Face recognition unlocking method and system based on intelligent door lock
By constructing and analyzing historical and representative activation combinations, and combining this with infrared sensor verification, the security risks of smart locks when users are coerced were resolved, and security protection was achieved in extreme situations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-14
- Publication Date
- 2026-04-03
AI Technical Summary
Existing smart locks can still open the door normally when the user is coerced, posing a serious security risk and failing to effectively identify whether the user is being coerced.
By constructing historical activation combinations and representative activation combinations, we can analyze the similarity and similarity ratio of the combinations to determine whether the user's expression is natural. If it is not natural, we can output an abnormal recognition signal to prevent the door lock from being opened. If necessary, we can further verify the information using infrared sensors and error checking features.
This improves the security of smart door locks in extreme security events, effectively identifies whether a user is being coerced, prevents unauthorized opening, and enhances the safety of users and people inside the house.
Smart Images
Figure CN121789326A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of smart door lock technology, and in particular to a face recognition unlocking method and system based on smart door locks. Background Technology
[0002] With the rapid development of smart security technology, smart door locks based on facial recognition have gradually become important access control devices in modern homes, offices, and other scenarios. Traditional facial recognition smart door locks typically use cameras to capture user facial images, and after steps such as liveness detection, feature extraction, and comparison, automatically control the lock to open after successful verification, achieving "seamless access" and improving the convenience and experience of entering and exiting.
[0003] However, existing technology has a significant security vulnerability: these facial recognition door locks only focus on whether the user is a "legitimate user" and whether they are a "real, living person" during the identification process. Once both verifications are successful, the door is opened. This mechanism is applicable in normal scenarios, but it can become a serious security flaw when the user is subjected to extreme security events such as coercion, tailgating, or hostage situations. Specifically, if a user is forced to face the door lock camera under duress, the system will still open the door normally because they are a legitimate user and a living person. This not only puts the user in a more dangerous situation but may also threaten the personal safety and property of other people inside the house. Therefore, current smart door locks still have certain security vulnerabilities and room for improvement. Summary of the Invention
[0004] To improve the security of smart door locks, this application provides a face recognition unlocking method and system based on smart door locks.
[0005] Firstly, this application provides a face recognition unlocking method based on a smart door lock, employing the following technical solution: A facial recognition unlocking method based on a smart door lock includes: Obtain facial recognition features; The user's recognition status is determined based on facial recognition features, and when the user's recognition status matches the preset successful recognition status, a representative feature is determined from the facial recognition features. On a preset timeline, a historical interval with a preset historical duration is constructed with the current time point as the endpoint, and the same activation time period is determined within the historical interval based on the current facial recognition features; Under the same activation period, each identification representative feature is combined to construct a historical activation combination, and the current identification representative features are combined to construct a representative activation combination; The similarity of the combination is determined by comparing and analyzing the representative activation combination and the historical activation combination. The historical activation combination with a similarity greater than the preset requirement similarity is defined as the historical similar combination. The historical similarity ratio is determined by calculating based on the historical similar combination and the historical activation combination. Determine whether the proportion of historical similarities is greater than the preset baseline demand proportion; If the proportion of historically similar demand is greater than the baseline demand proportion, then the door lock will be controlled to open. If the proportion of historical similarities is not greater than the baseline demand proportion, then an anomaly signal will be output.
[0006] Optionally, the step of determining combinatorial similarity by performing comparative analysis based on representative activation combinations and historical activation combinations includes: In the representative activation combination, a fixed number of representative features are randomly selected to construct a unit activation combination, and the comparison activation combination is determined based on the unit activation combination in the historical activation combination. The effective weight of each unit is determined based on a preset weight matching relationship. The unit alignment score is determined by comparing the unit activation combination and the alignment activation combination, and the combination similarity is determined by calculating all the unit alignment scores and the corresponding effective unit weights.
[0007] Optionally, it also includes a step for constructing weighted matching relationships, which includes: Two representative features are randomly selected from the unit activation combination to determine the relative correlation. Identifying representative features whose relative correlation is greater than the preset requirement correlation are categorized into a preset, initially empty feature categorization set; The number of individual features is determined by counting based on the identified representative features, and the number of inductive features is determined by counting based on the feature inductive set. The number of expressions is determined by calculating the number of monomeric features and the number of inductive features; The effective weight of a unit is determined based on the preset expression matching relationship, and a weight matching relationship is constructed based on the effective weight of the unit and the unit activation combination.
[0008] Optionally, the step of randomly selecting two representative features from the unit activation combination to determine the relative correlation includes: In the unit activation combination, a representative feature is randomly selected and defined as a fixed representative feature, and the remaining representative features are defined as variable representative features. A relative feature combination is constructed based on the fixed representative feature and the variable representative feature. Under the historical activation combination, the equivalent point is determined based on the fixed representative characteristics, and the equivalent main characteristics are determined based on the changing representative characteristics under the equivalent point; The relative degree of a single entity is determined by analyzing the changing representative features and the equivalent principal features, and the degree of correlation of the combination of relative features with the fixed representative features is determined by analyzing and calculating all the relative degrees of a single entity. The relative correlation is determined by calculating the correlation between the same relative feature and different fixed representative features.
[0009] Optionally, after identifying the abnormal signal output, the facial recognition unlocking method based on the smart door lock also includes: The incorrect verification features corresponding to the current face recognition features are determined based on the preset verification and matching relationship; On the timeline, a verification interval with a preset verification duration is constructed with the current time point as the leading point, and user input features are obtained in real time based on the error verification features within the verification interval; Determine if a user input feature matches an error verification feature; If a user input feature matches the error verification feature, then the door lock will be opened. If no user input feature matches the error check feature, an identification error signal is output.
[0010] Optionally, after an error signal is output, the facial recognition unlocking method based on the smart door lock further includes: Acquire external infrared features; Determine whether the external infrared features only contain users corresponding to facial recognition features; If the external infrared features only include the user corresponding to the facial recognition features, then an error recognition alert signal will be output. If the external infrared features include not only the user corresponding to the facial recognition features, the door lock will be controlled to "virtually" open, and an error recognition reminder signal will be output.
[0011] Secondly, this application provides a face recognition unlocking system based on a smart door lock, employing the following technical solution: A facial recognition unlocking system based on a smart door lock includes: The acquisition module is used to acquire facial recognition features; The processing module, connected to the acquisition and judgment modules, is used for information storage and processing; The judgment module, connected to the acquisition and processing modules, is used for judging information. The processing module determines the user's recognition status based on facial recognition features, and when the judgment module determines that the user's recognition status is consistent with the preset successful recognition status, it determines the recognition representative features from the facial recognition features. The processing module constructs a historical interval with a preset historical duration on a preset time axis, with the current time point as the endpoint, and determines the same activation time period in the historical interval based on the current face recognition features; The processing module combines each representative feature of identification to construct a historical activation combination under the same activation period, and combines the current representative features of identification to construct a representative activation combination; The processing module compares and analyzes representative activation combinations and historical activation combinations to determine the combination similarity. Historical activation combinations with a combination similarity greater than the preset required similarity are defined as historical similar combinations. The module also calculates the historical similarity ratio based on historical similar combinations and historical activation combinations. The judgment module determines whether the proportion of historical similarities is greater than the preset baseline requirement proportion; If the judgment module determines that the historical similarity ratio is greater than the baseline demand ratio, the processing module will control the door lock to open. If the judgment module determines that the proportion of historical similarities is not greater than the baseline requirement proportion, the processing module outputs an identification anomaly signal.
[0012] In summary, this application includes at least one of the following beneficial technical effects: During the facial recognition process, the user's current facial features can be compared with those of normal facial features in historical situations to determine if there are any unnatural situations. This can effectively identify situations of coercion and prevent the user from opening the door, thereby improving overall security. By analyzing the different facial expressions of different users, we can better understand and analyze various facial features, thereby improving the accuracy of data analysis. Attached Figure Description
[0013] Figure 1 This is a flowchart of a facial recognition unlocking method based on smart door locks.
[0014] Figure 2 This is a flowchart of a module based on a facial recognition unlocking method for smart door locks. Detailed Implementation
[0015] To make the purpose, technical solution, and advantages of this application clearer, the following is combined with Figures 1-2 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.
[0016] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.
[0017] This application discloses a facial recognition unlocking method based on a smart door lock, referring to... Figure 1 The method flow for unlocking smart door locks using facial recognition includes the following steps: Step S100: Obtain facial recognition features.
[0018] The facial recognition feature is the facial feature obtained by analyzing the image containing the user's face captured by the image acquisition device on the smart door lock. Specifically, the facial feature recognition can be trained by deep learning, so as to effectively extract features from the image. This method is a conventional technique for those skilled in the art and will not be elaborated here.
[0019] Step S101: Determine the user's recognition status based on the facial recognition features, and determine the representative features in the facial recognition features when the user's recognition status matches the preset successful recognition status.
[0020] User identification status refers to the state of the current user, determined by analyzing facial recognition features, to determine if the user is a registered user eligible for unlocking. This can be achieved by converting the facial recognition features into high-dimensional vector feature codes, then using the current feature codes and locally stored feature codes to calculate the cosine distance to determine the relative relationship. The calculated correlation is then compared with a pre-set threshold to confirm the identification status. Successful identification status refers to the user identification status when the current user is a registered user eligible for unlocking. When the user identification status matches the successful identification status, it indicates that the current user's identity is correct, and further analysis is needed to determine if the user has been coerced. Representative features are facial features that best reflect unnatural behavior in the user, such as the corner of the mouth. The specific locations are determined in advance by staff. In this case, it is only necessary to mark the features at the corresponding locations in the facial recognition features as representative features to facilitate the analysis of whether the user is behaving unnaturally.
[0021] Step S102: Construct a historical interval with a preset historical duration on the preset timeline, with the current time point as the endpoint, and determine the same activation time period in the historical interval based on the current face recognition features.
[0022] The timeline is a coordinate axis formed by combining various time points. This timeline points from the time points that have already passed to the time points that have not yet been reached, with the time points that have already passed being on the left, i.e., the beginning. The historical duration is the total duration since the smart door lock was put into use, as set by the staff. By constructing historical intervals, it is possible to acquire and analyze the data within the historical duration. The same activation period is the time period within the historical interval when the user corresponding to the current face uses the smart door lock to unlock and complete the opening operation.
[0023] Step S103: Under the same activation period, combine each identification representative feature to construct a historical activation combination, and combine the current identification representative features to construct a representative activation combination.
[0024] Historical activation combinations are combinations of representative features corresponding to a single activation period. By constructing representative activation combinations, it is possible to combine and identify the current representative features, thereby facilitating subsequent analysis.
[0025] Step S104: Compare and analyze the representative activation combination and the historical activation combination to determine the combination similarity, and define the historical activation combination with a combination similarity greater than the preset requirement similarity as the historical similar combination, and calculate the historical similarity ratio based on the historical similar combination and the historical activation combination.
[0026] Combination similarity is a parameter reflecting the degree of similarity between the overall performance of the active combination and the historical active combination. It can be determined by comparing the corresponding features one by one to determine the similarity of individual components, and then combining the individual component similarities to determine the combination similarity. Alternatively, the combination similarity can be determined using the method in steps S200-S202. Demand similarity is the minimum combination similarity set by the staff for determining that the performance of users under two combinations is relatively similar. Historical similar combinations are defined to identify and distinguish different historical active combinations, which facilitates subsequent analysis. Historical similarity ratio is the proportion of historical similar combinations in historical active combinations, which is determined by dividing the number of historical similar combinations by the number of historical active combinations.
[0027] Step S105: Determine whether the historical similarity ratio is greater than the preset baseline demand ratio.
[0028] The baseline requirement ratio is the minimum historical similarity ratio set by staff to determine that the current user's facial expressions and behavior are similar to those when the user is normally opening the door, i.e., the user does not have unnatural facial expressions. The purpose of this judgment is to determine whether the current user's facial expressions and behavior are unnatural and whether there may be a situation of coercion.
[0029] Step S1051: If the historical similarity ratio is greater than the baseline demand ratio, then control the door lock to open.
[0030] When the proportion of historical similarities is greater than the proportion of baseline demand, it indicates that the user's expression is not unnatural, and the possibility of being coerced is low. Therefore, it is sufficient to control the door lock to open normally.
[0031] Step S1052: If the proportion of historical similarities is not greater than the proportion of baseline demand, then output an identification anomaly signal.
[0032] When the proportion of historical similarities is not greater than the baseline demand proportion, it indicates that the user's facial expression is unnatural and may be under duress. Therefore, an abnormal recognition signal is output to identify this situation. This abnormal recognition signal can be a voice message saying "Face recognition failed, please try again later" or an electrical signal to facilitate further analysis.
[0033] The steps for determining combinatorial similarity by comparing representative activation combinations with historical activation combinations include: Step S200: Randomly select and identify representative features from the representative activation combinations according to a preset fixed number to construct unit activation combinations, and determine the comparison activation combinations based on the unit activation combinations from the historical activation combinations.
[0034] The fixed quantity is a set value set by the staff, which is less than the total number of representative features. Unit activation combinations are constructed to simulate the combination of different features, which is convenient for subsequent analysis. Comparison activation combinations are constructed to identify the feature combinations corresponding to the unit activation combinations, which is also convenient for subsequent analysis. The method for determining the comparison activation combinations in the historical activation combinations is to combine the representative features in the historical activation combinations that are located at the same face position as the representative features in the unit activation combinations.
[0035] Step S201: Determine the effective unit weight corresponding to the unit activation combination according to the preset weight matching relationship.
[0036] The effective weight of a unit is a weight value that reflects the importance of each unit activation combination in judging similar situations. Under different unit activation combinations, the effective weight of the unit is also different because the changes of each identification representative feature are different. The weight matching relationship between the two can be determined by the staff in advance through multiple experiments, or it can be determined by the method of steps S300-S304, so as to determine the appropriate matching relationship of the current user's adaptive weight.
[0037] Step S202: Perform comparative analysis based on the unit activation combination and the comparison activation combination to determine the unit comparison degree, and calculate the combination similarity based on all unit comparison degrees and the corresponding unit effective weights.
[0038] The unit similarity is a parameter value representing the similarity between unit activation combinations and paired activation combinations. It can be determined by comparing the identification representative features in the unit activation combination with the corresponding identification representative features in the paired activation combination, and then averaging all of them to determine the unit similarity. The combined similarity is determined by multiplying all the unit similarities by the corresponding effective unit weights and then summing them all.
[0039] It also includes a step for constructing weighted matching relationships, which includes: Step S300: Randomly select two representative features from the unit activation combination to determine the relative correlation.
[0040] The relative correlation is the degree of correlation between two representative features under the user's facial expression. For example, the higher the relative correlation, the stronger the correlation between the two representative features. That is, when one feature changes, the other feature is more likely to change accordingly. The relative correlation between the two can be determined through steps S400-S403.
[0041] Step S301: Summarize the representative features of the identification with a relative correlation degree greater than the preset requirement correlation degree into the preset initially empty feature summarization set.
[0042] Demand relevance is the minimum relative relevance set by staff for identifying two features as having a high degree of correlation. This is achieved by grouping highly correlated representative features into a feature set to distinguish them and facilitate subsequent analysis. The method for grouping representative features is as follows: For example, given three representative features A, B, and C, where the relative relevance between A and B is greater than the demand relevance, and the relative relevance between B and C is greater than the demand relevance, but the relative relevance between A and C is not greater than the demand relevance, since B can be grouped into a feature set along with both A and C, A, B, and C are grouped into a single feature set. When a representative feature does not have a relative relevance greater than the demand relevance with any other representative feature, that representative feature is grouped into a feature set independently.
[0043] Step S302: Count the number of individual features based on the identified representative features, and count the number of inductive features based on the feature induction set.
[0044] The number of individual features is the total number of representative features identified in a face, while the number of inductive features is the total number of non-empty feature inductive sets.
[0045] Step S303: Calculate and determine the expression quantity based on the number of monomeric features and the number of inductive features.
[0046] The number of expressions is the number of individual features minus the number of inductive features.
[0047] Step S304: Determine the effective weight of the unit corresponding to the number of expressions based on the preset expression matching relationship, and construct the weight matching relationship based on the effective weight of the unit and the unit activation combination.
[0048] Different expression counts indicate different correlations between features, and therefore different effective unit weights. The larger the expression count, the stronger the correlation between features, and the larger the effective unit weight. The expression matching relationship between the two is determined in advance by the staff. At this time, a more suitable weight matching relationship can be constructed for analysis based on the determined effective unit weight and the corresponding unit activation combination.
[0049] The steps for randomly selecting two representative features from the unit activation combination to determine the relative correlation include: Step S400: Randomly select one identification representative feature from the unit activation combination and define it as a fixed representative feature, and define the remaining identification representative features as variable representative features, and construct a relative feature combination based on the fixed representative feature and the variable representative feature.
[0050] By defining fixed representative features and variable representative features, different identification representative features can be identified and distinguished, which facilitates subsequent analysis; relative feature combination is the combination formed by fixed representative features and variable representative features.
[0051] Step S401: Under the historical activation combination, determine the equivalent point based on the fixed representative feature, and under the equivalent point, determine the equivalent principal feature based on the changing representative feature.
[0052] The equivalent point is the time point at which the feature corresponding to the fixed representative feature under the historical activation combination is consistent with the current fixed representative feature. The equivalent main feature is the feature at the same face position point as the changing representative feature under the equivalent point.
[0053] Step S402: Analyze the variable representative features and equivalent principal features to determine the relative degree of the individual, and analyze and calculate the combination correlation degree of the relative feature combination under the fixed representative features based on all the individual relative degrees.
[0054] Individual relative degree is the relative similarity between the variable representative feature and the equivalent principal feature, while the combined correlation degree is the average of all individual relative degrees.
[0055] Step S403: Calculate the relative correlation degree based on the combination correlation degree of the same relative feature with different fixed representative features.
[0056] A more accurate relative correlation can be obtained by averaging the combined correlation degrees calculated from different fixed representative features under the same combination of relative features.
[0057] After identifying abnormal signal output, the facial recognition unlocking method based on smart door locks also includes: Step S500: Determine the erroneous verification features corresponding to the current face recognition features based on the preset verification and matching relationship.
[0058] The error verification feature is the feature that the user corresponding to the face recognition feature needs to input in advance to verify the face recognition when an abnormal signal is output, without being coerced. Different users have different error verification features, and the verification and matching relationship between the two is input and stored by the user in advance.
[0059] Step S501: Construct a verification interval with a preset verification duration on the time axis, using the current time point as the leading point, and obtain user input features in real time based on error verification features within the verification interval.
[0060] The verification time is the time set by the staff for users to verify when an abnormal recognition signal is output. The verification interval is constructed to acquire and analyze the data within the verification time. The user input feature is the facial feature information at the location corresponding to the user's incorrect verification feature within the verification interval.
[0061] Step S502: Determine whether there exists a user input feature that matches the error verification feature.
[0062] The purpose of the judgment is to determine whether the user has verified that they were not coerced.
[0063] Step S5021: If a user input feature matches the error verification feature, then control the door lock to open.
[0064] When a user input feature matches the error verification feature, it indicates that the user has not been coerced, and the door lock can be opened normally.
[0065] Step S5022: If no user input feature matches the error verification feature, output an identification error signal.
[0066] When no user input feature matches the error check feature, it indicates a high probability that the user has been coerced. Therefore, an error recognition signal is output to identify this situation, facilitating further analysis and processing.
[0067] After identifying the error signal output, the facial recognition unlocking method based on smart door locks also includes: Step S600: Obtain external infrared features.
[0068] External infrared signatures are infrared signatures obtained using infrared sensors installed on the door lock.
[0069] Step S601: Determine whether the external infrared features only contain users corresponding to the facial recognition features.
[0070] The purpose of the assessment is to determine whether there are other individuals present, thereby analyzing the possibility of coercion.
[0071] Step S6011: If the external infrared features only include the user corresponding to the face recognition features, then output a recognition error reminder signal.
[0072] When the external infrared features only contain the user corresponding to the facial recognition features, it indicates that, based on the analysis, only the current user is present. In this case, the risk of coercion cannot be ruled out. Therefore, an error recognition alert signal is output to inform the user of the situation. If coercion exists, the coercing person is informed that the door lock system is malfunctioning and not caused by the coerced person. If there is no coercion, the user is informed of the specific situation of the door lock to facilitate subsequent unlocking.
[0073] Step S6012: If the external infrared features not only include the user corresponding to the facial recognition features, then control the door lock to "virtually" open and output a recognition error reminder signal.
[0074] When the external infrared signature includes not only the user corresponding to the facial recognition signature, it indicates that the current user is likely being coerced. In this case, the door lock is virtually opened to ensure that the sound of the latch springing open is emitted, but the door is not actually opened. This allows the coercing person to believe that the door itself is malfunctioning, rather than that the coerced person is not coercing, thus improving their own safety. At the same time, an error recognition reminder signal is output to further inform the coercing person that the door lock system itself is malfunctioning via voice.
[0075] Reference Figure 2 Based on the same inventive concept, embodiments of the present invention provide a facial recognition unlocking system based on a smart door lock, comprising: The acquisition module is used to acquire facial recognition features; The processing module, connected to the acquisition and judgment modules, is used for information storage and processing; The judgment module, connected to the acquisition and processing modules, is used for judging information. The processing module determines the user's recognition status based on facial recognition features, and when the judgment module determines that the user's recognition status is consistent with the preset successful recognition status, it determines the recognition representative features from the facial recognition features. The processing module constructs a historical interval with a preset historical duration on a preset time axis, with the current time point as the endpoint, and determines the same activation time period in the historical interval based on the current face recognition features; The processing module combines each representative feature of identification to construct a historical activation combination under the same activation period, and combines the current representative features of identification to construct a representative activation combination; The processing module compares and analyzes representative activation combinations and historical activation combinations to determine the combination similarity. Historical activation combinations with a combination similarity greater than the preset required similarity are defined as historical similar combinations. The module also calculates the historical similarity ratio based on historical similar combinations and historical activation combinations. The judgment module determines whether the proportion of historical similarities is greater than the preset baseline requirement proportion; If the judgment module determines that the historical similarity ratio is greater than the baseline demand ratio, the processing module will control the door lock to open. If the judgment module determines that the proportion of historical similarities is not greater than the baseline requirement proportion, the processing module outputs an abnormal identification signal. The combination similarity determination module is used to determine the combination similarity between combinations; The weight matching relationship construction module is used to construct appropriate weight matching relationships; The relative correlation determination module is used to determine the relative correlation between the representative features. The abnormal signal output module is used to analyze and process the situation after the abnormal signal is output. The error signal output module is used to analyze and process the situation after an error signal is output.
[0076] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
Claims
1. A facial recognition unlocking method based on a smart door lock, characterized in that, include: Obtain facial recognition features; The user's recognition status is determined based on facial recognition features, and when the user's recognition status matches the preset successful recognition status, a representative feature is determined from the facial recognition features. On a preset timeline, a historical interval with a preset historical duration is constructed with the current time point as the endpoint, and the same activation time period is determined within the historical interval based on the current facial recognition features; Under the same activation period, each identification representative feature is combined to construct a historical activation combination, and the current identification representative features are combined to construct a representative activation combination; The similarity of the combination is determined by comparing and analyzing the representative activation combination and the historical activation combination. The historical activation combination with a similarity greater than the preset requirement similarity is defined as the historical similar combination. The historical similarity ratio is determined by calculating based on the historical similar combination and the historical activation combination. Determine whether the proportion of historical similarities is greater than the preset baseline demand proportion; If the proportion of historically similar demand is greater than the baseline demand proportion, then the door lock will be controlled to open. If the proportion of historical similarities is not greater than the baseline demand proportion, then an anomaly signal will be output.
2. The facial recognition unlocking method based on a smart door lock according to claim 1, characterized in that, The steps for determining combinatorial similarity by comparing representative activation combinations with historical activation combinations include: In the representative activation combination, a fixed number of representative features are randomly selected to construct a unit activation combination, and the comparison activation combination is determined based on the unit activation combination in the historical activation combination. The effective weight of each unit is determined based on a preset weight matching relationship. The unit alignment score is determined by comparing the unit activation combination and the alignment activation combination, and the combination similarity is determined by calculating all the unit alignment scores and the corresponding effective unit weights.
3. The facial recognition unlocking method based on a smart door lock according to claim 2, characterized in that, It also includes a step for constructing weighted matching relationships, which includes: Two representative features are randomly selected from the unit activation combination to determine the relative correlation. Identifying representative features whose relative correlation is greater than the preset requirement correlation are categorized into a preset, initially empty feature categorization set; The number of individual features is determined by counting based on the identified representative features, and the number of inductive features is determined by counting based on the feature inductive set. The number of expressions is determined by calculating the number of monomeric features and the number of inductive features; The effective weight of a unit is determined based on the preset expression matching relationship, and a weight matching relationship is constructed based on the effective weight of the unit and the unit activation combination.
4. The facial recognition unlocking method based on a smart door lock according to claim 3, characterized in that, The steps for randomly selecting two representative features from the unit activation combination to determine the relative correlation include: In the unit activation combination, a representative feature is randomly selected and defined as a fixed representative feature, and the remaining representative features are defined as variable representative features. A relative feature combination is constructed based on the fixed representative feature and the variable representative feature. Under the historical activation combination, the equivalent point is determined based on the fixed representative characteristics, and the equivalent main characteristics are determined based on the changing representative characteristics under the equivalent point; The relative degree of a single entity is determined by analyzing the changing representative features and the equivalent principal features, and the degree of correlation of the combination of relative features with the fixed representative features is determined by analyzing and calculating all the relative degrees of a single entity. The relative correlation is determined by calculating the correlation between the same relative feature and different fixed representative features.
5. The facial recognition unlocking method based on a smart door lock according to claim 1, characterized in that, After identifying abnormal signal output, the facial recognition unlocking method based on smart door locks also includes: The incorrect verification features corresponding to the current face recognition features are determined based on the preset verification and matching relationship; On the timeline, a verification interval with a preset verification duration is constructed with the current time point as the leading point, and user input features are obtained in real time based on the error verification features within the verification interval; Determine if a user input feature matches an error verification feature; If a user input feature matches the error verification feature, then the door lock will be opened. If no user input feature matches the error check feature, an identification error signal is output.
6. The facial recognition unlocking method based on a smart door lock according to claim 5, characterized in that, After identifying the error signal output, the facial recognition unlocking method based on smart door locks also includes: Acquire external infrared features; Determine whether the external infrared features only contain users corresponding to facial recognition features; If the external infrared features only include the user corresponding to the facial recognition features, then an error recognition alert signal will be output. If the external infrared features include not only the user corresponding to the facial recognition features, the door lock will be controlled to "virtually" open, and an error recognition reminder signal will be output.
7. A facial recognition unlocking system based on a smart door lock, characterized in that, include: The acquisition module is used to acquire facial recognition features; The processing module, connected to the acquisition and judgment modules, is used for information storage and processing; The judgment module, connected to the acquisition and processing modules, is used for judging information. The processing module determines the user's recognition status based on facial recognition features, and when the judgment module determines that the user's recognition status is consistent with the preset successful recognition status, it determines the recognition representative features from the facial recognition features. The processing module constructs a historical interval with a preset historical duration on a preset time axis, with the current time point as the endpoint, and determines the same activation time period in the historical interval based on the current face recognition features; The processing module combines each representative feature of identification to construct a historical activation combination under the same activation period, and combines the current representative features of identification to construct a representative activation combination; The processing module compares and analyzes representative activation combinations and historical activation combinations to determine the combination similarity. Historical activation combinations with a combination similarity greater than the preset required similarity are defined as historical similar combinations. The module also calculates the historical similarity ratio based on historical similar combinations and historical activation combinations. The judgment module determines whether the proportion of historical similarities is greater than the preset baseline requirement proportion; If the judgment module determines that the proportion of historical similarities is greater than the baseline demand proportion, the processing module will control the door lock to open. If the judgment module determines that the proportion of historical similarities is not greater than the baseline requirement proportion, the processing module outputs an identification anomaly signal.