Intelligent driving identity information cross-domain authentication system based on face recognition

Abstract

The invention discloses an intelligent driving identity information cross-domain authentication system based on face recognition, and belongs to the field of intelligent driving. An intelligent driving identity information cross-domain authentication system based on face recognition comprises a face acquisition unit, a cross-domain authentication unit and an emergency processing unit. According to the method, the problem that the identity of a user is difficult to accurately recognize in a cross-domain manner in the prior art is solved, and whether the identity of the driver is legal or not can be verified by collecting and extracting face feature vectors in real time and comparing the face feature vectors with feature vectors in a cross-domain identity information database; intelligent driving identity information cross-domain authentication is achieved through face recognition, meanwhile, the execution state of comparison operation in the feature comparison process is monitored in real time, the identity verification process is recorded, the accuracy and reliability of identity authentication can be improved, when face feature comparison is abnormal, a secondary identity verification mechanism is started, and the safety of identity authentication is improved. The multi-factor verification mode greatly enhances the security of identity authentication, and effectively prevents illegal invasion.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to an intelligent driving identity information cross-domain authentication system based on face recognition. Background Art

[0002] With the rapid development and widespread application of intelligent driving technology, the intelligence level of vehicles and human-machine interaction capabilities have been significantly improved. As a key link in intelligent driving systems, identity authentication has become increasingly important. However, traditional intelligent driving identity authentication methods have the following limitations:

[0003] 1. Existing identity authentication methods are based on keys or cards. Keys or cards are easily lost or stolen, and there is a possibility that they can be copied and used by others, making it difficult to accurately identify the identity of the user.

[0004] 2. Existing identity authentication is limited to single-vehicle identity management and lacks cross-vehicle collaborative authentication capabilities, resulting in inefficient repeated identity information verification.

[0005] Therefore, it does not meet the existing needs. For this reason, we propose an intelligent driving identity information cross-domain authentication system based on face recognition. Summary of the Invention

[0006] The purpose of the present invention is to provide a cross-domain authentication system for intelligent driving identity information based on face recognition. By collecting facial images in real time and extracting the driver's facial feature vectors, the extracted facial feature vectors are compared one by one with the help of a cross-domain identity information database to efficiently compare and verify the driver's identity. In this way, cross-domain authentication of intelligent driving identity information is realized through face recognition. At the same time, secondary verification and other measures are initiated in abnormal situations to enhance the system's ability to deal with risks such as illegal intrusion, thereby solving the problems raised in the above-mentioned background technology.

[0007] To achieve the above-mentioned object, the present invention provides the following technical solution: an intelligent driving identity information cross-domain authentication system based on face recognition, the system comprising a face acquisition unit, a cross-domain authentication unit and an emergency processing unit;

[0008] The face acquisition unit is configured to acquire the driver's face image in real time through the vehicle-mounted camera and perform preprocessing operations, and extract the face feature vector from the preprocessed face image;

[0009] The cross-domain authentication unit is configured to receive the facial feature vector of the face collection unit through the cross-domain authentication server, compare the received facial feature vector with the feature vector in the cross-domain identity information database one by one, verify whether the driver's identity is legitimate according to a preset similarity threshold, and record the identity authentication process;

[0010] Among them, the cross-domain identity information database stores facial feature vectors and the fingerprint features corresponding to the driver;

[0011] Among them, the cosine similarity algorithm is used to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database. The basic similarity calculation is performed by using the cosine similarity algorithm. When any feature vector is updated, the vector difference before and after the update is retrieved to calculate the vector difference, and then the feature update strength is obtained. The pending similarity between the updated facial feature vector and the database feature vector is obtained, and the pending similarity is compensated and adjusted in combination with the feature update strength;

[0012] The emergency processing unit is configured to start a secondary identity verification mechanism when there is an abnormality in the comparison result between the face feature vector of the cross-domain authentication unit and the feature vector in the cross-domain identity information database.

[0013] Furthermore, the cross-domain authentication unit includes:

[0014] a local storage module configured to establish a cross-domain identity information database within the cross-domain authentication server, the cross-domain identity information database being used to store driver identity information and corresponding facial feature vectors registered by multiple cross-domain systems, wherein the driver identity information includes driver's license information, vehicle use permission information, and driving preference settings;

[0015] a feature comparison module configured to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database using a cosine similarity algorithm, and verify whether the driver's identity is legitimate using a preset similarity threshold;

[0016] If the similarity between the facial feature vector and the feature vector in the cross-domain identity information database is greater than or equal to the similarity threshold, the driver's identity is determined to be legal;

[0017] If the similarity between the facial feature vector and the feature vector in the cross-domain identity information database is lower than the similarity threshold, the driver's identity is determined to be illegal;

[0018] The verification record module is configured to build a distributed record book based on blockchain technology to record the identity verification process, including:

[0019] At the beginning of the authentication process, record the time the request was initiated, the source device ID, and the basic driver information;

[0020] During the comparison process, the key steps and intermediate results of feature comparison are recorded in real time;

[0021] After the comparison is completed, the final authentication result, the authentication completion time and cross-domain communication related information are recorded, where the cross-domain communication related information includes the interaction time and data transmission volume with the cross-domain authentication server.

[0022] Furthermore, the cosine similarity algorithm is used to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database, further comprising:

[0023] When any one of the facial feature vectors or the feature vectors in the cross-domain identity information database is updated, the feature vector before the update and the feature vector after the update are retrieved;

[0024] Performing difference processing on the feature vector before updating and the feature vector after updating to obtain a vector difference between the feature vector before updating and the feature vector after updating;

[0025] Obtaining a feature update strength using a vector difference between the feature vector before the update and the feature vector after the update;

[0026] Retrieving the similarity between the face feature vector obtained after any feature vector is updated and the feature vector in the cross-domain identity information database as the similarity to be processed;

[0027] Performing compensation adjustment on the to-be-processed similarity using the feature update strength to obtain a similarity value after compensation adjustment;

[0028] The similarity value after compensation adjustment is used as the final similarity value corresponding to the updated feature vector.

[0029] Furthermore, compensating and adjusting the similarity to be processed by using the feature update strength to obtain a similarity value after compensation adjustment includes:

[0030] Retrieving the similarity value before the feature vector is updated and the similarity value to be processed after the feature vector is updated;

[0031] Performing difference processing on the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated, to obtain the difference between the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated;

[0032] The similarity change direction parameter is obtained by using the numerical difference between the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated, wherein the similarity change direction parameter is sign(S c ); where S c The numerical difference between the similarity value before the feature vector is updated and the similarity value to be processed after the feature vector is updated;

[0033] Get the L2 norm between the eigenvector before and after the update || v new -v old ||2; where v new and v old Represent the feature vector before and after the update respectively;

[0034] Using the L2 norm ||v between the feature vector before and after the update new -v old ||2 Obtain feature change gradient Where Δt represents the update time interval;

[0035] Retrieve feature update strength;

[0036] Using the feature update strength and feature change gradient ||▽V|| combined with the similarity change direction parameter sign(S c ) performing compensation adjustment on the similarity to be processed to obtain a similarity value after compensation adjustment.

[0037] Furthermore, the emergency processing unit includes: an anomaly monitoring module configured to continuously monitor the comparison results transmitted by the cross-domain authentication unit in real time, immediately generate an anomaly warning signal once an anomaly is found, and analyze the anomaly of each comparison result to determine whether it is a difference in facial features itself, data transmission anomaly, or potential illegal intrusion behavior;

[0038] A secondary verification module is configured to immediately initiate a secondary identity verification process upon receiving an abnormality warning signal, while recording the time of the abnormality, the abnormality feature vector, and the vehicle location information;

[0039] The security linkage module is configured to immediately send a signal to lock the vehicle's starting system when there is an abnormality in the comparison result and the secondary verification fails. At the same time, it turns on the vehicle's anti-theft alarm device, emits a high-decibel alarm and flashes the alarm light, and sends the time of the abnormality, facial image, vehicle location information, identity authentication process and secondary verification process to the cross-domain authentication server, which notifies the vehicle owner and traffic management department.

[0040] Furthermore, the feature comparison module includes:

[0041] A threshold adjustment module is configured to dynamically adjust the similarity threshold according to the current authentication scenario, raising the similarity threshold for nighttime or outdoor scenarios and lowering the similarity threshold for daytime or indoor scenarios;

[0042] On the other hand, historical data is regularly analyzed to calculate the similarity distribution of different driver feature vectors, and the similarity threshold is dynamically adjusted based on the mean and standard deviation of the similarity;

[0043] The comparison monitoring module is configured to monitor the execution status of the comparison operation in real time during the feature comparison process, including the algorithm running time, memory usage, and comparison progress. If any abnormal delay, memory overflow, or comparison progress stagnation is found in the comparison process, an alarm mechanism will be immediately triggered to notify the cross-domain authentication server to intervene;

[0044] The report generation module is configured to generate a comparison result report after the comparison is completed, including the similarity value, the comparison time, and any abnormal situation descriptions that occurred during the comparison process. The generated comparison result report will be fed back to the threshold adjustment module as historical data to further optimize the adjustment strategy of the similarity threshold.

[0045] Furthermore, the verification record module includes:

[0046] Before the verification record is written into the blockchain distributed record book, the record data is encrypted using the AES algorithm. At the same time, the record data is digitally signed using the private key of the cross-domain authentication server.

[0047] Regularly synchronize the local blockchain distributed record book with other nodes in the network, and at the same time, back up and store the blockchain distributed record book.

[0048] Furthermore, upon receiving the abnormal warning signal, the secondary verification module immediately initiates the secondary identity verification process, specifically:

[0049] The vehicle's built-in fingerprint recognition device requires the driver to collect his fingerprint, which is then encrypted and transmitted to a cross-domain identity information database for comparison to further confirm the driver's identity.

[0050] At the same time, set the corresponding verification time limit and the number of failed attempts limit;

[0051] If the verification fails within the specified time or fails multiple times, it will be determined as an illegal intrusion and the emergency security measures of the security linkage module will be triggered.

[0052] Furthermore, the cross-domain authentication unit complies with a unified communication protocol, including:

[0053] HTTPS communication protocol based on TLS / SSL encryption protocol, used to establish a communication link between the local intelligent driving system and the cross-domain authentication server;

[0054] And before establishing a connection with the cross-domain authentication server, perform certificate verification to verify the legitimacy of the cross-domain authentication server's identity;

[0055] At the same time, it has the function of data packet encapsulation and decapsulation, encapsulates the facial feature vector, adds communication header information, and when receiving the response from the cross-domain authentication server, decapsulates the data packet, extracts the facial feature vector and performs integrity verification. If the data verification fails, it will re-request the face acquisition unit to transmit.

[0056] Furthermore, the face acquisition unit includes:

[0057] An image acquisition module is configured to use an onboard camera to capture the driver's facial image in real time, and is also equipped with an infrared fill light that automatically adjusts the light intensity and angle of the infrared fill light according to the vehicle environment;

[0058] The image processing module is configured to pre-process the face images collected in real time, including image enhancement and normalization processing, wherein:

[0059] Image enhancement: Brightness adjustment is performed by changing the brightness value of the face image, contrast of the face image is increased by using histogram equalization or Laplace filtering, and sharpening of the face image is performed by using the Laplace operator;

[0060] Normalization processing: used to normalize the face image after image enhancement to unify the size and grayscale range of the face image;

[0061] The feature extraction module is configured as a convolutional neural network algorithm based on deep learning and is trained with a large amount of existing facial data to extract facial feature vectors from preprocessed facial images. The extracted facial feature vectors are used for face recognition in the cross-domain authentication unit.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention collects and extracts the driver's facial feature vectors in real time to provide accurate information for subsequent authentication, compares the extracted facial feature vectors one by one with the help of a cross-domain identity information database, and verifies whether the driver's identity is legal based on a preset similarity threshold, thereby realizing cross-domain authentication of intelligent driving identity information through face recognition. During the feature comparison process, the execution status of the comparison operation is monitored in real time to avoid identity authentication failure or security vulnerabilities caused by problems in the comparison process. The identity authentication process is recorded in a distributed record book constructed based on blockchain technology to ensure data security and traceability. When an abnormality occurs in the facial feature comparison, a secondary identity verification mechanism is activated. The multi-factor verification method greatly enhances the security of identity authentication and effectively prevents illegal intrusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1This is a structural diagram of the cross-domain authentication system for intelligent driving identity information based on face recognition of the present invention. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] In order to solve the technical problems that existing technologies are difficult to accurately identify the identity of users and identity authentication is limited to bicycle identity management, please refer to Figure 1 , this embodiment provides the following technical solutions:

[0067] An intelligent driving identity information cross-domain authentication system based on face recognition, the system comprising a face acquisition unit, a cross-domain authentication unit, and an emergency processing unit;

[0068] The face acquisition unit is configured to acquire the driver's face image in real time through the vehicle-mounted camera and perform preprocessing operations, and extract the face feature vector from the preprocessed face image;

[0069] The cross-domain authentication unit is configured to receive the facial feature vector of the face collection unit through the cross-domain authentication server, compare the received facial feature vector with the feature vector in the cross-domain identity information database one by one, verify whether the driver's identity is legitimate according to a preset similarity threshold, and record the identity authentication process;

[0070] Among them, the cross-domain identity information database stores facial feature vectors and the fingerprint features corresponding to the driver;

[0071] Among them, the cosine similarity algorithm is used to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database. The basic similarity calculation is performed by using the cosine similarity algorithm. When any feature vector is updated, the vector difference before and after the update is retrieved to calculate the vector difference, and then the feature update strength is obtained. The pending similarity between the updated facial feature vector and the database feature vector is obtained, and the pending similarity is compensated and adjusted in combination with the feature update strength;

[0072] The emergency processing unit is configured to start a secondary identity verification mechanism when there is an abnormality in the comparison result between the face feature vector of the cross-domain authentication unit and the feature vector in the cross-domain identity information database.

[0073] The technical effect of the above technical solution is as follows: the face acquisition unit uses the on-board camera to collect the driver's face image in real time and performs preprocessing operations to extract the driver's face feature vector, providing accurate information for subsequent face authentication. The cross-domain authentication unit receives the feature vector through the cross-domain authentication server and compares it with the feature vector in the cross-domain identity information database one by one, thereby verifying whether the driver's identity is legal based on a preset similarity threshold. Through face recognition technology, accurate identification of the driver's identity is achieved, effectively preventing illegal driving behavior and ensuring the safety of vehicles and traffic. The driver's identity information can be recognized and used in multiple cross-domain systems, which improves the compatibility and convenience of the system and facilitates the driver's switching and use between different systems. If the cross-domain authentication unit finds an abnormality in the comparison result, the emergency processing unit will activate the secondary identity verification mechanism to further confirm the driver's identity, thereby timely discovering and responding to potential illegal intrusions, further improving the security and reliability of the system and ensuring driving safety.

[0074] Cross-domain authentication unit, including:

[0075] a local storage module configured to establish a cross-domain identity information database within the cross-domain authentication server, the cross-domain identity information database being used to store driver identity information and corresponding facial feature vectors registered by multiple cross-domain systems, wherein the driver identity information includes driver's license information, vehicle use permission information, and driving preference settings;

[0076] a feature comparison module configured to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database using a cosine similarity algorithm, and verify whether the driver's identity is legitimate using a preset similarity threshold;

[0077] If the similarity between the facial feature vector and the feature vector in the cross-domain identity information database is greater than or equal to the similarity threshold, the driver's identity is determined to be legal;

[0078] If the similarity between the facial feature vector and the feature vector in the cross-domain identity information database is lower than the similarity threshold, the driver's identity is determined to be illegal;

[0079] The verification record module is configured to build a distributed record book based on blockchain technology to record the identity verification process, including:

[0080] At the beginning of the authentication process, record the time the request was initiated, the source device ID, and the basic driver information;

[0081] During the comparison process, the key steps and intermediate results of feature comparison are recorded in real time;

[0082] After the comparison is completed, the final authentication result, the authentication completion time and cross-domain communication related information are recorded, where the cross-domain communication related information includes the interaction time and data transmission volume with the cross-domain authentication server.

[0083] The technical effects of the above technical solution are as follows: the local storage module establishes a cross-domain identity information database in the cross-domain authentication server, which can store the driver identity information and corresponding facial feature vectors registered by multiple cross-domain systems, and realize the centralized management of driver identity information. The feature comparison module adopts the cosine similarity algorithm to scientifically calculate the similarity between the received facial feature vector and the feature vector in the database, and accurately determine whether the driver's identity is legal based on the preset similarity threshold, effectively preventing illegal driving behavior and ensuring the safety of vehicles and traffic. The verification record module builds a distributed record ledger based on blockchain technology, comprehensively records each key link of the identity authentication process, ensures the transparency, non-tamperability and traceability of the identity authentication process, provides a reliable basis for post-audit, problem investigation and responsibility tracing, and enhances the credibility and reliability of identity authentication.

[0084] Specifically, the cosine similarity algorithm is used to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database, and further includes:

[0085] When any one of the facial feature vectors or the feature vectors in the cross-domain identity information database is updated, the feature vector before the update and the feature vector after the update are retrieved;

[0086] Performing difference processing on the feature vector before updating and the feature vector after updating to obtain a vector difference between the feature vector before updating and the feature vector after updating;

[0087] Obtaining a feature update strength using a vector difference between the feature vector before the update and the feature vector after the update;

[0088] The feature update strength is obtained by the following formula:

[0089]

[0090] Among them, ΔV represents the feature update strength; v new and v old They represent the feature vector before and after updating respectively; E(v new ) represents the feature information entropy of the updated feature vector; specifically, It is used to eliminate the influence of the modulus of the feature vector itself on the judgment of the degree of difference, so that this part of the results focuses on the relative change amplitude of the feature vector update. 10 [1+E(v new)] Through logarithmic operation, the feature information entropy is converted into a gain term for the feature update strength. The larger the feature information entropy, the more complex the information contained in the updated feature, and the greater the contribution to the feature update strength, which reflects the logic of the influence of the intrinsic information richness of the feature on the update strength. The above formula comprehensively considers the relative difference amplitude of the feature vector update and the information complexity of the updated feature to obtain the feature update strength ΔV. The relative difference amplitude is multiplied by the information entropy gain term to achieve a comprehensive measurement of the degree of change brought to the vector by the feature update from the two dimensions of "spatial difference" and "information richness difference" of the feature vector, providing a quantitative basis for subsequent similarity compensation adjustment, so that the impact of feature update can be reasonably incorporated into the similarity evaluation system.

[0091] Retrieving the similarity between the face feature vector obtained after any feature vector is updated and the feature vector in the cross-domain identity information database as the similarity to be processed;

[0092] Performing compensation adjustment on the to-be-processed similarity using the feature update strength to obtain a similarity value after compensation adjustment;

[0093] The similarity value after compensation adjustment is used as the final similarity value corresponding to the updated feature vector.

[0094] The technical solution achieves the following: By constructing a feature update strength formula, combined with vector difference and feature information entropy, it accurately quantifies the degree of change brought about by feature vector updates. This compensates for the fact that traditional similarity calculations fail to consider the dynamic impact of feature updates, making cross-domain facial feature matching more responsive to real-world dynamic changes and improving the accuracy of similarity assessment in applications such as identity recognition. By compensating and adjusting the updated similarity based on the feature update strength, the similarity value dynamically reflects the true matching relationship after the feature update, enhancing the adaptability and reliability of cross-domain identity information databases in feature update scenarios and ensuring the stability and accuracy of business processes such as identity recognition and verification. On the other hand, in existing cross-domain identity recognition scenarios, when facial feature vectors (or database feature vectors) are updated, traditional cosine similarity algorithms fail to consider the "dynamic differences between features before and after the update," resulting in the updated similarity failing to reflect the true matching relationship. For example, a small feature update without compensation may result in a misidentification, or a large feature update may be ignored, leading to identity misidentification. After the feature vector is updated, the information entropy (complexity) it contains will change the nature of the feature, but the traditional similarity calculation only focuses on the "vector space distance" and ignores the impact of the "feature intrinsic information richness" on the matching results. For example, a feature update with high information entropy may be misjudged as a large difference due to changes in details, or the matching degree should be improved due to rich details but is not reflected. The relative difference of features before and after the update is refined to eliminate the interference of feature modulus on difference judgment, log 10 [1+E(v new )] Incorporating "feature complexity changes" allows similarity compensation to be deeply associated with "update amplitude + information richness", solving the problem of "similarity after update is disconnected from the true matching relationship", and ensuring the accuracy of identity recognition when features are dynamically updated. Incorporating "feature complexity changes" allows similarity compensation to be deeply associated with "update amplitude + information richness", solving the problem of "similarity after update is disconnected from the true matching relationship", and ensuring the accuracy of identity recognition when features are dynamically updated. The above technical solution in this embodiment can adaptively balance the impact of "feature updates of different complexities" on similarity through information entropy compensation, improve the robustness of identity recognition in cross-domain scenarios, and solve the problem of "matching deviation caused by cross-domain feature complexity differences" in traditional methods.

[0095] Specifically, using the feature update strength to compensate and adjust the similarity to be processed, and obtaining a similarity value after compensation and adjustment, includes:

[0096] Retrieving the similarity value before the feature vector is updated and the similarity value to be processed after the feature vector is updated;

[0097] Performing difference processing on the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated, to obtain the difference between the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated;

[0098] The similarity change direction parameter is obtained by using the numerical difference between the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated, wherein the similarity change direction parameter is sign(S c ); where S c The numerical difference between the similarity value before the feature vector is updated and the similarity value to be processed after the feature vector is updated;

[0099] Get the L2 norm between the eigenvector before and after the update || v new -v old ||2; where v new and v old Represent the feature vector before and after the update respectively;

[0100] Using the L2 norm ||v between the feature vector before and after the update new -v old ||2 Obtain feature change gradient Where Δt represents the update time interval;

[0101] Retrieve feature update strength;

[0102] Use the features to update the intensity and feature change gradient Combined with the similarity change direction parameter sign(S c ) performing compensation adjustment on the similarity to be processed to obtain a similarity value after compensation adjustment.

[0103] The similarity value after compensation adjustment is obtained by the following formula:

[0104]

[0105] Among them, S comp Indicates the similarity value after compensation adjustment; S new represents the similarity to be processed; λ represents the update compensation coefficient, which ranges from 0.3 to 0.8; ΔV represents the feature update strength; σ v Represents a historical feature fluctuation benchmark of a feature vector in a cross-domain identity information database corresponding to the driver; Represents the feature change gradient. Specifically, Chinese sign(S c ) determines whether the compensation is "positive enhancement" (for example, the similarity should be higher after the update, so the compensation is increased) or "negative inhibition" (for example, the similarity should be lower after the update, so the compensation is reduced) to solve the problem of "the compensation direction being out of line with actual needs". Used to solve the problem of "single update intensity leading to over-compensation / under-compensation"; When the feature change gradient is large (updates quickly and may be abnormal), the compensation term approaches 1, moderately enhancing the similarity correction; when the gradient is small (updates smoothly), the compensation term approaches 0, weakening the impact of the gradient on compensation, and solving the problem of "feature mutation causing compensation logic failure".

[0106] The technical effect of the above technical solution is: comprehensively considering multiple dimensional factors such as the direction of similarity change before and after feature update, feature change gradient, feature update intensity, and historical feature fluctuation benchmark, the similarity to be processed after the feature vector is updated is compensated and adjusted, solving the problem that traditional methods do not fully consider the impact of dynamic changes in features on similarity, so that the compensated similarity is more in line with the actual matching relationship of features, and improves the accuracy and reliability of similarity evaluation in scenarios such as cross-domain identity recognition. Starting from the similarity numerical difference caused by feature update, combined with dynamic parameters such as the gradient and intensity of feature changes, a complete compensation adjustment logic is constructed, so that similarity compensation can adapt to the process of continuous evolution of features, ensuring the stability and accuracy of identity recognition, verification and other businesses in scenarios with frequent feature updates, and enhancing the dynamic adaptability of cross-domain identity information databases. At the same time, in cross-domain identity recognition, the feature update rules of different domains are significantly different (such as frequent feature updates in domain A and slow feature updates in domain B). The above technical solution of this embodiment adapts the feature evolution characteristics of different domains through "rate-historical benchmark" and adapts the updated similarity offset through "direction-intensity", thereby enhancing the universality of identity recognition in cross-domain scenarios and solving the problem of "uneven cross-domain compensation effect" of traditional methods.

[0107] The emergency processing unit includes: an anomaly monitoring module, which is configured to continuously monitor the comparison results transmitted by the cross-domain authentication unit in real time, immediately generate an anomaly warning signal if an anomaly is found, and analyze the anomaly of each comparison result to determine whether it is a difference in facial features itself, an anomaly in data transmission, or a potential illegal intrusion;

[0108] A secondary verification module is configured to immediately initiate a secondary identity verification process upon receiving an abnormality warning signal, while recording the time of the abnormality, the abnormality feature vector, and the vehicle location information;

[0109] The security linkage module is configured to immediately send a signal to lock the vehicle's starting system when there is an abnormality in the comparison result and the secondary verification fails. At the same time, it turns on the vehicle's anti-theft alarm device, emits a high-decibel alarm and flashes the alarm light, and sends the time of the abnormality, facial image, vehicle location information, identity authentication process and secondary verification process to the cross-domain authentication server, which notifies the vehicle owner and traffic management department.

[0110] The technical effects of the above technical solution are as follows: the anomaly monitoring module continuously monitors the comparison results transmitted by the cross-domain authentication unit in real time, can quickly detect anomalies and immediately generate anomaly warning signals, buying time for subsequent emergency processing, and analyzing the anomaly situation, which helps to take more targeted emergency measures. After receiving the anomaly warning signal, the secondary verification module immediately initiates the secondary identity verification process, which can further confirm the driver's identity and improve the accuracy and reliability of identity verification. At the same time, it records the time of the anomaly occurrence, the anomaly feature vector and the vehicle location information, providing important basis for subsequent investigation and processing. When the comparison result is abnormal and the secondary verification fails, the security linkage module immediately sends a signal to lock the vehicle's starting system, preventing illegal driving behavior from occurring, effectively protecting the safety of the vehicle and personnel, and at the same time activates the vehicle's anti-theft alarm device, which can deter potential illegal intruders and attract the attention of people around. The relevant anomaly information is sent to the cross-domain authentication server, which notifies the vehicle owner and traffic management department, realizing rapid transmission and sharing of information, facilitating the vehicle owner and traffic management department to take corresponding measures in a timely manner, and further improving security emergency capabilities.

[0111] Feature comparison module, including:

[0112] A threshold adjustment module is configured to dynamically adjust the similarity threshold according to the current authentication scenario, raising the similarity threshold for nighttime or outdoor scenarios and lowering the similarity threshold for daytime or indoor scenarios;

[0113] On the other hand, historical data is regularly analyzed to calculate the similarity distribution of different driver feature vectors, and the similarity threshold is dynamically adjusted based on the mean and standard deviation of the similarity;

[0114] The comparison monitoring module is configured to monitor the execution status of the comparison operation in real time during the feature comparison process, including the algorithm running time, memory usage, and comparison progress. If any abnormal delay, memory overflow, or comparison progress stagnation is found in the comparison process, an alarm mechanism will be immediately triggered to notify the cross-domain authentication server to intervene;

[0115] The report generation module is configured to generate a comparison result report after the comparison is completed, including the similarity value, the comparison time, and any abnormal situation descriptions that occurred during the comparison process. The generated comparison result report will be fed back to the threshold adjustment module as historical data to further optimize the adjustment strategy of the similarity threshold.

[0116] The technical effects of the above technical solution are as follows: the threshold adjustment module can dynamically adjust the similarity threshold according to the current authentication scenario, so that the system can adapt to different environmental conditions and improve the accuracy and reliability of identity authentication. It can also regularly analyze historical data and dynamically adjust the similarity threshold according to the similarity distribution, further optimizing the threshold setting, so that the system can better adapt to the characteristic differences of different drivers, and improve the adaptability and accuracy of identity authentication. The comparison monitoring module monitors the execution status of the comparison operation in real time, so as to promptly discover and handle abnormal situations in the comparison process, ensure the stability and reliability of the comparison process, and prevent system failures or freezes from affecting the timeliness and accuracy of identity authentication. The report generation module generates a detailed comparison result report after the comparison is completed, providing accurate data support for subsequent analysis and optimization, and feeding back the comparison result report to the threshold adjustment module to form a closed-loop optimization mechanism, making the threshold adjustment more scientific and reasonable, and further improving the accuracy and stability of identity authentication.

[0117] Verification record module, including:

[0118] Before the verification record is written into the blockchain distributed record book, the record data is encrypted using the AES algorithm. At the same time, the record data is digitally signed using the private key of the cross-domain authentication server.

[0119] Regularly synchronize the local blockchain distributed record book with other nodes in the network, and at the same time, back up and store the blockchain distributed record book.

[0120] The technical effects of the above technical solution are: through the distributed ledger of the blockchain, each link of the identity authentication process can be easily traced, providing a reliable basis for subsequent audits, problem troubleshooting and responsibility tracing. The AES algorithm is used to encrypt the recorded data to ensure the security and confidentiality of the data before it is written to the blockchain distributed record ledger, preventing data leakage or tampering. The private key of the cross-domain authentication server is used to digitally sign the recorded data, which can verify the legitimacy of the data source and the integrity of the data, ensure the authenticity of the recorded data, and regularly synchronize the blockchain distributed record ledger with other nodes in the network to ensure data consistency and reliability. Even if a node fails or data is lost, it can be restored through other nodes, and the blockchain distributed record ledger is backed up and stored, further enhancing the fault tolerance of the data and ensuring the long-term availability and security of the data.

[0121] After receiving the abnormal warning signal, the secondary verification module immediately starts the secondary identity verification process, specifically:

[0122] The vehicle's built-in fingerprint recognition device requires the driver to collect his fingerprint, which is then encrypted and transmitted to a cross-domain identity information database for comparison to further confirm the driver's identity.

[0123] At the same time, set the corresponding verification time limit and the number of failed attempts limit;

[0124] If the verification fails within the specified time or fails multiple times, it will be determined as an illegal intrusion and the emergency security measures of the security linkage module will be triggered.

[0125] The technical effect of the above technical solution is: when there is an abnormality in the face recognition comparison result, a secondary identity verification is performed through the vehicle's built-in fingerprint recognition device. By combining the two biometric technologies of fingerprint recognition and face recognition, multi-factor identity authentication is achieved, which significantly improves the accuracy and security of identity authentication and effectively prevents illegal intrusions. The verification time limit and the number of failed attempts are set to ensure the efficiency of the secondary verification process, avoid illegal intruders from cracking identity authentication through repeated attempts, and at the same time ensure that legitimate users complete identity verification within a reasonable time, thereby improving user experience. If the verification fails within the specified time or fails multiple times, it will be determined as an illegal intrusion, and the emergency security measures of the security linkage module will be triggered to promptly stop illegal driving and protect the safety of vehicles and personnel.

[0126] The cross-domain authentication unit follows a unified communication protocol, including:

[0127] HTTPS communication protocol based on TLS / SSL encryption protocol, used to establish a communication link between the local intelligent driving system and the cross-domain authentication server;

[0128] And before establishing a connection with the cross-domain authentication server, perform certificate verification to verify the legitimacy of the cross-domain authentication server's identity;

[0129] At the same time, it has the function of data packet encapsulation and decapsulation, encapsulates the facial feature vector, adds communication header information, and when receiving the response from the cross-domain authentication server, decapsulates the data packet, extracts the facial feature vector and performs integrity verification. If the data verification fails, it will re-request the face acquisition unit to transmit.

[0130] The technical effects of the above technical solution are: the HTTPS communication protocol based on the TLS / SSL encryption protocol ensures the security of the communication link between the local intelligent driving system and the cross-domain authentication server, prevents data from being stolen or tampered with during transmission, performs certificate verification before establishing a connection, verifies the legitimacy of the identity of the cross-domain authentication server, prevents man-in-the-middle attacks, and further improves the security of communication. The data packet encapsulation and decapsulation function can encapsulate the facial feature vector and add communication header information to ensure the integrity and consistency of data transmission.

[0131] Face acquisition unit, including:

[0132] An image acquisition module is configured to use an onboard camera to capture the driver's facial image in real time, and is also equipped with an infrared fill light that automatically adjusts the light intensity and angle of the infrared fill light according to the vehicle environment;

[0133] The image processing module is configured to pre-process the face images collected in real time, including image enhancement and normalization processing, wherein:

[0134] Image enhancement: Brightness adjustment is performed by changing the brightness value of the face image, contrast of the face image is increased by using histogram equalization or Laplace filtering, and sharpening of the face image is performed by using the Laplace operator;

[0135] Normalization processing: used to normalize the face image after image enhancement to unify the size and grayscale range of the face image;

[0136] The feature extraction module is configured as a convolutional neural network algorithm based on deep learning and is trained with a large amount of existing facial data to extract facial feature vectors from preprocessed facial images. The extracted facial feature vectors are used for face recognition in the cross-domain authentication unit.

[0137] The technical effects of the above technical solution are as follows: the image acquisition module uses the on-board camera to collect the driver's facial image in real time, and is equipped with an infrared fill light, which can automatically adjust the light intensity and angle according to the vehicle environment, ensuring that clear facial images can be collected under different lighting conditions (such as at night or in a dimly lit environment), thereby improving the quality and usability of facial images. The image processing module uses image enhancement technology to adjust the brightness and contrast of the facial image and perform sharpening operations, effectively improving the image quality, making the image clearer and more vivid, enhancing the recognizability of facial features, and facilitating subsequent feature extraction, thereby improving the accuracy of face recognition. The normalization process unifies the size and grayscale range of the facial image, eliminates the differences in images under different acquisition conditions, improves the consistency and stability of the image, and provides a good foundation for subsequent feature extraction. The feature extraction module is based on a deep learning convolutional neural network algorithm. After training with a large amount of facial data, it can accurately extract distinctive facial feature vectors from the preprocessed image, provide reliable facial feature information for the cross-domain authentication unit, and ensure the accuracy and effectiveness of identity authentication.

[0138] Working principle: The face acquisition unit uses the on-board camera to collect the driver's face image in real time and perform pre-processing operations to extract the driver's face feature vector, providing accurate information for subsequent face authentication. The cross-domain authentication unit compares the extracted face feature vector with the feature vector in the cross-domain identity information database one by one, and judges the legitimacy of the identity based on the similarity threshold, thereby realizing accurate identification of the driver's identity through face recognition technology, effectively preventing illegal driving behavior. If the cross-domain authentication unit finds an abnormality in the comparison result, the emergency processing unit will activate the secondary identity verification mechanism to further confirm the driver's identity, so as to timely discover and respond to potential illegal intrusions, further improving the security and reliability of the system and ensuring driving safety.

[0139] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0140] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. Intelligent driving identity information cross-domain authentication system based on face recognition, characterized by: The system includes a face acquisition unit, a cross-domain authentication unit and an emergency processing unit; The face acquisition unit is configured to acquire the driver's face image in real time through the vehicle-mounted camera and perform preprocessing operations, and extract the face feature vector from the preprocessed face image; The cross-domain authentication unit is configured to receive the facial feature vector of the face collection unit through the cross-domain authentication server, compare the received facial feature vector with the feature vector in the cross-domain identity information database one by one, verify whether the driver's identity is legitimate according to a preset similarity threshold, and record the identity authentication process; Among them, the cross-domain identity information database stores facial feature vectors and the fingerprint features corresponding to the driver; Among them, the cosine similarity algorithm is used to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database. The basic similarity calculation is performed by using the cosine similarity algorithm. When any feature vector is updated, the vector difference before and after the update is retrieved to calculate the vector difference, and then the feature update strength is obtained. The pending similarity between the updated facial feature vector and the database feature vector is obtained, and the pending similarity is compensated and adjusted in combination with the feature update strength; The emergency processing unit is configured to start a secondary identity verification mechanism when there is an abnormality in the comparison result between the face feature vector of the cross-domain authentication unit and the feature vector in the cross-domain identity information database.

2. The cross-domain authentication system for intelligent driving identity information based on face recognition according to claim 1 is characterized in that: The cross-domain authentication unit includes: a local storage module configured to establish a cross-domain identity information database within the cross-domain authentication server, the cross-domain identity information database being used to store driver identity information and corresponding facial feature vectors registered by multiple cross-domain systems, wherein the driver identity information includes driver's license information, vehicle use permission information, and driving preference settings; a feature comparison module configured to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database using a cosine similarity algorithm, and verify whether the driver's identity is legitimate using a preset similarity threshold; If the similarity between the facial feature vector and the feature vector in the cross-domain identity information database is greater than or equal to the similarity threshold, the driver's identity is determined to be legal; If the similarity between the facial feature vector and the feature vector in the cross-domain identity information database is lower than the similarity threshold, the driver's identity is determined to be illegal; The verification record module is configured to build a distributed record book based on blockchain technology to record the identity verification process, including: At the beginning of the authentication process, record the time the request was initiated, the source device ID, and the basic driver information; During the comparison process, the key steps and intermediate results of feature comparison are recorded in real time; After the comparison is completed, the final authentication result, the authentication completion time and cross-domain communication related information are recorded, where the cross-domain communication related information includes the interaction time and data transmission volume with the cross-domain authentication server.

3. The cross-domain authentication system for intelligent driving identity information based on face recognition according to claim 2 is characterized in that: The cosine similarity algorithm is used to calculate the similarity between the received facial feature vector and the feature vector in the cross-domain identity information database, and further includes: When any one of the facial feature vectors or the feature vectors in the cross-domain identity information database is updated, the feature vector before the update and the feature vector after the update are retrieved; Performing difference processing on the feature vector before updating and the feature vector after updating to obtain a vector difference between the feature vector before updating and the feature vector after updating; Obtaining a feature update strength using a vector difference between the feature vector before the update and the feature vector after the update; Retrieving the similarity between the face feature vector obtained after any feature vector is updated and the feature vector in the cross-domain identity information database as the similarity to be processed; Performing compensation adjustment on the to-be-processed similarity using the feature update strength to obtain a similarity value after compensation adjustment; The similarity value after compensation adjustment is used as the final similarity value corresponding to the updated feature vector.

4. The cross-domain authentication system for intelligent driving identity information based on face recognition according to claim 3 is characterized in that: Compensating and adjusting the similarity to be processed by using the feature update strength to obtain a similarity value after compensation and adjustment, including: Retrieving the similarity value before the feature vector is updated and the similarity value to be processed after the feature vector is updated; Performing difference processing on the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated, to obtain the difference between the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated; The similarity change direction parameter is obtained by using the numerical difference between the similarity value before the feature vector is updated and the value of the similarity to be processed after the feature vector is updated, wherein the similarity change direction parameter is sign(S c ); where S c The numerical difference between the similarity value before the feature vector is updated and the similarity value to be processed after the feature vector is updated; Get the L2 norm between the eigenvector before and after the update || v new -v old ||2; where v new and v old Represent the feature vector before and after the update respectively; Using the L2 norm ||v between the feature vector before and after the update new -v old ||2 Obtain feature change gradient Where Δt represents the update time interval; Retrieve feature update strength; Use the features to update the intensity and feature change gradient Combined with the similarity change direction parameter sign(S c ) performing compensation adjustment on the similarity to be processed to obtain a similarity value after compensation adjustment.

5. The cross-domain authentication system for intelligent driving identity information based on face recognition according to claim 1 is characterized in that: The emergency processing unit includes: The anomaly monitoring module is configured to continuously monitor the comparison results transmitted by the cross-domain authentication unit in real time. Once an anomaly is found, it immediately generates an anomaly warning signal and analyzes the anomaly of each comparison result to determine whether it is a difference in facial features, data transmission anomalies, or potential illegal intrusion; A secondary verification module is configured to immediately initiate a secondary identity verification process upon receiving an abnormality warning signal, while recording the time of the abnormality, the abnormality feature vector, and the vehicle location information; The security linkage module is configured to immediately send a signal to lock the vehicle's starting system when there is an abnormality in the comparison result and the secondary verification fails. At the same time, it turns on the vehicle's anti-theft alarm device, emits a high-decibel alarm and flashes the alarm light, and sends the time of the abnormality, facial image, vehicle location information, and identity authentication process and secondary verification process to the cross-domain authentication server, which notifies the vehicle owner and traffic management department.

6. The intelligent driving identity information cross-domain authentication system based on face recognition according to claim 2 is characterized in that: The feature comparison module includes: A threshold adjustment module is configured to dynamically adjust the similarity threshold according to the current authentication scenario, raising the similarity threshold for nighttime or outdoor scenarios and lowering the similarity threshold for daytime or indoor scenarios; On the other hand, historical data is regularly analyzed to calculate the similarity distribution of different driver feature vectors, and the similarity threshold is dynamically adjusted based on the mean and standard deviation of the similarity; The comparison monitoring module is configured to monitor the execution status of the comparison operation in real time during the feature comparison process, including the algorithm running time, memory usage, and comparison progress. If any abnormal delay, memory overflow, or comparison progress stagnation is found in the comparison process, an alarm mechanism will be immediately triggered to notify the cross-domain authentication server to intervene; The report generation module is configured to generate a comparison result report after the comparison is completed, including the similarity value, the comparison time, and any abnormal situation descriptions that occurred during the comparison process. The generated comparison result report will be fed back to the threshold adjustment module as historical data to further optimize the adjustment strategy of the similarity threshold.

7. The intelligent driving identity information cross-domain authentication system based on face recognition according to claim 2 is characterized in that: The verification record module includes: Before the verification record is written into the blockchain distributed record book, the record data is encrypted using the AES algorithm. At the same time, the record data is digitally signed using the private key of the cross-domain authentication server. Regularly synchronize the local blockchain distributed record book with other nodes in the network, and at the same time, back up and store the blockchain distributed record book.

8. The intelligent driving identity information cross-domain authentication system based on face recognition according to claim 5 is characterized in that: After receiving the abnormal warning signal, the secondary verification module immediately starts the secondary identity verification process, specifically: The vehicle's built-in fingerprint recognition device requires the driver to collect his fingerprint, which is then encrypted and transmitted to a cross-domain identity information database for comparison to further confirm the driver's identity. At the same time, set the corresponding verification time limit and the number of failed attempts limit; If the verification fails within the specified time or fails multiple times, it will be determined as an illegal intrusion and the emergency security measures of the security linkage module will be triggered.

9. The cross-domain authentication system for intelligent driving identity information based on face recognition according to claim 1 is characterized in that: The cross-domain authentication unit complies with a unified communication protocol, including: HTTPS communication protocol based on TLS / SSL encryption protocol, used to establish a communication link between the local intelligent driving system and the cross-domain authentication server; And before establishing a connection with the cross-domain authentication server, perform certificate verification to verify the legitimacy of the cross-domain authentication server's identity; At the same time, it has the function of data packet encapsulation and decapsulation, encapsulates the facial feature vector, adds communication header information, and when receiving the response from the cross-domain authentication server, decapsulates the data packet, extracts the facial feature vector and performs integrity verification. If the data verification fails, it will re-request the face acquisition unit to transmit.

10. The intelligent driving identity information cross-domain authentication system based on face recognition according to claim 1 is characterized in that: The face collection unit includes: An image acquisition module is configured to use an onboard camera to capture the driver's facial image in real time, and is also equipped with an infrared fill light that automatically adjusts the light intensity and angle of the infrared fill light according to the vehicle environment; The image processing module is configured to pre-process the face images collected in real time, including image enhancement and normalization processing, wherein: Image enhancement: Brightness adjustment is performed by changing the brightness value of the face image, contrast of the face image is increased by using histogram equalization or Laplace filtering, and sharpening of the face image is performed by using the Laplace operator; Normalization processing: used to normalize the face image after image enhancement to unify the size and grayscale range of the face image; The feature extraction module is configured as a convolutional neural network algorithm based on deep learning and is trained with a large amount of existing facial data to extract facial feature vectors from preprocessed facial images. The extracted facial feature vectors are used for face recognition in the cross-domain authentication unit.

Images