Smart lock security management method and system based on facial recognition

Abstract

The present invention relates to a smart lock security management method and system based on facial recognition, comprising the following steps: performing multi-angle image acquisition on a user's face to obtain an original facial image; performing light compensation and image enhancement processing on the original facial image to obtain a facial feature image group; performing three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence; performing authenticity authentication on the user based on the facial micro-expression change sequence to obtain an authentication result; and controlling the opening or closing operation of the smart lock based on the authentication result; wherein, if the authentication result is successful, an unlocking instruction is sent to the smart lock actuator to complete the unlocking action; if the authentication result fails, a security alarm is triggered and abnormal event information is recorded, thereby solving the technical problem that existing facial recognition smart locks mostly rely on two-dimensional image comparison technology and lack effective analysis of facial depth information and dynamic features.

Description

Technical Field

[0001] The present invention relates to the technical field of smart locks, and in particular to a smart lock security management method and system based on facial recognition. Background Art

[0002] With the rapid development of information technology and artificial intelligence, traditional mechanical locks are no longer able to meet the dual demands of security and convenience in modern society. Smart locks are becoming increasingly important in home and commercial security. Among numerous biometric technologies, facial recognition is widely used in smart lock systems due to its contactless nature, ease of use, and high user acceptance. However, most current facial recognition-based smart locks still face numerous security risks and technical bottlenecks, particularly limited ability to defend against forgery attacks (such as photo and video playback attacks). This severely restricts their widespread adoption in high-security scenarios.

[0003] Existing facial recognition smart locks mostly rely on two-dimensional image comparison technology, lacking effective analysis of facial depth and dynamic features. These systems are susceptible to factors such as lighting changes, posture differences, and masquerading attacks, resulting in high false positive rates. Furthermore, most systems only extract static features during authentication, neglecting the value of dynamic physiological features such as facial micro-expressions, thus reducing the system's anti-spoofing capabilities and recognition accuracy. These technical limitations make some products significantly vulnerable to malicious attacks and can even lead to serious security incidents.

[0004] To improve the security and reliability of facial recognition smart locks, more advanced 3D reconstruction and dynamic feature analysis technologies are urgently needed to enhance the system's ability to accurately identify real users and effectively defend against various counterfeit attacks. Building a robust and anti-interference authentication mechanism by combining multi-angle image acquisition, light compensation, image enhancement, and micro-expression sequence analysis is a key research direction currently awaiting breakthroughs in the intelligent security field. Summary of the Invention

[0005] The main purpose of the present invention is to provide a smart lock security management method based on facial recognition, which solves the technical problem that existing facial recognition smart locks mostly rely on two-dimensional image comparison technology and lack effective analysis of facial depth information and dynamic features.

[0006] To achieve the above objectives, the present invention provides a smart lock security management method based on facial recognition, wherein a face acquisition camera is provided on the smart lock, comprising the following steps:

[0007] The face acquisition camera is used to capture multi-angle images of the user's face to obtain the original facial image;

[0008] Performing light compensation and image enhancement processing on the original facial image to obtain a facial feature image group;

[0009] Performing three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence;

[0010] Performing authenticity authentication on the user based on the facial micro-expression change sequence to obtain an authentication result;

[0011] The opening or closing operation of the smart lock is controlled based on the identity authentication result; if the identity authentication result is successful, an unlocking instruction is sent to the smart lock actuator to complete the unlocking action; if the identity authentication result fails, a security alarm is triggered and abnormal event information is recorded.

[0012] Furthermore, the face acquisition camera is a multi-angle camera array, and the face acquisition camera is used to acquire multi-angle images of the user's face to obtain the original facial image, including:

[0013] Use a multi-angle camera array to synchronously shoot the user's face to obtain multiple facial images from different perspectives;

[0014] Performing time stamping and image splicing processing on the multiple facial images from different perspectives to obtain an initial complete original facial image;

[0015] Image quality assessment and preliminary screening are performed based on the initial complete facial original image, and images with low quality due to motion blur or insufficient lighting are removed to obtain the facial original image.

[0016] Furthermore, the light compensation and image enhancement processing is performed on the original facial image to obtain a facial feature image group, including:

[0017] Performing adaptive histogram equalization on the original facial image to obtain a light-equalized image, and performing non-local mean denoising on the light-equalized image to obtain a denoised facial image;

[0018] Performing illumination non-uniformity correction on the denoised facial image using a multi-scale retinex algorithm to obtain an illumination-corrected image, and performing high-frequency detail enhancement processing based on the illumination-corrected image to obtain a detail-enhanced image;

[0019] Locating facial key points in the detail-enhanced image, determining coordinates of facial feature points based on the facial key points, and geometrically correcting and aligning the facial region in the original facial image based on the facial feature point coordinates to obtain a corrected and aligned facial image;

[0020] The corrected and aligned facial images are subjected to bilateral filtering to obtain edge-preserving smooth images, and adaptive contrast enhancement is performed based on the edge-preserving smooth images to obtain a facial feature image group, wherein the facial feature image group includes a plurality of facial images under different lighting conditions and expression states.

[0021] Furthermore, the three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence includes:

[0022] Performing multi-view stereo matching on the facial feature image group to obtain a facial depth map, and performing point cloud generation and triangular mesh reconstruction based on the facial depth map to obtain a three-dimensional facial model;

[0023] Performing continuous frame motion estimation on the three-dimensional facial model to obtain a facial micro-motion vector field, and performing spatiotemporal filtering processing based on the facial micro-motion vector field to obtain a facial micro-expression feature map;

[0024] Identifying the user's muscle movement pattern based on the facial micro-expression feature map to obtain facial muscle movement parameters, and performing action unit encoding based on the facial muscle movement parameters to obtain a facial expression action coding sequence;

[0025] Based on the facial expression action coding sequence, time series segmentation and cluster analysis are performed to obtain a facial micro-expression change sequence, wherein the facial micro-expression change sequence includes facial micro-expression feature change information within multiple time periods.

[0026] Furthermore, the performing continuous frame motion estimation on the three-dimensional facial model to obtain a facial micro-motion vector field includes:

[0027] Performing pixel brightness gradient calculation on the three-dimensional facial model to obtain a facial pixel gradient matrix, and performing a temporal difference operation on the facial pixel gradient matrix to obtain an inter-frame brightness change map;

[0028] Constructing a motion constraint equation for the facial region based on the inter-frame brightness change map to obtain an optical flow constraint equation group, and performing a variational solution on the optical flow constraint equation group to obtain a pixel displacement vector;

[0029] Performing multi-scale decomposition on the pixel displacement vector using a pyramid layering strategy to obtain a multi-level motion feature map, and performing motion consistency test on the multi-level motion feature map to obtain a motion feature consistency matrix;

[0030] Performing vector field reconstruction on the motion feature consistency matrix to obtain a facial micro-motion vector field.

[0031] Furthermore, the authenticity authentication of the user is performed based on the facial micro-expression change sequence to obtain an authentication result, including:

[0032] Performing dynamic time warping alignment on the facial micro-expression change sequence to obtain an aligned micro-expression sequence, and performing a differential operation on the aligned micro-expression sequence to obtain a micro-expression differential feature, wherein the micro-expression differential feature is a micro-expression change rate at adjacent time points in the aligned micro-expression sequence;

[0033] Performing singular value decomposition on the micro-expression differential features to obtain a micro-expression feature matrix, and performing feature dimensionality reduction based on the micro-expression feature matrix to obtain reduced-dimensional micro-expression features;

[0034] Performing local sensitive hash coding on the dimensionality-reduced micro-expression feature to obtain a micro-expression hash fingerprint, and performing similarity calculation based on the micro-expression hash fingerprint to obtain a similarity matching score;

[0035] An identity threshold is determined based on the similarity matching score to obtain the identity authentication result.

[0036] Furthermore, performing a differential operation on the aligned micro-expression sequence to obtain micro-expression differential features includes:

[0037] Performing time window sliding segmentation on the aligned micro-expression sequence to obtain a plurality of micro-expression sub-sequences, and calculating a local micro-expression intensity curve based on the micro-expression sub-sequences to obtain a micro-expression intensity time series graph;

[0038] Performing multi-scale analysis on the micro-expression intensity time series graph by wavelet decomposition to obtain a micro-expression frequency component graph, and extracting high-frequency components based on the micro-expression frequency component graph to obtain a micro-expression high-frequency feature sequence;

[0039] Applying adaptive threshold segmentation to the micro-expression high-frequency feature sequence to obtain micro-expression peak points, and constructing a micro-expression change topology map based on the micro-expression peak points to obtain a micro-expression topological structure description;

[0040] The gradient change rate between adjacent micro-expression peak points of the user is calculated based on the micro-expression topological structure description to obtain a micro-expression differential feature.

[0041] The present invention also provides a smart lock security management system based on facial recognition, wherein the smart lock is provided with a face acquisition camera, comprising:

[0042] The acquisition module is used to acquire multi-angle images of the user's face through a face acquisition camera to obtain the original facial image;

[0043] an enhancement module, configured to perform light compensation and image enhancement processing on the original facial image to obtain a facial feature image group;

[0044] An extraction module, configured to perform three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence;

[0045] A verification module, configured to verify the authenticity of the user based on the facial micro-expression change sequence and obtain an authentication result;

[0046] A control module is used to control the opening or closing operation of the smart lock based on the identity authentication result; if the identity authentication result is successful, an unlocking instruction is sent to the smart lock actuator to complete the unlocking action; if the identity authentication result fails, a security alarm is triggered and abnormal event information is recorded.

[0047] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0048] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0049] The present invention provides a smart lock security management method based on facial recognition, comprising the following steps: performing multi-angle image acquisition on the user's face to obtain an original facial image; performing light compensation and image enhancement processing on the original facial image to obtain a facial feature image group; performing three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence; performing authenticity authentication on the user based on the facial micro-expression change sequence to obtain an authentication result; and controlling the opening or closing operation of the smart lock based on the authentication result; wherein, if the authentication result is successful, an unlocking instruction is sent to the smart lock actuator to complete the unlocking action; if the authentication result fails, a security alarm is triggered and abnormal event information is recorded. This solves the technical problem that existing facial recognition smart locks rely heavily on two-dimensional image comparison technology and lack effective analysis of facial depth information and dynamic features. This achieves the goal that by extracting and analyzing the dynamic changes of the user's facial micro-expressions, the system can capture more subtle physiological feature changes, further improve the recognition ability of the recognition system, and especially exhibit a stronger anti-interference performance in responding to camouflage attacks. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 1 is a schematic diagram of the steps of a smart lock security management method based on facial recognition in one embodiment of the present invention;

[0051] Figure 2 This is a structural block diagram of a smart lock security management system based on facial recognition in one embodiment of the present invention;

[0052] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0053] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0055] like Figure 1 As shown, Figure 1 This is a smart lock security management method based on facial recognition in one embodiment of the present invention, wherein a face acquisition camera is provided on the smart lock, comprising the following steps:

[0056] Step S1: Use a face acquisition camera to capture multi-angle images of the user's face to obtain an original facial image.

[0057] Specifically, in this method, a facial capture camera captures multi-angle images of the user's face to obtain original facial images. This process is the initial input link of the entire identity verification process. Its purpose is to obtain multi-perspective visual information of the user's face to support subsequent higher-level image processing and feature analysis. In a specific implementation, the facial capture camera is deployed on the front panel of the smart lock and has a certain degree of angle adjustment capability or adopts a wide-angle lens design. This allows it to automatically capture facial images from multiple directions such as the front, side, and oblique angles when the user approaches, thereby forming a set of spatially diverse original facial images. These original images not only contain information about the user's facial contours and facial structure, but also retain surface texture features under lighting conditions, providing basic data support for subsequent light compensation and image enhancement. For example, in an actual application scenario, when a user stands in front of a smart lock and attempts to unlock it, the camera automatically starts and captures the user's facial images from multiple angles, such as the front, left and right sides of the face, to ensure that the collected information fully reflects the user's true facial features and avoid recognition failures due to posture changes or occlusion. The design of this step effectively improves the system's ability to understand facial spatial structure, providing a high-quality data foundation for three-dimensional reconstruction and micro-expression extraction.

[0058] Step S2: performing light compensation and image enhancement processing on the original facial image to obtain a facial feature image group.

[0059] Specifically, the purpose of performing light compensation and image enhancement on the original facial image is to improve image quality under complex lighting conditions, thereby providing a clear and stable set of facial feature images for subsequent 3D reconstruction and micro-expression extraction. In a specific implementation, the system first performs lighting analysis on the received original facial image to identify shadow areas, overexposed areas, or low-contrast areas in the image. It then uses technical means such as adaptive histogram equalization and the Retinex algorithm to perform dynamic range compression and brightness correction on the image, thereby eliminating ambient light interference and enhancing facial detail texture information. On this basis, the system further uses methods such as sharpening filtering and edge enhancement to highlight facial features and structural features, ultimately generating a set of facial feature images with high definition and good lighting consistency. For example, in actual application scenarios, when a user approaches a smart lock at night or in a backlit environment, the original facial image captured by the face acquisition camera may have uneven brightness or local blurring. At this time, performing the above-mentioned light compensation and image enhancement operations can significantly improve the image quality, making key areas such as the eyes, nose, and mouth more clear and distinguishable, thereby ensuring the accuracy and stability of subsequent 3D modeling and micro-expression analysis.

[0060] Step S3: performing three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence.

[0061] Specifically, the three-dimensional reconstruction and micro-expression extraction based on the facial feature image set is intended to model and analyze facial structure and dynamic changes from multiple angles and levels, thereby obtaining a sequence of facial micro-expression changes that can be used to determine identity authenticity. In specific implementations, the system first uses computer vision algorithms to detect and match feature points on the facial feature image set obtained above. In combination with multi-view geometry principles, it constructs a three-dimensional geometric model of the user's face. This model not only contains the spatial positional relationship between facial contours and organs, but also restores facial depth information and surface texture details. On this basis, the system further uses facial action unit recognition technology based on optical flow or deep learning to continuously track and analyze the user's facial muscle movements over a short period of time, extracting sequences of micro-expression changes such as blinking, slight twitching of the corners of the mouth, and frowning, which are difficult to disguise. For example, in actual application scenarios, when a user stands in front of a smart lock and completes image acquisition, the system automatically creates a three-dimensional model of their face and simultaneously analyzes subtle changes in their facial expressions during the verification process, such as slight facial twitches due to nervousness or unnatural muscle movements when trying to imitate others' expressions. This information will be integrated into a sequence of facial micro-expression changes, which serves as an important basis for subsequent identity authenticity verification.

[0062] Step S4: performing authenticity authentication on the user based on the facial micro-expression change sequence to obtain an authentication result.

[0063] Specifically, user authenticity verification based on the facial micro-expression change sequence is performed by analyzing the non-autonomous and difficult-to-imitate facial dynamic features displayed by the user during the identity recognition process to determine whether the user is a real living person, thereby obtaining a reliable identity verification result. In a specific implementation, the system inputs the extracted facial micro-expression change sequence into a pre-trained deep learning model or pattern recognition algorithm. By learning the micro-expression behaviors of a large number of real users and forged attack samples, the model can effectively distinguish between natural human faces and facial changes produced by forgeries such as photos, videos, and masks. In combination with time series analysis technology, the duration, frequency, and spatial distribution of micro-expressions are modeled to determine whether the current user's biometric characteristics conform to the behavioral patterns of real people. For example, in a real application scenario, when an attacker attempting to deceive a smart lock using a photo of someone else stands in front of the device, because the photo cannot produce realistic micro-expression changes, the system will detect the lack of or abnormal facial muscle movement, and thus reject the unlock request and record it as an abnormal event. Conversely, when a legitimate user performs identity verification normally, the system can accurately recognize their unique micro-expression dynamic patterns, completing a highly secure identity confirmation process.

[0064] Step S5, controlling the opening or closing operation of the smart lock based on the identity authentication result; wherein, if the identity authentication result is successful, an unlocking instruction is sent to the smart lock actuator to complete the unlocking action; if the identity authentication result fails, a security alarm is triggered and abnormal event information is recorded.

[0065] Specifically, the control of the opening or closing operation of the smart lock based on the identity authentication result is to use the output of the aforementioned identity authentication link as the core basis for the system to perform physical actions, thereby realizing the automation and intelligent security management of the access control system. In the specific implementation, when the system completes the analysis of the facial micro-expression change sequence and obtains the identity authentication result, it will immediately send the corresponding control instruction to the actuator inside the smart lock according to the result. If the identity authentication result is successful, the system will send an unlocking instruction to the motor drive unit through the communication module to drive the mechanical structure inside the lock body to complete the unlocking action; on the contrary, if the identity authentication fails, the system will not only prevent the unlocking operation, but also trigger the preset security alarm mechanism, such as starting a buzzer, flashing light or pushing an abnormal notification to the user terminal, and at the same time record the time, image data and device status information of this verification failure to a local or cloud database for subsequent tracing and risk analysis. For example, in actual application scenarios, when a legitimate user stands in front of a smart lock for identity recognition, the system quickly completes the unlocking operation after confirming that the user is a real user; when an attacker tries to impersonate someone else using a photo or mask, the system recognizes that the sequence of changes in their facial micro-expressions is abnormal, determines that the identity authentication has failed, and then refuses to unlock the lock and triggers an alarm, effectively preventing illegal intrusions.

[0066] In a specific embodiment, the face acquisition camera is a multi-angle camera array, and the multi-angle image acquisition of the user's face by the face acquisition camera to obtain the original facial image includes:

[0067] Use a multi-angle camera array to synchronously shoot the user's face to obtain multiple facial images from different perspectives;

[0068] Performing time stamping and image splicing processing on the multiple facial images from different perspectives to obtain an initial complete original facial image;

[0069] Image quality assessment and preliminary screening are performed based on the initial complete facial original image, and images with low quality due to motion blur or insufficient lighting are removed to obtain the facial original image.

[0070] Specifically, the face acquisition camera is a multi-angle camera array, which is designed to synchronously capture user facial images from different spatial angles through multiple cameras, thereby improving the accuracy and completeness of subsequent three-dimensional reconstruction and identity verification. In a specific implementation, the process of acquiring multi-angle images of the user's face through the face acquisition camera to obtain the original facial image includes: first, using the multi-angle camera array to simultaneously shoot the user's face from multiple directions such as the front, left and right sides, and upper and lower angles when the user is close to the smart lock and in the recognition area, so as to obtain multiple facial images with spatial differences; then, the system will add a unified timestamp to these images and fuse them into an initial complete facial original image through an image stitching algorithm to construct more comprehensive facial structure information; on this basis, the system will further perform image quality assessment on the initial complete facial original image to detect whether there are problems such as motion blur, uneven lighting, and images that are too dark or too bright, and preliminarily screen the images according to preset quality scoring standards, eliminate low-quality images that do not meet the requirements, and finally retain high-quality facial original images that can be used for subsequent processing. For example, in actual application scenarios, when a user stands in front of a smart lock and prepares to unlock it, the multi-angle camera array will automatically start and synchronously capture their facial images from different directions, ensuring that even if the user's head is slightly tilted or there is a slight occlusion, complete and clear facial information can be obtained, avoiding recognition failures due to single-perspective imaging defects, thereby significantly improving the system's robustness and user experience.

[0071] In a specific embodiment, performing light compensation and image enhancement processing on the original facial image to obtain a facial feature image group includes:

[0072] Performing adaptive histogram equalization on the original facial image to obtain a light-equalized image, and performing non-local mean denoising on the light-equalized image to obtain a denoised facial image;

[0073] Performing illumination non-uniformity correction on the denoised facial image using a multi-scale retinex algorithm to obtain an illumination-corrected image, and performing high-frequency detail enhancement processing based on the illumination-corrected image to obtain a detail-enhanced image;

[0074] Locating facial key points in the detail-enhanced image, determining coordinates of facial feature points based on the facial key points, and geometrically correcting and aligning the facial region in the original facial image based on the facial feature point coordinates to obtain a corrected and aligned facial image;

[0075] The corrected and aligned facial images are subjected to bilateral filtering to obtain edge-preserving smooth images, and adaptive contrast enhancement is performed based on the edge-preserving smooth images to obtain a facial feature image group, wherein the facial feature image group includes a plurality of facial images under different lighting conditions and expression states.

[0076] Specifically, performing light compensation and image enhancement on the original facial image to obtain a facial feature image group is a critical preprocessing step in the entire identity verification process. Its purpose is to improve the quality of the original image through a series of image processing techniques, so that it has higher accuracy and stability in subsequent high-level analysis tasks such as three-dimensional reconstruction and micro-expression extraction. In the specific implementation process, this step first performs adaptive histogram equalization (CLAHE) processing on the original facial image to improve the overall brightness distribution of the image, making areas that originally appeared dim or too bright due to uneven lighting more clear and discernible, thereby obtaining a light-balanced image; then, on this basis, a non-local mean denoising algorithm is further adopted to effectively suppress random noise interference in the image, retain more real facial texture information, and ultimately obtain a denoised facial image. Next, the system uses a multi-scale Retinex algorithm to correct for illumination unevenness in the denoised facial images. This process isolates and normalizes the illumination components within the image, eliminating the effects of undesirable factors such as shadows and reflections, resulting in a corrected image. Furthermore, the system performs high-frequency detail enhancement on the corrected image to highlight key structural information such as facial contours and facial features, resulting in a detail-enhanced image. This provides richer visual information for subsequent feature extraction. To ensure spatial consistency between images, the system further locates facial key points in the detail-enhanced image, such as the eyes, nose tip, and mouth corners, which are landmarks with stable positions and shapes. The coordinates of these key points are then determined based on these key points. Subsequently, geometric transformations such as affine transformations or thin-plate spline interpolation are used to geometrically correct and align the facial regions in the original facial images, ensuring that all images maintain consistent pose and proportions within a unified spatial coordinate system. This results in a corrected and aligned facial image. This step is crucial for improving 3D reconstruction accuracy and extracting sequences of micro-expression changes, especially when the user's face exhibits slight deflections or expression changes, effectively reducing recognition errors caused by pose differences. Subsequently, the system performs bilateral filtering on the corrected and aligned facial images. This technology can better preserve edge information while smoothing the image, avoiding the problem of blurred details caused by traditional filtering methods, thereby obtaining a smooth image with preserved edges. Finally, the system performs adaptive contrast enhancement based on the smoothed image, dynamically adjusting the grayscale distribution range according to the image content, so that the details of the dark and bright parts of the image are clearer, and finally generates a facial feature image group, which contains multiple high-quality facial images under different lighting conditions and expressions, providing a solid data foundation for subsequent identity authenticity verification. For example, in actual application scenarios, when a user approaches a smart lock in a backlit environment in the evening, the original facial image captured by the face acquisition camera may have problems of local overexposure and shadow occlusion, resulting in a decrease in image quality.At this point, the system sequentially executes the aforementioned light compensation and image enhancement processes: First, it uses adaptive histogram equalization and a multi-scale Retinex algorithm to mitigate uneven illumination. Then, it combines non-local means denoising and bilateral filtering to reduce noise interference while enhancing facial detail and edge clarity. Finally, through precise positioning and geometric alignment of key points, images captured at different angles are unified into a standard pose, ultimately generating a set of high-quality facial feature images. These images are not only visually clearer but also highly semantically consistent, significantly improving the accuracy of 3D modeling and micro-expression analysis, thereby ensuring the security and reliability of identity verification.

[0077] In a specific embodiment, the three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence includes:

[0078] Performing multi-view stereo matching on the facial feature image group to obtain a facial depth map, and performing point cloud generation and triangular mesh reconstruction based on the facial depth map to obtain a three-dimensional facial model;

[0079] Performing continuous frame motion estimation on the three-dimensional facial model to obtain a facial micro-motion vector field, and performing spatiotemporal filtering processing based on the facial micro-motion vector field to obtain a facial micro-expression feature map;

[0080] Identifying the user's muscle movement pattern based on the facial micro-expression feature map to obtain facial muscle movement parameters, and performing action unit encoding based on the facial muscle movement parameters to obtain a facial expression action coding sequence;

[0081] Based on the facial expression action coding sequence, time series segmentation and cluster analysis are performed to obtain a facial micro-expression change sequence, wherein the facial micro-expression change sequence includes facial micro-expression feature change information within multiple time periods.

[0082] Specifically, the three-dimensional reconstruction and micro-expression extraction based on the facial feature image set to obtain a sequence of facial micro-expression changes is the core technical component of achieving high-security identification throughout the entire identity verification process. This step aims to construct the three-dimensional structure of the user's face using the high-quality facial feature image set. Based on this, the dynamic changes in micro-expressions, which are difficult to manually control in a short period of time, are analyzed to obtain physiological behavioral characteristics that can be used for authenticity judgment. In its implementation, the system first performs multi-view stereo matching on the facial feature image set. This involves calculating the spatial correspondence between images taken from different angles through feature point matching and disparity estimation algorithms, thereby generating a corresponding facial depth map. Subsequently, a dense point cloud is generated based on this depth map, and a triangulated mesh reconstruction method is used to construct a continuous, smooth three-dimensional facial model from these discrete points. This model not only contains facial contour information but also reflects the spatial distribution of skin texture and organ morphology. Next, to capture dynamic facial changes during the identity verification process, the system performs continuous frame motion estimation on the three-dimensional facial model. This involves tracking the position changes of facial vertices in adjacent time frames to construct a micro-motion vector field for each region of the facial surface, which describes the local motion trends of facial muscles. Based on this, the system further performs spatiotemporal filtering on the micro-motion vector field to remove abnormal fluctuations caused by environmental interference or device noise, retaining physiologically meaningful micro-expression changes and ultimately obtaining a stable facial micro-expression feature map. This processing provides accurate motion parameters for subsequent action unit recognition. The system then identifies the user's facial muscle movement patterns based on this facial micro-expression feature map. Specifically, the system analyzes the displacement and deformation direction of key facial regions (such as the orbicularis oculi and orbicularis oris muscles) to derive corresponding facial muscle movement parameters. Incorporating action unit theory (AU) from psychology, these movement parameters are mapped to a standardized expression coding system to generate a facial expression action code sequence. This code sequence digitally represents the user's specific facial movement combination, such as blinking, frowning, and raising the corners of the mouth, and is highly semantically interpretable. Finally, to characterize the temporal patterns of a user's facial microexpressions, the system performs temporal segmentation and clustering analysis based on the facial expression action code sequence. This involves dividing the continuous action code into segments according to time windows and using a clustering algorithm (such as K-means or DBSCAN) to classify microexpression patterns within similar time periods. This identifies multiple representative microexpression change states and their durations, ultimately forming a facial microexpression change sequence. This sequence records the changes in the user's facial microexpression characteristics at different times during the identity verification process, forming an important basis for identity verification.For example, in a real-world application scenario, when a user stands in front of a smart lock to complete identity recognition, the system reconstructs a three-dimensional face model based on their facial feature image set and tracks subtle facial changes during the verification process in real time. If the user blinks slightly more frequently or their mouth corners twitch involuntarily due to nervousness, the system can accurately identify and incorporate this into the sequence of facial micro-expression changes. Conversely, if an attacker attempts to deceive the system using photos or videos, the system will not be able to detect effective micro-expression changes due to the lack of real muscle movement, and will therefore determine it as a non-live attack, effectively improving the security level of identity verification.

[0083] In a specific embodiment, performing continuous frame motion estimation on the three-dimensional facial model to obtain a facial micro-motion vector field includes:

[0084] Performing pixel brightness gradient calculation on the three-dimensional facial model to obtain a facial pixel gradient matrix, and performing a temporal difference operation on the facial pixel gradient matrix to obtain an inter-frame brightness change map;

[0085] Constructing a motion constraint equation for the facial region based on the inter-frame brightness change map to obtain an optical flow constraint equation group, and performing a variational solution on the optical flow constraint equation group to obtain a pixel displacement vector;

[0086] Performing multi-scale decomposition on the pixel displacement vector using a pyramid layering strategy to obtain a multi-level motion feature map, and performing motion consistency test on the multi-level motion feature map to obtain a motion feature consistency matrix;

[0087] Performing vector field reconstruction on the motion feature consistency matrix to obtain a facial micro-motion vector field.

[0088] Specifically, in the process of performing continuous frame motion estimation on the 3D facial model and obtaining the facial micro-motion vector field, the pixel brightness gradients of the 3D facial model must first be calculated. This is to capture detailed changes in facial components. By analyzing the brightness variations around each pixel, a facial pixel gradient matrix that reflects facial structural features is generated. Next, to capture the dynamic characteristics of facial changes over time, a temporal difference operation is performed on this facial pixel gradient matrix, comparing the brightness differences between adjacent frames to generate an inter-frame brightness change map. This step is crucial for accurately identifying subtle brightness changes caused by facial muscle movement, as even the slightest changes in expression can cause localized brightness changes, which are the basis for inferring facial motion patterns in subsequent steps. Based on the resulting inter-frame brightness change map, a set of optical flow constraint equations is constructed and solved variationally to obtain pixel displacement vectors. This involves imposing specific motion constraints on the facial region and using the principles of optical flow to estimate the velocity vector, or pixel displacement vector, for each point in the image sequence. Specifically, the optical flow constraint equations, established based on optical flow theory, describe the relationship between the positional changes of the same object between two consecutive frames and the assumption of brightness invariance. By applying variational methods to this set of equations, we can effectively estimate the movement direction and distance of facial points between adjacent frames, providing key data for understanding facial dynamics. To more accurately analyze complex facial motion patterns, a pyramid-based hierarchical strategy is used to perform a multi-scale decomposition of the pixel displacement vectors obtained above, generating a multi-level motion feature map. This strategy allows the algorithm to gradually approximate the true facial motion state from a coarse-to-fine approach, first processing motion trends at a large scale and low resolution, and then gradually refining the motion details at a small scale and high resolution. This not only improves computational efficiency but also effectively avoids local extrema that may occur directly at high resolution. Subsequently, a motion consistency check is performed on the multi-level motion feature map to ensure that the extracted motion features are physically coherent and reasonable, forming a reliable motion feature consistency matrix. This step is crucial for eliminating false motion signals caused by noise or non-facial elements, thereby improving the accuracy of the final results. Finally, the motion feature consistency matrix is ​​used as input for vector field reconstruction to generate the final facial micro-motion vector field. This process involves integrating all verified motion features and converting them into a vector field form that can be intuitively represented in three-dimensional space. The facial micro-motion vector field is not just a collection of independent displacement vectors, but rather a comprehensive depiction of the complex motion patterns of the entire facial surface that evolve over time. For example, in the application scenario of a smart door lock, when a user attempts to unlock the door, the system collects a 3D model of their face in real time and uses the above technology to track subtle facial movements.If the user's real expression includes movements such as blinking or smiling, the system can accurately identify the changes in these micro-expressions by analyzing the facial micro-motion vector field. Even very brief or subtle movements can be captured. This is important for distinguishing real users from attackers who try to deceive the system using photos or videos, because the latter cannot simulate the natural and unique micro-expression dynamics of real faces.

[0089] In a specific embodiment, performing authenticity authentication on the user based on the facial micro-expression change sequence to obtain an authentication result includes:

[0090] Performing dynamic time warping alignment on the facial micro-expression change sequence to obtain an aligned micro-expression sequence, and performing a differential operation on the aligned micro-expression sequence to obtain a micro-expression differential feature, wherein the micro-expression differential feature is a micro-expression change rate at adjacent time points in the aligned micro-expression sequence;

[0091] Performing singular value decomposition on the micro-expression differential features to obtain a micro-expression feature matrix, and performing feature dimensionality reduction based on the micro-expression feature matrix to obtain reduced-dimensional micro-expression features;

[0092] Performing local sensitive hash coding on the dimensionality-reduced micro-expression feature to obtain a micro-expression hash fingerprint, and performing similarity calculation based on the micro-expression hash fingerprint to obtain a similarity matching score;

[0093] An identity threshold is determined based on the similarity matching score to obtain the identity authentication result.

[0094] Specifically, verifying the user's authenticity based on the sequence of facial micro-expression changes and obtaining an authentication result is a key step in achieving high-security identity recognition for the entire smart lock system. The core of this step is to determine whether the user is a real living person and has a legitimate identity by modeling and comparing the dynamic changes in micro-expressions on the user's face over a short period of time, which are difficult to manually control. In specific implementation, the sequence of facial micro-expression changes is first subjected to dynamic time warping (DTW) alignment. The purpose is to unify micro-expression changes of different time lengths and rhythms onto a standard timeline, eliminating timing differences caused by the varying speeds of user movements, and thus obtaining a structurally consistent aligned micro-expression sequence. Next, the system performs a differential operation on this aligned micro-expression sequence to calculate the micro-expression change rate between adjacent time points, namely the micro-expression differential feature. This step can highlight the acceleration characteristics of facial muscle movement and help capture more subtle and difficult-to-imitate physiological behavior patterns. To further extract discriminative feature expressions, the system performs singular value decomposition (SVD) on the micro-expression differential features. Using matrix decomposition techniques, it extracts the main change directions and energy distribution, forming a compact yet information-rich micro-expression feature matrix. Furthermore, the system performs feature dimensionality reduction on the micro-expression feature matrix, removing redundant dimensions and retaining the most representative low-dimensional feature vectors to generate reduced-dimensional micro-expression features. This not only improves subsequent matching efficiency but also enhances the model's robustness to noise and abnormal data. Next, the system performs locality-sensitive hashing (LSH) encoding on the reduced-dimensional micro-expression features, mapping them into a set of micro-expression hash fingerprints with proximity-preserving properties. This encoding method significantly reduces data storage and computational complexity while ensuring similar hash values ​​for similar samples. The system then calculates similarity between these micro-expression hash fingerprints and pre-stored legitimate user templates, using methods such as cosine similarity or Euclidean distance to assess the degree of match between the current user's micro-expression behavior and the target identity, ultimately generating a quantitative similarity match score. Finally, the system performs an identity threshold determination based on the similarity match score, setting a reasonable security threshold as the basis for identity verification decisions: if the similarity match score exceeds the threshold, the identity verification is considered successful and unlocking is allowed; otherwise, it is considered an illegal access attempt, triggering an alarm mechanism and recording the abnormal event. This process completes a closed-loop process from the original micro-expression change sequence to the final identity authentication result, ensuring that the system has a high degree of discrimination against forgery attacks. For example, in a real application scenario, when a legitimate user stands in front of a smart lock for identity verification, the system automatically captures their facial image and, through the aforementioned stages of processing, generates a corresponding sequence of facial micro-expression changes. Because the micro-expression changes of real users have natural temporal continuity and specific muscle movement patterns, the system will obtain a higher score when performing similarity matching, successfully passing identity verification and completing unlocking.However, if an attacker attempts to deceive the system using photos or videos, due to the lack of real micro-expression dynamic changes, the micro-expression change sequence generated by the attacker cannot match the template of the legitimate user, and the similarity score is far below the set threshold. The system will then reject the unlocking request and activate the alarm mechanism, effectively preventing illegal intrusion.

[0095] In a specific embodiment, performing a differential operation on the aligned micro-expression sequence to obtain a micro-expression differential feature includes:

[0096] Performing time window sliding segmentation on the aligned micro-expression sequence to obtain a plurality of micro-expression sub-sequences, and calculating a local micro-expression intensity curve based on the micro-expression sub-sequences to obtain a micro-expression intensity time series graph;

[0097] Performing multi-scale analysis on the micro-expression intensity time series graph by wavelet decomposition to obtain a micro-expression frequency component graph, and extracting high-frequency components based on the micro-expression frequency component graph to obtain a micro-expression high-frequency feature sequence;

[0098] Applying adaptive threshold segmentation to the micro-expression high-frequency feature sequence to obtain micro-expression peak points, and constructing a micro-expression change topology map based on the micro-expression peak points to obtain a micro-expression topological structure description;

[0099] The gradient change rate between adjacent micro-expression peak points of the user is calculated based on the micro-expression topological structure description to obtain a micro-expression differential feature.

[0100] Specifically, performing differential operations on the aligned micro-expression sequences to obtain micro-expression differential features is an important technical step in the identity authentication process for characterizing the dynamic changes in the user's facial micro-expressions. The core goal of this process is to extract micro-expression behavior features with high discriminative capabilities through in-depth analysis in the time dimension, thereby providing an accurate basis for subsequent identity authenticity judgment. In a specific implementation, the aligned micro-expression sequences are first subjected to a time window sliding segmentation process, that is, continuous micro-expression changes are divided into multiple overlapping time segments according to a fixed or adaptive length to form multiple micro-expression sub-sequences. This can more carefully capture the local change trends of facial micro-movements in different time periods; then, based on each micro-expression sub-sequence, the system calculates its corresponding local micro-expression intensity curve, which reflects the amplitude of image brightness changes caused by facial muscle movement in a short period of time, and finally generates a micro-expression intensity time series diagram to quantitatively describe the activity level of the user's facial micro-expressions. To further explore the frequency information in micro-expression changes, the system performs multi-scale analysis on the micro-expression intensity time series using wavelet decomposition. This process decomposes the original signal into multiple frequency components, allowing the low-frequency component to reflect the overall trend while the high-frequency component retains rapidly changing details. The system then focuses on extracting the high-frequency components, which often correspond to transient movements of the user's facial muscles, such as blinking and mouth twitching, which are difficult to control. This results in a high-frequency feature sequence of micro-expressions. This step is crucial for identifying forgery attacks, as photos or videos often cannot reproduce these natural micro-movements of real faces. Next, to accurately extract key change points from the high-frequency features, the system applies an adaptive threshold segmentation algorithm to the micro-expression high-frequency feature sequence, automatically identifying micro-expression peaks exceeding a set intensity threshold. These peaks represent moments when the user's facial micro-movements occur significantly. The system then constructs a micro-expression change topology map based on these peaks. This topology map graphically displays the spatiotemporal correlations and structural relationships between micro-expression events, forming a topological description of the user's micro-expression dynamic patterns and helping to reveal their unique facial behavioral characteristics. Finally, based on the micro-expression topological structure description, the system further calculates the gradient change rate between adjacent micro-expression peak points of the user, that is, the speed of change in the intensity of facial micro-expressions per unit time, and outputs this as the micro-expression differential feature. This feature not only reflects the speed of the user's facial movements, but also reflects the coherence and naturalness of their movements, making it an important indicator for determining identity authenticity.For example, in real-world applications, when a legitimate user stands in front of a smart lock for identity verification, their facial micro-expressions exhibit natural rhythms and complex topological structures. When calculating the gradient rate of change, the system generates a set of micro-expression differential features with rich dynamic characteristics, resulting in a high-quality match with the pre-stored template. However, if an attacker attempts to deceive the system using a photo or video, their facial micro-expressions lack realistic dynamic changes, resulting in sparse or even missing micro-expression peaks and an abnormal gradient rate of change, ultimately failing identity verification. This identity verification mechanism based on micro-expression differential features significantly improves the system's security and effectively prevents forgery attacks.

[0101] The above describes the smart lock security management method based on facial recognition in the embodiment of the present invention. The following describes the smart lock security management system based on facial recognition in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a smart lock security management system based on facial recognition includes:

[0102] The acquisition module 21 is used to acquire multi-angle images of the user's face through a face acquisition camera to obtain an original facial image;

[0103] An enhancement module 22 is configured to perform light compensation and image enhancement processing on the original facial image to obtain a facial feature image group;

[0104] An extraction module 23 is used to perform three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence;

[0105] A verification module 24 is used to verify the authenticity of the user based on the facial micro-expression change sequence and obtain an authentication result;

[0106] The control module 25 is used to control the opening or closing operation of the smart lock based on the identity authentication result; if the identity authentication result is successful, an unlocking instruction is sent to the smart lock actuator to complete the unlocking action; if the identity authentication result fails, a security alarm is triggered and abnormal event information is recorded.

[0107] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0108] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided, wherein the internal structure of the computer device can be as follows: Figure 3As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0109] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0110] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0111] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media provided herein and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0112] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0113] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A smart lock security management method based on facial recognition, characterized in that: The smart lock is provided with a face collection camera, including the following steps: The face acquisition camera is used to capture multi-angle images of the user's face to obtain the original facial image; Performing light compensation and image enhancement processing on the original facial image to obtain a facial feature image group; Performing three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence; Performing authenticity authentication on the user based on the facial micro-expression change sequence to obtain an authentication result; Controlling the opening or closing operation of the smart lock based on the identity verification result; wherein, if the identity verification result is successful, sending an unlocking instruction to the smart lock actuator to complete the unlocking action; if the identity verification result fails, triggering a security alarm and recording abnormal event information; The three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence includes: Performing multi-view stereo matching on the facial feature image group to obtain a facial depth map, and performing point cloud generation and triangular mesh reconstruction based on the facial depth map to obtain a three-dimensional facial model; Performing continuous frame motion estimation on the three-dimensional facial model to obtain a facial micro-motion vector field, and performing spatiotemporal filtering processing based on the facial micro-motion vector field to obtain a facial micro-expression feature map; Identifying the user's muscle movement pattern based on the facial micro-expression feature map to obtain facial muscle movement parameters, and performing action unit encoding based on the facial muscle movement parameters to obtain a facial expression action coding sequence; Based on the facial expression action coding sequence, time series segmentation and cluster analysis are performed to obtain a facial micro-expression change sequence, wherein the facial micro-expression change sequence includes facial micro-expression feature change information within multiple time periods.

2. The smart lock security management method based on facial recognition according to claim 1 is characterized in that: The face acquisition camera is a multi-angle camera array, and the face acquisition camera is used to acquire multi-angle images of the user's face to obtain an original facial image, including: Use a multi-angle camera array to synchronously shoot the user's face to obtain multiple facial images from different perspectives; Performing time stamping and image splicing processing on the multiple facial images from different perspectives to obtain an initial complete original facial image; Image quality assessment and preliminary screening are performed based on the initial complete facial original image, and images with low quality due to motion blur or insufficient lighting are removed to obtain the facial original image.

3. The smart lock security management method based on facial recognition according to claim 1 is characterized in that: The step of performing light compensation and image enhancement processing on the original facial image to obtain a facial feature image group includes: performing adaptive histogram equalization on the original facial image to obtain a light-equalized image, and performing non-local mean denoising on the light-equalized image to obtain a denoised facial image; Performing illumination non-uniformity correction on the denoised facial image using a multi-scale retinex algorithm to obtain an illumination-corrected image, and performing high-frequency detail enhancement processing based on the illumination-corrected image to obtain a detail-enhanced image; Locating facial key points in the detail-enhanced image, determining coordinates of facial feature points based on the facial key points, and geometrically correcting and aligning the facial region in the original facial image based on the facial feature point coordinates to obtain a corrected and aligned facial image; The corrected and aligned facial images are subjected to bilateral filtering to obtain edge-preserving smooth images, and adaptive contrast enhancement is performed based on the edge-preserving smooth images to obtain a facial feature image group, wherein the facial feature image group includes a plurality of facial images under different lighting conditions and expression states.

4. The smart lock security management method based on facial recognition according to claim 1 is characterized in that: The step of performing continuous frame motion estimation on the three-dimensional facial model to obtain a facial micro-motion vector field includes: Performing pixel brightness gradient calculation on the three-dimensional facial model to obtain a facial pixel gradient matrix, and performing a temporal difference operation on the facial pixel gradient matrix to obtain an inter-frame brightness change map; Constructing a motion constraint equation for the facial region based on the inter-frame brightness change map to obtain an optical flow constraint equation group, and performing a variational solution on the optical flow constraint equation group to obtain a pixel displacement vector; Performing multi-scale decomposition on the pixel displacement vector using a pyramid layering strategy to obtain a multi-level motion feature map, and performing motion consistency test on the multi-level motion feature map to obtain a motion feature consistency matrix; Performing vector field reconstruction on the motion feature consistency matrix to obtain a facial micro-motion vector field.

5. The smart lock security management method based on facial recognition according to claim 1 is characterized in that: The performing of authenticity authentication on the user based on the facial micro-expression change sequence to obtain an authentication result includes: Performing dynamic time warping alignment on the facial micro-expression change sequence to obtain an aligned micro-expression sequence, and performing a differential operation on the aligned micro-expression sequence to obtain a micro-expression differential feature, wherein the micro-expression differential feature is a micro-expression change rate at adjacent time points in the aligned micro-expression sequence; Performing singular value decomposition on the micro-expression differential features to obtain a micro-expression feature matrix, and performing feature dimensionality reduction based on the micro-expression feature matrix to obtain reduced-dimensional micro-expression features; Performing local sensitive hash coding on the dimensionality-reduced micro-expression feature to obtain a micro-expression hash fingerprint, and performing similarity calculation based on the micro-expression hash fingerprint to obtain a similarity matching score; An identity threshold is determined based on the similarity matching score to obtain the identity authentication result.

6. The smart lock security management method based on facial recognition according to claim 5 is characterized in that: The performing a differential operation on the aligned micro-expression sequence to obtain micro-expression differential features includes: Performing time window sliding segmentation on the aligned micro-expression sequence to obtain a plurality of micro-expression sub-sequences, and calculating a local micro-expression intensity curve based on the micro-expression sub-sequences to obtain a micro-expression intensity time series graph; Performing multi-scale analysis on the micro-expression intensity time series graph by wavelet decomposition to obtain a micro-expression frequency component graph, and extracting high-frequency components based on the micro-expression frequency component graph to obtain a micro-expression high-frequency feature sequence; Applying adaptive threshold segmentation to the micro-expression high-frequency feature sequence to obtain micro-expression peak points, and constructing a micro-expression change topology map based on the micro-expression peak points to obtain a micro-expression topological structure description; The gradient change rate between adjacent micro-expression peak points of the user is calculated based on the micro-expression topological structure description to obtain a micro-expression differential feature.

7. A smart lock security management system based on facial recognition, characterized in that: The smart lock is provided with a face acquisition camera, including: The acquisition module is used to acquire multi-angle images of the user's face through a face acquisition camera to obtain the original facial image; an enhancement module, configured to perform light compensation and image enhancement processing on the original facial image to obtain a facial feature image group; An extraction module, configured to perform three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence; A verification module, configured to verify the authenticity of the user based on the facial micro-expression change sequence and obtain an authentication result; A control module for controlling the opening or closing operation of the smart lock based on the identity verification result; wherein, if the identity verification result is successful, an unlocking instruction is sent to the smart lock actuator to complete the unlocking action; if the identity verification result fails, a security alarm is triggered and abnormal event information is recorded; The three-dimensional reconstruction and micro-expression extraction based on the facial feature image group to obtain a facial micro-expression change sequence includes: Performing multi-view stereo matching on the facial feature image group to obtain a facial depth map, and performing point cloud generation and triangular mesh reconstruction based on the facial depth map to obtain a three-dimensional facial model; Performing continuous frame motion estimation on the three-dimensional facial model to obtain a facial micro-motion vector field, and performing spatiotemporal filtering processing based on the facial micro-motion vector field to obtain a facial micro-expression feature map; Identifying the user's muscle movement pattern based on the facial micro-expression feature map to obtain facial muscle movement parameters, and performing action unit encoding based on the facial muscle movement parameters to obtain a facial expression action coding sequence; Based on the facial expression action coding sequence, time series segmentation and cluster analysis are performed to obtain a facial micro-expression change sequence, wherein the facial micro-expression change sequence includes facial micro-expression feature change information within multiple time periods.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images