: All historical input and output data and daily behavior characteristics are used as the input of the matching quantity prediction model. The matching quantity prediction model takes the feature matching quantity and N corresponding to each set of historical input and output data and daily behavior characteristics as the output, and takes the actual feature matching quantity and N corresponding to each set of historical input and output data and daily behavior characteristics as the prediction target. The training target is to minimize the sum of the second prediction accuracies of all predicted feature matching quantities and N.

[0019] Among them, the calculation formula for the second prediction accuracy is: ,in, is the number of each set of historical input and output data and daily behavior characteristics, is the second prediction accuracy, For the The predicted feature matching amount and N corresponding to the historical input and output data and daily behavior characteristics, For the The actual feature matching quantity corresponding to the historical input and output data of the group and the daily behavior characteristics is N; the matching quantity prediction model is trained until the sum of the second prediction accuracy reaches convergence and the training is stopped; the matching quantity prediction model is a long short-term memory network model.

[0020] like Figure 2 As shown, the activation methods of the face card reader include: The recognition range of the face reader is divided into M block areas of equal area, where N is less than or equal to M, and N≠0, M≠0, and one block area is one unit. When the sensor of the face reader detects that the resident walks to the Mth unit, the face reader is pre-activated. When the sensor of the face reader detects that the resident walks to the Nth unit, the face reader is activated.

[0021] Pre-activation is the standby state, in which the power is still on and the ability to receive signals is maintained. It can start and respond quickly in a short time, but it is not completely turned on. By activating the face reader only when the resident approaches a specific range of the access control (the Nth unit), the face recognition device can be prevented from being in standby or working state for a long time, reducing energy consumption. When the resident approaches the Nth unit within the recognition range, the system can identify the identity and features of the resident in advance by activating the face reader. Combined with the feature matching amount output by the matching amount prediction model, it can be determined in time whether it is a legal resident, ensuring that only verified persons can enter the community. Activating the face reader at an appropriate distance in advance can ensure that the system has entered the standby or recognition state when the resident reaches the recognition range, improving the smoothness of passage and user experience, and reducing the waiting time of residents. If the face reader is activated directly when the resident just enters the recognition range, more facial feature readings will be required due to distance reasons, which may waste more time and cause the resident's non-synesthesia effect to deteriorate. At the same time, it also prevents the huge increase in feature matching when there are too many people. For example, when there are too many people and the face reader is not activated at an appropriate distance, that is, when multiple residents appear in the same picture, the following will appear. Figure 3 The scenario shown here will increase the recognition time, increase the system burden, and also cause the residents' synaesthesia effect to deteriorate; It should be noted that the working principle of face reader recognition is to take photos for recognition, so when the distance is far, miscellaneous features will appear (such as irrelevant features such as vehicles, notebooks and clothes), and the extracted features will also appear in a blurred visual state.

[0022] The current entry and exit location can be located by the positioning software (such as GPS) on the resident's mobile phone, and then linked with the face card reader. The positioning software and the distance sensor on the face card reader can jointly locate the resident's location more accurately, and it is also a solution to Figure 3 The key to the scenario shown.

[0023] The feature matching amount is the value output based on the working principle of the face reader. When the face reader is taking pictures, each photo (each frame) taken will be feature matched according to the photos stored in the face recognition database. For example, the features on the photo (face shape, eyes, eyebrows and nose bridge shape, etc.) will be compared one by one with the features on the photo stored in the face recognition database. The recognition is successful if the number of correct comparison results exceeds a certain proportion, and the comparison will not be a single feature of a single photo. The feature matching amount is the maximum feature matching amount when determining feature matching. When the feature matching amount is small, the matching speed will increase, thereby reducing the time of blocking residents. If the recognition is completed and the access control is unlocked before the user reaches the face reader, the illusion of seamless door opening can be created, that is, there is no synesthesia, which improves the smoothness of passage and user experience.

[0024] The training methods of the human recognition model include: All facial feature images and surveillance images are used as the input of the human body recognition model. The human body recognition model predicts the corresponding alarm strategy for each group of facial feature images and surveillance images as output, and uses the actual alarm strategy corresponding to each group of facial feature images and surveillance images as the prediction target. Minimizing the sum of the third prediction accuracies of all predicted alarm strategies is used as the training goal.

[0025] Among them, the calculation formula for the third prediction accuracy is: ,in, The number of each group of facial feature pictures and surveillance pictures, is the third prediction accuracy, For the The predicted alarm strategy corresponding to the facial feature image and the surveillance image, For the The actual alarm strategy corresponding to the group of facial feature pictures and the surveillance pictures is obtained; the human body recognition model is trained until the sum of the third prediction accuracy reaches convergence and the training is stopped; the human body recognition model is a convolutional neural network model.

[0026] Facial features are feature data used by the model for facial recognition. Monitoring images are used to check the dynamics of the human body and the crowd. The dynamics of the human body, such as shaking and falling and unrelieved vertical falls, and the dynamics of the crowd, such as doctors, emergency stretchers and emergency vehicles. The combination of facial features and monitoring images can determine whether the resident is in crisis, such as blue face and shaking and falling. If the resident is judged to be old, he may have an acute illness. At this time, an alarm is sent to the community security and the door control is automatically unlocked, allowing rescue personnel to enter unimpeded and preventing the security from forgetting to unlock the door control when they detect the critical condition of the resident and go to perform first aid, thereby delaying the best rescue time. If the appearance of a doctor and an emergency stretcher is identified, or even an emergency vehicle is detected, an alarm is sent to the community security and the door control is automatically unlocked. The automatic unlocking of the door control is to prevent the following scenarios: When the rescue vehicle arrives, if the access control is determined by security throughout the process, or if one access control unlocking method is followed completely without a quick emergency method such as automatic access control unlocking, two scenarios will occur.

[0027] The first scenario: The access control is entirely decided by the security guard. If the security guard panics, there may be mistakes in releasing the access control. The community access control is generally composed of a personnel passage and a vehicle passage. If the personnel passage is mistakenly released at this time, it will cause a huge waste of emergency time.

[0028] The second scenario: If one access control unlocking method is followed completely and there is no quick emergency method such as automatic access control unlocking, the security guard must go out to unlock the access control from the inside. The first scenario may occur and the security guard will also need time to unlock the access control from the inside, which will also cause a great waste of emergency time.

[0029] Since the area supervision and resident security in this embodiment are centered on execution efficiency in order to avoid synergy between residents, some special scenarios may be overlooked, so an alarm strategy is used to deal with special scenarios that may arise.

[0030] Embodiment 2: like Figure 4 As shown, based on the same inventive concept as the smart community access control and security management method in the aforementioned embodiment, the present application provides a smart community access control and security management system. The system and method embodiments in the present application are based on the same inventive concept. The system includes a regional supervision system and a resident security system: The regional regulatory system includes: The first collection module is used to collect community environment data and event ratings; Security rating module: The security rating module is used to input community environment data and event ratings into a pre-built security level prediction model and output the security level; Resident security systems include: The second collection module is used to collect the historical entry and exit data and daily behavior characteristics of the residents; Model training module: The model training module uses historical input and output data and daily behavior characteristics to train the matching quantity prediction model; A third collection module, the third collection module is used to collect the current entry and exit position and the current time point; The non-synaesthesia preparation module is used to input the current entry and exit position and the current time point into the matching amount prediction model, and output the feature matching amount and N; Activate the recognition module, the activation recognition module is used to activate the face card reader according to the value N; A matching quantity prediction module, which is used to adjust the feature matching quantity of the face card reader according to the feature matching quantity output by the matching quantity prediction model; A fourth collection module, which is used to collect facial feature images and surveillance-shot images from a face card reader; Emergency alarm module: The emergency alarm module is used to input facial feature images and surveillance images into a pre-built human recognition model and output an alarm strategy; the alarm strategy includes not outputting an alarm and outputting an alarm. If the alarm strategy is to output an alarm, the alarm will be pushed to community security and the access control will be automatically unlocked.

[0031] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0032] What has been described above is only a preferred specific implementation manner of the embodiments of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can make equivalent substitutions or changes according to the technical scheme and concept of the present application within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A smart community access control security management method, characterized in that: Methods include area supervision and resident security: Regional supervision includes: Collect community environment data and event ratings; Input community environment data and event ratings into the pre-built security level prediction model to output the security level; Resident security includes: Collect residents’ historical entry and exit data and daily behavior characteristics; Use historical entry and exit data and daily behavior characteristics to train a matching quantity prediction model; Collect the current entry and exit location and current time point; Input the current entry and exit position and the current time point into the matching quantity prediction model, and output the feature matching quantity and N; Activate the face reader according to the value N; Adjust the feature matching amount of the face reader according to the feature matching amount output by the matching amount prediction model; Collect facial feature images from the face card reader and surveillance images; The facial feature images and surveillance images are input into the pre-built human recognition model to output the alarm strategy; the alarm strategy includes not outputting the alarm and outputting the alarm. If the alarm strategy is to output the alarm, the alarm will be pushed to the community security and the access control will be automatically unlocked.

2. The method according to claim 1, characterized in that The training method of the security level prediction model includes: All community environment data and event ratings are used as inputs of a security level prediction model, which takes the security level corresponding to each set of community environment data and event rating predictions as output, the actual security level corresponding to each set of community environment data and event ratings as a prediction target, and minimizing the sum of the first prediction accuracies of all predicted security levels as a training target; the security level prediction model is trained until the sum of the first prediction accuracies reaches convergence, and the training is stopped; the security level prediction model is a decision tree model.

3. The method according to claim 1, characterized in that The activation method of the face card reader includes: The recognition range of the face reader is divided into M block areas of equal area, where N is less than or equal to M, and N≠0, M≠0, and one block area is one unit. When the sensor of the face reader detects that the resident walks to the Mth unit, the face reader is pre-activated. When the sensor of the face reader detects that the resident walks to the Nth unit, the face reader is activated.

4. The method according to claim 1, characterized in that The training method of the matching quantity prediction model includes: All historical entry and exit data and daily behavior characteristics are used as inputs of a matching quantity prediction model, the matching quantity prediction model takes the feature matching quantity and N corresponding to the prediction of each group of historical entry and exit data and daily behavior characteristics as outputs, takes the actual feature matching quantity and N corresponding to each group of historical entry and exit data and daily behavior characteristics as prediction targets, and takes minimizing the sum of the second prediction accuracies of all predicted feature matching quantities and N as a training target; the matching quantity prediction model is trained until the sum of the second prediction accuracies reaches convergence, and the training is stopped; the matching quantity prediction model is a long short-term memory network model.

5. The method according to claim 1, characterized in that The training method of the human body recognition model comprises: All facial feature images and surveillance images are used as inputs of a human body recognition model; the human body recognition model uses the alarm strategy corresponding to the prediction of each group of facial feature images and surveillance images as output, uses the actual alarm strategy corresponding to each group of facial feature images and surveillance images as the prediction target, and uses minimizing the sum of the third prediction accuracies of all predicted alarm strategies as the training target; the human body recognition model is trained until the sum of the third prediction accuracies reaches convergence and the training is stopped; the human body recognition model is a convolutional neural network model.

6. An intelligent community access control and security management system, characterized in that: The system includes regional supervision system and resident security system: The regional regulatory system includes: A first collection module, the first collection module is used to collect community environment data and event ratings; A security rating module, which is used to input community environment data and event ratings into a pre-built security level prediction model and output a security level; Resident security systems include: A second collection module, the second collection module is used to collect historical entry and exit data and daily behavior characteristics of residents; A model training module, wherein the model training module uses historical input and output data and daily behavior characteristics to train a matching quantity prediction model; A third collection module, the third collection module is used to collect the current entry and exit position and the current time point; A non-synaesthesia preparation module, wherein the non-synaesthesia preparation module is used to input the current entry and exit position and the current time point into the matching amount prediction model, and output the feature matching amount and N; An activation recognition module is used to activate the face card reader according to the value N; A matching amount prediction module, the matching amount prediction module is used to adjust the feature matching amount of the face card reader according to the feature matching amount output by the matching amount prediction model; A fourth collection module, the fourth collection module is used to collect facial feature pictures and surveillance pictures in the face card reader; Emergency alarm module, which is used to input facial feature images and surveillance images into a pre-built human recognition model and output an alarm strategy; the alarm strategy includes not outputting an alarm and outputting an alarm. If the alarm strategy is to output an alarm, the alarm is pushed to community security and the access control is automatically unlocked.

Images