Security door alarm method and system based on visual identification control

Through visual identification control and behavioral analysis models, combined with sensor data for multiple verification and abnormal monitoring, the problem of safety hazards of existing security door unlocking methods is solved, and a higher safety and convenient door opening method is achieved.

CN119049199AInactive Publication Date: 2024-11-29NANTONG JINMUDAN DOORS&WINDOWS MFG CO LTD

Patent Information

Application Number
CN202411556886.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-04
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The unlocking methods of existing security doors such as fingerprints and facial recognition pose security risks, which may be copied or forged, and passwords may also be guessed or obtained, resulting in security risks.

Method used

An anti-theft door alarm method based on visual recognition control is adopted. Face data and behavioral characteristics are collected through visual recognition equipment, a behavior analysis model is established, and multiple verification and abnormal monitoring are carried out in combination with sensor data. If the risk assessment score exceeds the safety threshold, the alarm is issued.

Benefits of technology

It improves the safety of security doors, effectively deals with potential safety risks, and provides convenient and fast door opening methods to improve user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119049199A_ABST
    Figure CN119049199A_ABST
Patent Text Reader

Abstract

The invention discloses a security door alarm method and system based on visual recognition control, and relates to the technical field of face recognition, and the method comprises the steps: obtaining a family member face database and a family member database, carrying out the analysis based on the family member database, obtaining a family member behavior feature set, and carrying out the alarm based on the family member behavior feature set. Analyzing passwords and fingerprints in door opening verification to judge whether the door opening verification is abnormal or not, if yes, performing face verification through visual identification equipment, performing family member face database searching and comparison on face data identified outside the security door based on a family member face database, if the comparison fails, marking the security door as an abnormal person, and if not, identifying the security door as the abnormal person. And starting abnormal monitoring, performing risk assessment on the identified personnel according to an abnormal behavior analysis model, determining a risk assessment score, and giving an alarm when the risk assessment score exceeds a safety threshold. The anti-theft door has the advantages that the risk that the password is stolen is reduced, and the safety of the anti-theft door is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of face recognition, and in particular to an anti-theft door alarm method based on visual recognition control. Background Art

[0002] Anti-theft doors usually refer to doors with anti-theft functions installed at the entrance of a building. Their design and structure can effectively block or delay illegal intruders from entering the building, playing a role in protecting property and personnel safety. Such doors are usually made of special locks and materials to increase the difficulty of destruction and intrusion and improve overall security.

[0003] With the development of technology, the current unlocking mode of anti-theft doors has evolved from key opening to fingerprint opening, password opening, and face opening. These new unlocking methods have improved the convenience of life to a certain extent. However, these unlocking methods may also face some security risks. For example, fingerprint and face information may be copied or forged in some cases, and passwords may be guessed or obtained through other means, posing certain security risks. Summary of the invention

[0004] In order to solve the above technical problems, a method and system for anti-theft door alarm based on visual recognition control are provided. This technical solution solves the above technical problems. However, these unlocking methods may also face some security risks. For example, fingerprints and face information may be copied or forged in some cases, and passwords may also be guessed or obtained through other means, which may pose certain security risks.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is: A burglary-proof door alarm method based on visual recognition control, comprising: Collect the facial data of family members through visual recognition equipment and store them as a family member facial database; Collect door opening verification behavior data of family members through anti-theft door verification sensors and pressure sensors, and store them in the family member database; Based on the door opening verification behavior data of each family member in the family member database, a behavior analysis model is established to analyze the behavior characteristics of each family member and obtain a set of family member behavior characteristics; Based on the collection of family member behavior characteristics, analyze whether there is any abnormality in the door opening verification for the family member corresponding to the password and fingerprint who is currently verifying the door. If not, no verification intervention is performed. If yes, facial verification is performed through the visual recognition device. If face verification is performed through visual recognition equipment, the face data recognized outside the anti-theft door is searched and compared with the family member face database based on the family member face database. If the comparison is successful, the intervention ends. If the comparison fails, the person is marked as an abnormal person and abnormal monitoring is started; If abnormal monitoring is performed, the video images in the monitoring area are collected through visual recognition equipment, an abnormal behavior analysis model is established, and risk assessment is performed on the people identified outside the anti-theft door to determine the risk assessment score; Based on the analysis and assessment score, an alarm is issued when the risk assessment score exceeds the safety threshold.

[0006] Preferably, based on the door opening verification behavior data of each family member in the family member database, a behavior analysis model is established to analyze the behavior characteristics of each family member, and obtaining a family member behavior characteristic set specifically includes: Based on the door opening verification habits of family members recorded by the anti-theft door verification sensor and the pressure sensor, the door opening verification habits of family members are bound to the corresponding family members and combined into a family member behavior experience package; Based on the family member behavior experience package, the regression time series features, fingerprint features, password features and fingerprint pressure features corresponding to each family member are marked and constructed as a family member behavior feature set; The behavioral feature set of family members is combined into the original collection of family members, and then divided into a training set and a validation set; Establish behavioral analysis models; Based on the behavior analysis model, the training model takes the behavior characteristics of family members in the training set as input, and the training model takes the information of each family member as output; the training goal is to identify the corresponding family members through the behavior characteristics of family members; the model training ends with minimizing the sum of error functions; Among them, the behavior analysis model is: ; In the formula, To enter the behavioral characteristics of family members Predict family members under the condition Probability , is the output category of the model, For behavioral characteristics, is the natural number base, is the weight vector associated with family members, For family members The associated bias term.

[0007] Preferably, based on the behavior analysis model, the family member corresponding to the password in the door opening verification is obtained, and whether there is an abnormality in the door opening verification is analyzed. If not, no verification intervention is performed. If so, face verification is performed through a visual recognition device, which specifically includes: Based on the behavior analysis model, the fingerprint or password being verified for the door opening password is analyzed to determine whether it is the current family member; If it is the current family member, the behavior analysis model is used to analyze the family member corresponding to the fingerprint or password currently verifying the door password; Based on the corresponding family members, the behavioral characteristics of the family members are searched from the family member behavior experience package to determine whether the behavioral characteristics match. If so, no verification intervention is performed. If not, face verification is performed through visual recognition equipment.

[0008] Preferably, if face verification is performed by a visual recognition device, based on the family member face database, the face data recognized outside the anti-theft door is searched and compared with the family member face database. If the comparison is successful, the intervention is terminated. If the comparison fails, it is marked as an abnormal person. The specific opening of abnormal monitoring includes: Based on visual recognition equipment, facial data of people entering the face verification area is captured; The captured facial data is converted into grayscale images, and each facial feature vector is marked as the facial data to be verified; Based on the facial feature vectors in the face data to be verified and the facial feature vectors of the facial data in the family member face database, the similarity between the face data to be verified and the facial data in the family member face database is calculated through a similarity formula to obtain a facial similarity index; Based on the face similarity index, when the face similarity index is greater than the unlocking threshold, face verification unlocking is performed; when the face similarity index is less than the unlocking threshold, secondary verification is performed; If the face similarity index of the secondary verification is still lower than the unlocking threshold, the person detected by visual recognition will be marked as an abnormal person and abnormal monitoring will be enabled; The similarity formula is: ; Where S is the face similarity index, The face data to be verified Individual face feature vector, The first face data in the family face database Individual face feature vector, is the total number of face feature vectors.

[0009] Preferably, if abnormal monitoring is performed, video images in the area are collected by visual recognition equipment, an abnormal behavior analysis model is established, and risk assessment is performed on the persons identified outside the anti-theft door. The risk assessment score is determined specifically including: Based on visual recognition equipment, the area outside the anti-theft door is divided into several abnormal monitoring areas; Based on the abnormal persons in the area, obtain the video images of abnormal persons, intercept them by frame, mark the key points of each type, determine the coordinate values ​​of the key points, and obtain the human body posture data; Based on human body posture data, the coordinate values ​​of key points in the human body posture data are moved in the visual key frame to determine the motion trajectory; Based on human posture data and motion trajectory, the abnormal behavior analysis model is used to analyze the associated actions between abnormal persons and anti-theft doors. According to the risk score of the associated actions, the risk assessment of the persons identified outside the anti-theft door is performed to determine the risk assessment score; ; In the formula, is the risk score of the associated action, The risk weight of the associated action between the kth abnormal person and the anti-theft door, is the associated actions between k abnormal persons and anti-theft doors.

[0010] Furthermore, a burglar alarm system based on visual recognition control is proposed, which is used to implement the burglar alarm method based on visual recognition control as described above, comprising: A face data collection module is used to collect face data of family members through visual recognition equipment and store it as a family member face database; The family member database module is used to collect the door opening verification behavior data of family members through the anti-theft door verification sensor and the pressure sensor, and store it in the family member database; A feature analysis module, the feature analysis module is electrically connected to the family member database module, and the feature analysis module is used to establish a behavior analysis model based on the door opening verification behavior data of each family member in the family member database, analyze the behavior characteristics of each family member, and obtain a family member behavior characteristic set; The abnormality judgment module is electrically connected to the feature analysis module. The abnormality judgment module is used to analyze whether there is an abnormality in the door opening verification for the family member corresponding to the password and fingerprint in the door opening verification based on the collection of family member behavior features. If not, no verification intervention is performed. If so, face verification is performed through a visual recognition device; The exception handling module is electrically connected to the exception judgment module and the face data acquisition module. The exception handling module is used to search and compare the face data recognized outside the anti-theft door with the family member face database based on the family member face database if the face verification is performed through the visual recognition device. If the comparison is successful, the intervention is terminated. If the comparison fails, the person is marked as an abnormal person and abnormal monitoring is started; The abnormal monitoring module is electrically connected to the abnormal processing module. The abnormal monitoring module is used to collect video images in the monitoring area through visual recognition equipment when performing abnormal monitoring, establish an abnormal behavior analysis model, conduct risk assessment on the personnel identified outside the anti-theft door, and determine the risk assessment score; The alarm module is electrically connected to the abnormality monitoring module and is used to generate an alarm when the risk assessment score exceeds a safety threshold based on the analysis and assessment score.

[0011] Optionally, the feature analysis module includes: An experience acquisition unit, based on the door opening verification habits of family members recorded by the anti-theft door verification sensor and the pressure sensor, binds the door opening verification habits of family members with corresponding family members and combines them into a family member behavior experience package; The feature marking unit marks the regression time series feature, fingerprint feature, password feature and fingerprint pressure feature corresponding to each family member based on the family member behavior experience package, and constructs it into a family member behavior feature set; The collection division unit combines the behavioral feature set of family members into the original collection of family members, and divides it into a training set and a validation set; Model unit, building a behavior analysis model; The training unit is based on the behavior analysis model. The training model takes the behavior characteristics of family members in the training set as input, and the training model takes the information of each family member as output; the training goal is to be able to identify the corresponding family members based on the behavior characteristics of family members; and the model training ends with minimizing the sum of error functions.

[0012] Optionally, the abnormality judgment module includes: The verification analysis unit analyzes the fingerprint or password being verified for the door password based on the behavior analysis model to determine whether it is the current family member; The member verification unit, if it is the current family member, analyzes the family member corresponding to the fingerprint or password currently verifying the door opening password through the behavior analysis model; The feature verification unit searches for the behavioral features of the family members from the family members' behavioral experience packages based on the corresponding family members, and determines whether the behavioral features match. If so, no verification intervention is performed. If not, face verification is performed through a visual recognition device.

[0013] Optionally, the exception handling module includes: A face capture unit, based on a visual recognition device, captures the face data of a person entering the face verification area; The face data processing unit converts the captured face data into grayscale images and marks each face feature vector as face data to be verified; A similarity calculation unit, based on the face feature vector in the face data to be verified and the face feature vector of the face data in the family member face database, calculates the similarity between the face data to be verified and the face data in the family member face database through a similarity formula to obtain a face similarity index; A face verification unit, based on a face similarity index, performs face verification unlocking when the face similarity index is greater than an unlocking threshold, and performs secondary verification when the face similarity index is less than the unlocking threshold; The monitoring activation unit, if the face similarity index of the secondary verification is still lower than the unlocking threshold, will mark the person detected by visual recognition as an abnormal person and activate abnormal monitoring.

[0014] Optionally, the abnormal monitoring module includes: The monitoring area division unit divides the outside of the anti-theft door into several abnormal monitoring areas based on the visual recognition device; The human posture analysis unit obtains video images of abnormal persons based on the abnormal persons in the area, intercepts them frame by frame, marks key points for each type, determines the coordinate values ​​of key points, and obtains human posture data; The motion trajectory analysis unit is based on the human body posture data, and moves the position in the visual key frame according to the coordinate values ​​of the key points in the human body posture data to determine the motion trajectory; The abnormal behavior analysis unit analyzes the associated actions between abnormal persons and anti-theft doors based on human posture data and motion trajectories through an abnormal behavior analysis model. According to the risk scores of the associated actions, a risk assessment is performed on the persons identified outside the anti-theft doors to determine the risk assessment scores.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention proposes a burglar alarm solution based on visual recognition control. The solution establishes a family member database, combines visual recognition equipment and behavior analysis models, realizes multiple verifications and abnormal monitoring, and improves the security of burglar alarm doors. At the same time, through risk assessment and timely alarm mechanism, it effectively responds to potential security risks, and provides a convenient and fast way to open the door, improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1It is a flow chart of a method for alarming anti-theft doors based on visual recognition control; Figure 2 A flow chart of the method for establishing a behavior analysis model; Figure 3 Flowchart of the method for analyzing whether there is an abnormality in door opening verification; Figure 4 A flow chart of the method for searching and comparing the family member face database; Figure 5 A flow chart of the risk assessment method for persons identified outside the anti-theft door; Figure 6 This is a framework diagram of an anti-theft door alarm system based on visual recognition control. DETAILED DESCRIPTION

[0017] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0018] Reference Figure 1 As shown, a method for alarming an anti-theft door based on visual recognition control comprises: Collect the facial data of family members through visual recognition equipment and store them as a family member facial database; Collect door opening verification behavior data of family members through anti-theft door verification sensors and pressure sensors, and store them in the family member database; Based on the door opening verification behavior data of each family member in the family member database, a behavior analysis model is established to analyze the behavior characteristics of each family member and obtain a set of family member behavior characteristics; Based on the collection of family member behavior characteristics, analyze whether there is any abnormality in the door opening verification for the family member corresponding to the password and fingerprint who is currently verifying the door. If not, no verification intervention is performed. If yes, facial verification is performed through the visual recognition device. If face verification is performed through visual recognition equipment, the face data recognized outside the anti-theft door is searched and compared with the family member face database based on the family member face database. If the comparison is successful, the intervention ends. If the comparison fails, the person is marked as an abnormal person and abnormal monitoring is started; If abnormal monitoring is performed, the video images in the monitoring area are collected through visual recognition equipment, an abnormal behavior analysis model is established, and risk assessment is performed on the people identified outside the anti-theft door to determine the risk assessment score; Based on the analysis and assessment score, an alarm is issued when the risk assessment score exceeds the safety threshold.

[0019] This solution establishes a family member face database by entering the face data of family members, and generates a behavior database and trains the analysis model by real-time collection and pre-entry of behavioral characteristics. After the system recognizes a face outside the anti-theft door, it will compare it with the family member face database. If the match is successful, the monitoring will be lifted; otherwise, it will be marked as a person. If monitoring is activated, the system will use the behavioral analysis model to conduct a risk assessment on the person, determine their risk assessment score, and trigger an alarm when the score exceeds the safety threshold.

[0020] Reference Figure 2 As shown, based on the door opening verification behavior data of each family member in the family member database, a behavior analysis model is established to analyze the behavior characteristics of each family member, and the family member behavior characteristic set is obtained, which specifically includes: Based on the door opening verification habits of family members recorded by the anti-theft door verification sensor and the pressure sensor, the door opening verification habits of family members are bound to the corresponding family members and combined into a family member behavior experience package; Based on the family member behavior experience package, the regression time series features, fingerprint features, password features and fingerprint pressure features corresponding to each family member are marked and constructed as a family member behavior feature set; The behavioral feature set of family members is combined into the original collection of family members, and then divided into a training set and a validation set; Establish behavioral analysis models; Based on the behavior analysis model, the training model takes the behavior characteristics of family members in the training set as input, and the training model takes the information of each family member as output; the training goal is to identify the corresponding family members through the behavior characteristics of family members; the model training ends with minimizing the sum of error functions; Among them, the behavior analysis model is: ; In the formula, To enter the behavioral characteristics of family members Predict family members under the condition Probability , is the output category of the model, For behavioral characteristics, is the natural number base, is the weight vector associated with family members, For family members The associated bias term.

[0021] This solution establishes a behavior analysis model by analyzing and utilizing the behavior habits of family members when opening the door based on the records of anti-theft door verification sensors and pressure sensors, and conducts learning and analysis on historical behavior habits, so that the behavior analysis model can identify the identities of family members and avoid the loss of verification passwords and economic losses.

[0022] It can be understood that the family member behavior experience package is recorded by visual recognition equipment according to the historical door opening habits of family members, and packaged into the door opening habit characteristics corresponding to each family member; for example: when a family member uses the correct fingerprint or password to verify the door opening during non-return time, in order to avoid password theft, the visual recognition device can be mobilized to perform a secondary facial confirmation on the person who is verifying the door opening password to see if he is a family member. If the facial recognition fails, the fingerprint or password door opening verification will be rejected.

[0023] Reference Figure 3 As shown, based on the behavior analysis model, the family member corresponding to the password in the door opening verification is obtained, and the door opening verification is analyzed to see if there is any abnormality. If not, no verification intervention is performed. If so, face verification is performed through the visual recognition device, which specifically includes: Based on the behavior analysis model, the fingerprint or password being verified for the door opening password is analyzed to determine whether it is the current family member; If it is the current family member, the behavior analysis model is used to analyze the family member corresponding to the fingerprint or password currently verifying the door password; Based on the corresponding family members, the behavioral characteristics of the family members are searched from the family member behavior experience package to determine whether the behavioral characteristics match. If so, no verification intervention is performed. If not, face verification is performed through visual recognition equipment.

[0024] It is understandable that the behavioral characteristics of a family member are usually highly representative. The fingerprint or password currently being verified for the door opening password can be analyzed to determine which family member it is, and whether there is any abnormality in the door opening password verification based on the behavioral characteristics of the family member. For example: a family member's historical door opening habit is to use fingerprint verification to open the door, but it is detected that the door opening password verification this time is the door opening password of the family member, then it can be determined that this door opening verification is in an abnormal state. If it is a fingerprint verification, then the pressure value of the historical family member's fingerprint verification can be used to verify whether there is any abnormality.

[0025] Reference Figure 4 If face verification is performed through visual recognition equipment, the face data recognized outside the anti-theft door is searched and compared with the family member face database based on the family member face database. If the comparison is successful, the intervention ends. If the comparison fails, it is marked as an abnormal person. The specific opening of abnormal monitoring includes: Based on visual recognition equipment, facial data of people entering the face verification area is captured; The captured facial data is converted into grayscale images, and each facial feature vector is marked as the facial data to be verified; Based on the facial feature vectors in the face data to be verified and the facial feature vectors of the facial data in the family member face database, the similarity between the face data to be verified and the facial data in the family member face database is calculated through a similarity formula to obtain a facial similarity index; Based on the face similarity index, when the face similarity index is greater than the unlocking threshold, face verification unlocking is performed; when the face similarity index is less than the unlocking threshold, secondary verification is performed; If the face similarity index of the secondary verification is still lower than the unlocking threshold, the person detected by visual recognition will be marked as an abnormal person and abnormal monitoring will be enabled; The similarity formula is: ; Where S is the face similarity index, The face data to be verified Individual face feature vector, The first face data in the family face database Individual face feature vector, is the total number of face feature vectors.

[0026] Reference Figure 5 As shown, if abnormal monitoring is performed, the video images in the area are collected through visual recognition equipment, an abnormal behavior analysis model is established, and risk assessment is performed on the people identified outside the anti-theft door. The risk assessment score is determined specifically including: Based on visual recognition equipment, the area outside the anti-theft door is divided into several abnormal monitoring areas; Based on the abnormal persons in the area, obtain the video images of abnormal persons, intercept them by frame, mark the key points of each type, determine the coordinate values ​​of the key points, and obtain the human body posture data; Based on human body posture data, the coordinate values ​​of key points in the human body posture data are moved in the visual key frame to determine the motion trajectory; Based on human posture data and motion trajectory, the abnormal behavior analysis model is used to analyze the associated actions between abnormal persons and anti-theft doors. According to the risk score of the associated actions, the risk assessment of the persons identified outside the anti-theft door is performed to determine the risk assessment score; ; In the formula, is the risk score of the associated action, The risk weight of the associated action between the kth abnormal person and the anti-theft door, is the associated actions between k abnormal persons and anti-theft doors.

[0027] This solution uses visual recognition equipment to monitor the area outside the security door. When an abnormal person is detected, the video image is obtained and key point markers are extracted to obtain human posture data. The movement trajectory is determined by the tracking algorithm, and the abnormal behavior is identified by the abnormal behavior analysis model. Finally, the risk of the person is assessed based on the risk score of the associated action, realizing real-time monitoring of the person outside the security door and improving the security of the security door.

[0028] Further, see Figure 6 As shown, based on the same inventive concept of an anti-theft door alarm method based on visual recognition control, a anti-theft door alarm system based on visual recognition control is proposed, comprising: A face data collection module is used to collect face data of family members through visual recognition equipment and store it as a family member face database; The family member database module is used to collect the door opening verification behavior data of family members through the anti-theft door verification sensor and the pressure sensor, and store it in the family member database; A feature analysis module, the feature analysis module is electrically connected to the family member database module, and the feature analysis module is used to establish a behavior analysis model based on the door opening verification behavior data of each family member in the family member database, analyze the behavior characteristics of each family member, and obtain a family member behavior characteristic set; The abnormality judgment module is electrically connected to the feature analysis module. The abnormality judgment module is used to analyze whether there is an abnormality in the door opening verification for the family member corresponding to the password and fingerprint in the door opening verification based on the collection of family member behavior features. If not, no verification intervention is performed. If so, face verification is performed through a visual recognition device; The exception handling module is electrically connected to the exception judgment module and the face data acquisition module. The exception handling module is used to search and compare the face data recognized outside the anti-theft door with the family member face database based on the family member face database if the face verification is performed through the visual recognition device. If the comparison is successful, the intervention is terminated. If the comparison fails, the person is marked as an abnormal person and abnormal monitoring is started; The abnormal monitoring module is electrically connected to the abnormal processing module. The abnormal monitoring module is used to collect video images in the monitoring area through visual recognition equipment when performing abnormal monitoring, establish an abnormal behavior analysis model, conduct risk assessment on the personnel identified outside the anti-theft door, and determine the risk assessment score; The alarm module is electrically connected to the abnormality monitoring module and is used to generate an alarm when the risk assessment score exceeds a safety threshold based on the analysis and assessment score.

[0029] The feature analysis module includes: An experience acquisition unit, based on the door opening verification habits of family members recorded by the anti-theft door verification sensor and the pressure sensor, binds the door opening verification habits of family members with corresponding family members and combines them into a family member behavior experience package; The feature marking unit marks the regression time series feature, fingerprint feature, password feature and fingerprint pressure feature corresponding to each family member based on the family member behavior experience package, and constructs it into a family member behavior feature set; The collection division unit combines the behavioral feature set of family members into the original collection of family members, and divides it into a training set and a validation set; Model unit, building a behavior analysis model; The training unit is based on the behavior analysis model. The training model takes the behavior characteristics of family members in the training set as input, and the training model takes the information of each family member as output; the training goal is to be able to identify the corresponding family members based on the behavior characteristics of family members; and the model training ends with minimizing the sum of error functions.

[0030] The abnormal judgment module includes: The verification analysis unit analyzes the fingerprint or password being verified for the door password based on the behavior analysis model to determine whether it is the current family member; The member verification unit, if it is the current family member, analyzes the family member corresponding to the fingerprint or password currently verifying the door opening password through the behavior analysis model; The feature verification unit searches for the behavioral features of the family members from the family members' behavioral experience packages based on the corresponding family members, and determines whether the behavioral features match. If so, no verification intervention is performed. If not, face verification is performed through a visual recognition device.

[0031] The exception handling module includes: A face capture unit, based on a visual recognition device, captures the face data of a person entering the face verification area; The face data processing unit converts the captured face data into grayscale images and marks each face feature vector as face data to be verified; A similarity calculation unit, based on the face feature vector in the face data to be verified and the face feature vector of the face data in the family member face database, calculates the similarity between the face data to be verified and the face data in the family member face database through a similarity formula to obtain a face similarity index; A face verification unit, based on a face similarity index, performs face verification unlocking when the face similarity index is greater than an unlocking threshold, and performs secondary verification when the face similarity index is less than the unlocking threshold; The monitoring activation unit, if the face similarity index of the secondary verification is still lower than the unlocking threshold, will mark the person detected by visual recognition as an abnormal person and activate abnormal monitoring.

[0032] The abnormal monitoring module includes: The monitoring area division unit divides the outside of the anti-theft door into several abnormal monitoring areas based on the visual recognition device; The human posture analysis unit obtains video images of abnormal persons based on the abnormal persons in the area, intercepts them frame by frame, marks key points for each type, determines the coordinate values ​​of key points, and obtains human posture data; The motion trajectory analysis unit is based on the human body posture data, and moves the position in the visual key frame according to the coordinate values ​​of the key points in the human body posture data to determine the motion trajectory; The abnormal behavior analysis unit analyzes the associated actions between abnormal persons and anti-theft doors based on human posture data and motion trajectories through an abnormal behavior analysis model. According to the risk scores of the associated actions, a risk assessment is performed on the persons identified outside the anti-theft doors to determine the risk assessment scores.

[0033] The use process of a burglar alarm system based on visual recognition control is as follows: Step 1: Collect the facial data of family members through visual recognition equipment and store it as a family member facial database; Step 2: Collect the door opening verification behavior data of family members through the anti-theft door verification sensor and pressure sensor, and store it in the family member database; Step 3: Based on the door opening verification habits of family members recorded by the anti-theft door verification sensor and the pressure sensor, the door opening verification habits of family members are bound to the corresponding family members and combined into a family member behavior experience package; Step 4: Based on the family member behavior experience package, mark the regression time series features, fingerprint features, password features and fingerprint pressure features corresponding to each family member, and construct them into a family member behavior feature set; Step 5: Combine the behavioral feature sets of family members into the original collection of family members, and divide it into a training set and a validation set; Step 6: Establish a behavior analysis model; Step 7: Based on the behavior analysis model, the training model takes the behavior characteristics of family members in the training set as input, and the training model takes the information of each family member as output; the training goal is to identify the corresponding family members based on the behavior characteristics of family members; the model training ends with minimizing the sum of the error function; Step 8: Based on the behavior analysis model, analyze the fingerprint or password being verified for the door opening password to determine whether it is the current family member; Step 9: If it is the current family member, analyze the family member corresponding to the fingerprint or password currently verifying the door password through the behavior analysis model; Step 10: Based on the corresponding family members, the behavior characteristics of the family members are searched from the family member behavior experience package to determine whether the behavior characteristics match. If so, no verification intervention is performed. If not, face verification is performed through a visual recognition device; Step 11: Based on the visual recognition device, the facial data of the person entering the face verification area is captured; Step 12: Convert the captured facial data into grayscale images and mark each facial feature vector as the facial data to be verified; Step 13: Based on the facial feature vector in the face data to be verified and the facial feature vector of the facial data in the family member face database, the similarity between the face data to be verified and the facial data in the family member face database is calculated by a similarity formula to obtain a facial similarity index; Step 14: Based on the face similarity index, when the face similarity index is greater than the unlocking threshold, face verification unlocking is performed; when the face similarity index is less than the unlocking threshold, secondary verification is performed; Step 15: If the face similarity index of the secondary verification is still lower than the unlocking threshold, the person detected by visual recognition is marked as an abnormal person and abnormal monitoring is turned on; Step 16: Based on the visual recognition device, the area outside the anti-theft door is divided into several abnormal monitoring areas; Step 17: Based on the abnormal persons in the area, obtain the video images of the abnormal persons, intercept them by frame, mark the key points of each type, determine the coordinate values ​​of the key points, and obtain the human body posture data; Step 18: Based on the human body posture data, the coordinate values ​​of the key points in the human body posture data are moved in the visual key frame to determine the motion trajectory; Step 19: Based on the human posture data and motion trajectory, the associated actions between the abnormal person and the anti-theft door are analyzed through the abnormal behavior analysis model. According to the risk score of the associated action, the person identified outside the anti-theft door is risk assessed to determine the risk assessment score.

[0034] In summary, the advantages of the present invention are: by establishing a family member database, combining visual recognition equipment and behavior analysis models, multiple verifications and abnormal monitoring are achieved to improve the security of anti-theft doors. At the same time, through risk assessment and timely alarm mechanism, potential security risks are effectively dealt with, and a convenient and fast door opening method is provided to improve user experience.

[0035] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for alarming anti-theft doors based on visual recognition control, characterized in that: include: Collect family members' facial data through visual recognition equipment and store them as a family members' facial database; Collect door opening verification behavior data of family members through anti-theft door verification sensors and pressure sensors, and store them in the family member database; Based on the door opening verification behavior data of each family member in the family member database, a behavior analysis model is established to analyze the behavior characteristics of each family member and obtain a set of family member behavior characteristics; Based on the collection of family member behavior characteristics, analyze whether there is any abnormality in the door opening verification for the family member corresponding to the password and fingerprint who is currently verifying the door. If not, no verification intervention is performed. If yes, facial verification is performed through the visual recognition device. If face verification is performed through visual recognition equipment, the face data recognized outside the anti-theft door is searched and compared with the family member face database based on the family member face database. If the comparison is successful, the intervention ends. If the comparison fails, the person is marked as an abnormal person and abnormal monitoring is started; If abnormal monitoring is performed, the video images in the monitoring area are collected through visual recognition equipment, an abnormal behavior analysis model is established, and risk assessment is performed on the people identified outside the anti-theft door to determine the risk assessment score; Based on the analysis and assessment score, an alarm is issued when the risk assessment score exceeds the safety threshold.

2. The anti-theft door alarm method based on visual recognition control according to claim 1 is characterized in that: Based on the door opening verification behavior data of each family member in the family member database, a behavior analysis model is established to analyze the behavior characteristics of each family member, and the family member behavior characteristic set is obtained, which specifically includes: Based on the door opening verification habits of family members recorded by the anti-theft door verification sensor and the pressure sensor, the door opening verification habits of family members are bound to the corresponding family members and combined into a family member behavior experience package; Based on the family member behavior experience package, the regression time series features, fingerprint features, password features and fingerprint pressure features corresponding to each family member are marked and constructed as a family member behavior feature set; The behavioral feature set of family members is combined into the original collection of family members, and then divided into a training set and a validation set; Establish behavioral analysis models; Based on the behavior analysis model, the training model takes the behavior characteristics of family members in the training set as input, and the training model takes the information of each family member as output; the training goal is to identify the corresponding family members through the behavior characteristics of family members; the model training ends with minimizing the sum of error functions; Among them, the behavior analysis model is: ; In the formula, To enter the behavioral characteristics of family members Predict family members under the condition Probability , is the output category of the model, For behavioral characteristics, is the natural number base, is the weight vector associated with family members, For family members The associated bias term.

3. The anti-theft door alarm method based on visual recognition control according to claim 2 is characterized in that: Based on the behavior analysis model, obtain the family member corresponding to the password in the door opening verification, and analyze whether there is any abnormality in the door opening verification. If not, no verification intervention is performed. If so, face verification is performed through the visual recognition device. Specifically, it includes: Based on the behavior analysis model, the fingerprint or password being verified for the door password is analyzed to determine whether it is the current family member; If it is the current family member, the behavior analysis model is used to analyze the family member corresponding to the fingerprint or password currently verifying the door opening password; Based on the corresponding family members, the behavioral characteristics of the family members are searched from the family member behavior experience package to determine whether the behavioral characteristics match. If so, no verification intervention is performed. If not, face verification is performed through visual recognition equipment.

4. The anti-theft door alarm method based on visual recognition control according to claim 3 is characterized in that: If face verification is performed through visual recognition equipment, the face data recognized outside the anti-theft door is searched and compared with the family member face database based on the family member face database. If the comparison is successful, the intervention ends. If the comparison fails, the person is marked as an abnormal person. The specific abnormal monitoring includes: Based on visual recognition equipment, facial data of people entering the face verification area is captured; The captured facial data is converted into grayscale images, and each facial feature vector is marked as the facial data to be verified; Based on the facial feature vectors in the face data to be verified and the facial feature vectors of the facial data in the family member face database, the similarity between the face data to be verified and the facial data in the family member face database is calculated through a similarity formula to obtain a facial similarity index; Based on the face similarity index, when the face similarity index is greater than the unlocking threshold, face verification unlocking is performed; when the face similarity index is less than the unlocking threshold, secondary verification is performed; If the face similarity index of the secondary verification is still lower than the unlocking threshold, the person detected by visual recognition will be marked as an abnormal person and abnormal monitoring will be turned on; The similarity formula is: ; Where S is the face similarity index, The face data to be verified Individual face feature vector, The first face data in the family face database Individual face feature vector, is the total number of face feature vectors.

5. The anti-theft door alarm method based on visual recognition control according to claim 4 is characterized in that: If abnormal monitoring is performed, the video images in the area are collected through visual recognition equipment, an abnormal behavior analysis model is established, and risk assessment is performed on the people identified outside the anti-theft door. The risk assessment score is determined to include: Based on visual recognition equipment, the area outside the anti-theft door is divided into several abnormal monitoring areas; Based on the abnormal persons in the area, obtain the video images of abnormal persons, intercept them by frame, mark the key points of each type, determine the coordinate values ​​of the key points, and obtain the human body posture data; Based on human body posture data, the coordinate values ​​of key points in the human body posture data are moved in the visual key frame to determine the motion trajectory; Based on human posture data and motion trajectory, the abnormal behavior analysis model is used to analyze the associated actions between abnormal persons and anti-theft doors. According to the risk score of the associated actions, the risk assessment of the persons identified outside the anti-theft door is performed to determine the risk assessment score; ; In the formula, is the risk score of the associated action, The risk weight of the associated action between the kth abnormal person and the anti-theft door, is the associated actions between k abnormal persons and anti-theft doors.

6. A burglar alarm system based on visual recognition control, characterized in that: A method for realizing an anti-theft door alarm based on visual recognition control as claimed in any one of claims 1 to 5, comprising: A face data collection module is used to collect face data of family members through visual recognition equipment and store it as a family member face database; The family member database module is used to collect the door opening verification behavior data of family members through the anti-theft door verification sensor and the pressure sensor, and store it in the family member database; A feature analysis module, the feature analysis module is electrically connected to the family member database module, and the feature analysis module is used to establish a behavior analysis model based on the door opening verification behavior data of each family member in the family member database, analyze the behavior characteristics of each family member, and obtain a family member behavior characteristic set; The abnormality judgment module is electrically connected to the feature analysis module. The abnormality judgment module is used to analyze whether there is an abnormality in the door opening verification for the family member corresponding to the password and fingerprint in the door opening verification based on the collection of family member behavior features. If not, no verification intervention is performed. If so, face verification is performed through a visual recognition device; The exception handling module is electrically connected to the exception judgment module and the face data acquisition module. The exception handling module is used to search and compare the face data recognized outside the anti-theft door with the family member face database based on the family member face database if the face verification is performed through the visual recognition device. If the comparison is successful, the intervention is terminated. If the comparison fails, the person is marked as an abnormal person and abnormal monitoring is started; The abnormal monitoring module is electrically connected to the abnormal processing module. The abnormal monitoring module is used to collect video images in the monitoring area through visual recognition equipment when performing abnormal monitoring, establish an abnormal behavior analysis model, conduct risk assessment on the personnel identified outside the anti-theft door, and determine the risk assessment score; The alarm module is electrically connected to the abnormality monitoring module and is used to generate an alarm when the risk assessment score exceeds a safety threshold based on the analysis and assessment score.

7. The anti-theft door alarm system based on visual recognition control according to claim 6 is characterized in that: The feature analysis module includes: An experience acquisition unit, based on the door opening verification habits of family members recorded by the anti-theft door verification sensor and the pressure sensor, binds the door opening verification habits of family members with corresponding family members and combines them into a family member behavior experience package; The feature marking unit marks the regression time series feature, fingerprint feature, password feature and fingerprint pressure feature corresponding to each family member based on the family member behavior experience package, and constructs it into a family member behavior feature set; The collection division unit combines the behavioral feature set of family members into the original collection of family members, and divides it into a training set and a validation set; Model unit, building a behavior analysis model; The training unit is based on the behavior analysis model. The training model takes the behavior characteristics of family members in the training set as input, and the training model takes the information of each family member as output; the training goal is to be able to identify the corresponding family members based on the behavior characteristics of family members; and the model training ends with minimizing the sum of error functions.

8. The anti-theft door alarm system based on visual recognition control according to claim 6 is characterized in that: The abnormal judgment module includes: The verification analysis unit analyzes the fingerprint or password being verified for the door password based on the behavior analysis model to determine whether it is the current family member; The member verification unit, if it is the current family member, analyzes the family member corresponding to the fingerprint or password currently verifying the door opening password through the behavior analysis model; The feature verification unit searches for the behavioral features of the family members from the family members' behavioral experience packages based on the corresponding family members, and determines whether the behavioral features match. If so, no verification intervention is performed. If not, face verification is performed through a visual recognition device.

9. The anti-theft door alarm system based on visual recognition control according to claim 6 is characterized in that: The exception handling module includes: A face capture unit, based on a visual recognition device, captures the face data of a person entering the face verification area; The face data processing unit converts the captured face data into grayscale images and marks each face feature vector as face data to be verified; A similarity calculation unit, based on the face feature vector in the face data to be verified and the face feature vector of the face data in the family member face database, calculates the similarity between the face data to be verified and the face data in the family member face database through a similarity formula to obtain a face similarity index; A face verification unit, based on a face similarity index, performs face verification unlocking when the face similarity index is greater than an unlocking threshold, and performs secondary verification when the face similarity index is less than the unlocking threshold; The monitoring activation unit, if the face similarity index of the secondary verification is still lower than the unlocking threshold, will mark the person detected by visual recognition as an abnormal person and activate abnormal monitoring.

10. The anti-theft door alarm system based on visual recognition control according to claim 6, characterized in that: The abnormal monitoring module includes: The monitoring area division unit divides the outside of the anti-theft door into several abnormal monitoring areas based on the visual recognition device; The human posture analysis unit obtains video images of abnormal persons based on the abnormal persons in the area, intercepts them frame by frame, marks key points for each type, determines the coordinate values ​​of key points, and obtains human posture data; The motion trajectory analysis unit is based on the human body posture data, and moves the position in the visual key frame according to the coordinate values ​​of the key points in the human body posture data to determine the motion trajectory; The abnormal behavior analysis unit analyzes the associated actions between abnormal persons and anti-theft doors based on human posture data and motion trajectories through an abnormal behavior analysis model. According to the risk scores of the associated actions, a risk assessment is performed on the persons identified outside the anti-theft doors to determine the risk assessment scores.

Citation Information

Patent Citations

  • Smart door lock unlocking method and smart door lock

    CN107633573A

  • Monitoring system and method for judging video behavior based on deep learning

    CN108600701A

  • Method for identifying family members by intelligent door lock

    CN110060389A

  • Hierarchical authority unlocking method and device and storage medium

    CN111553192A

  • Human body abnormal behavior recognition alarm system and method under panoramic monitoring based on posture estimation

    CN112991656A

Cited By

  • Operation control method for anti-theft safety door

    CN120014746A

  • Operation control method for anti-theft safety door

    CN120014746B