Intelligent lock unlocking method and system based on face image processing

Abstract

The invention relates to the technical field of identity verification, in particular to an intelligent lock unlocking method and system based on face image processing, and the method comprises the following steps: obtaining face images of multiple angles through an image collection device, extracting face key features of a user, including a face contour, face symmetry and positions of five sense organs, and generating a face feature data set; according to the method, the limitation of traditional two-dimensional plane recognition is broken through through acquisition of multi-angle face images, construction of a multi-dimensional feature data set, realization of feature fusion by utilizing difference evaluation and weight distribution, enhancement of the stability and the anti-interference capability of biological feature information and extraction of three-dimensional geometric information and depth data; and in combination with a dynamic illumination threshold monitoring mechanism, the influence of environment variables on recognition precision is suppressed, and the resistance to plane attacks is improved, so that the security level and scene adaptability are both improved while the convenience of the identity authentication system is kept.

Description

Technical Field

[0001] The present invention relates to the technical field of identity authentication, and particularly to an intelligent lock unlocking method and system based on face image processing. Background Art

[0002] The technical field of identity authentication focuses on using biometric features, behavioral features, and various identity authentication means to confirm the identity of users. The core content involves collecting and analyzing multiple biometric features such as images, voices, fingerprints, or irises of users to accurately verify the identity of users. With the increasing demand for information security, identity authentication technology is gradually developing towards a more efficient, secure, and convenient direction. By using high-precision image processing, pattern recognition, and signal processing, it is possible to verify and authorize personal identities without the need to carry any physical media, and it is widely used in security systems, access control management, financial payment, and public service fields, aiming to enhance security and facilitate user operations.

[0003] Among them, the intelligent lock unlocking method based on face image processing refers to unlocking control of an intelligent lock by acquiring and recognizing a user's face image. The method covers using face recognition technology for user identity authentication. Specifically, it involves collecting a user's face image and comparing it with the pre-stored face features in a database to confirm the identity. During the collection, processing, and comparison of face images, various technical means such as image preprocessing, feature extraction, and comparison algorithms are used to extract facial feature information and compare it with the user's facial information in the database to confirm the identity. The process relies on image processing algorithms and pattern recognition technology to ensure the accuracy and security of the unlocking process.

[0004] Traditional intelligent lock unlocking methods have deficiencies in multiple aspects. The acquisition of single-perspective two-dimensional images is easily restricted by pose changes and lighting conditions. When there is an angular deviation between the user's face and the acquisition device or in a backlight environment, the extraction of key features may fail. The static biometric database lacks the ability to adapt to dynamic environments and cannot respond to changes in light intensity in real time, resulting in a sharp drop in the recognition rate in low-light or strong-light environments. The two-dimensional planar feature comparison has anti-counterfeiting loopholes and it is difficult to distinguish a real face from a high-precision imitation. The traditional technology lacks an evaluation mechanism for the stability of feature data. When a user's local facial features change due to trauma, legitimate users are wrongly rejected. When dealing with complex environmental variables, it relies on fixed thresholds and lacks a dynamic adjustment mechanism, leading to a decline in user experience and an accumulation of security risks. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art, embodiments of the present invention provide an intelligent lock unlocking method and system based on face image processing. The technical solutions are as follows:

[0006] To achieve the above object, the present invention adopts the following technical solution, an intelligent lock unlocking method based on face image processing, comprising the following steps:

[0007] S1: Through an image acquisition device, obtain facial images from multiple angles, extract the key facial features of the user, including facial contour, facial symmetry, and the positions of facial features, and generate a facial feature dataset;

[0008] S2: Based on the facial feature dataset, calculate the difference degree of facial images from each perspective, evaluate the stability and richness of features in multiple perspective images, assign weights to multiple datasets and perform fusion to generate a dataset fusion result;

[0009] S3: Through the dataset fusion result, extract the geometric information of the face from facial images from multiple perspectives, calculate the depth information of multiple positions on the face, and generate facial three-dimensional feature data;

[0010] S4: Invoke the facial three-dimensional feature data, combine with a light sensor to continuously monitor the ambient light intensity, set a light tolerance threshold and send a rejection signal by analyzing the face recognition accuracy under various light intensities, and generate a rejection signal judgment result;

[0011] S5: Based on the rejection signal judgment result, analyze the matching degree of multiple local feature regions, face contour and three-dimensional structure features in the image, calculate the matching degree score of the image, and send a control instruction to the intelligent lock to generate intelligent lock control parameters.

[0012] As a further solution of the present invention, the facial feature dataset includes facial images from multiple perspectives, extracted facial feature data, and facial contour information. The dataset fusion result includes feature difference degree, stability evaluation result, and dataset weight assignment result. The facial three-dimensional feature data includes depth change data, curvature feature information, and convex and concave region marking results. The rejection signal judgment result includes light adaptability information, recognition accuracy data, and rejection signal sending records. The intelligent lock control parameters include unlocking instructions, rejection instructions, and authentication status.

[0013] As a further solution of the present invention, the step of obtaining facial images from multiple angles through an image acquisition device, extracting the key facial features of the user, including facial contour, facial symmetry, and the positions of facial features, and generating a facial feature dataset is specifically as follows:

[0014] S101: Use an image acquisition device to obtain facial images of the user from multiple angles, and record the shooting angle and distance information of each image to generate facial image data;

[0015] S102: Invoke the facial image data, preprocess the image, including removing noise and performing image normalization, unifying the image size and color space, and generating a preprocessed image;

[0016] S103: Invoke the preprocessed image, extract multiple key feature points on the face, including eyes, nose, mouth, and chin, identify the facial features, and analyze the facial contour and symmetry of the face, and generate a facial feature dataset.

[0017] As a further solution of the present invention, based on the facial feature dataset, calculate the difference degree of facial images from each perspective, evaluate the stability and richness of features in multiple perspective images, assign weights to multiple datasets and perform fusion, and the steps of generating a dataset fusion result are specifically as follows:

[0018] S201: Invoke the facial feature dataset, calculate the difference degree of facial images from each perspective according to the facial image features under multiple perspectives, and obtain perspective difference degree data;

[0019] S202: Based on the perspective difference degree data, evaluate the stability and richness of features in multiple perspective facial images, and generate a feature stability evaluation result;

[0020] S203: According to the feature stability evaluation result, assign fusion weights to the feature datasets of multiple perspective images, perform fusion processing on multiple datasets, and generate a dataset fusion result.

[0021] As a further solution of the present invention, through the dataset fusion result, extract the geometric information of the face from facial images of multiple perspectives, calculate the depth information of multiple positions on the face, and the steps of generating facial three-dimensional feature data are specifically as follows:

[0022] S301: Based on the dataset fusion result, calculate the depth information of multiple positions on the face according to the facial images of multiple perspectives, and generate facial depth data;

[0023] S302: Based on the facial depth data, calculate the curvature information of multiple positions on the face, and generate facial curvature data;

[0024] S303: According to the facial curvature data, identify multiple convex and concave regions on the user's face, and generate facial three-dimensional feature data.

[0025] As a further solution of the present invention, invoke the facial three-dimensional feature data, combine with a light sensor, real-time monitor the ambient light intensity, set a light tolerance threshold and send a rejection signal by analyzing the face recognition accuracy under multiple light intensities, and the steps of generating a rejection signal judgment result are specifically as follows:

[0026] S401: Obtain the three-dimensional facial feature data, and in combination with the light sensor, collect the ambient light intensity data in real time to generate ambient light data;

[0027] S402: Based on the ambient light data, calculate the face recognition accuracy under multiple light intensities, analyze the recognition effects under multiple light conditions, and generate light recognition accuracy data;

[0028] S403: According to the light recognition accuracy data, set a light tolerance threshold and compare it with the real-time light intensity. When the light condition does not meet the image recognition requirement, send a rejection signal and generate a judgment result of the rejection signal.

[0029] As a further solution of the present invention, the specific formula for calculating the face recognition accuracy under multiple light intensities is:

[0030]

[0031] Calculate the recognition accuracy;

[0032] where P(I) represents the recognition accuracy under the current light intensity I in the face recognition test, N correct represents the number of correctly recognized faces under this light intensity, N total represents the total number of face recognitions under this light intensity, I is the current experimental light intensity, I ideal represents the optimal ideal light intensity, α is a weight coefficient used to adjust the influence of light error on the recognition accuracy, β is a coefficient used to adjust the influence of light intensity change on the recognition accuracy, I j is the light intensity value measured during the experiment, I j-1 is the previous light intensity value, m represents the number of light intensity data points, and j represents the index of the light intensity data test.

[0033] As a further solution of the present invention, based on the judgment result of the rejection signal, analyze the matching degrees of multiple local feature regions, face contours, and three-dimensional structure features in the image, calculate the matching degree score of the image, and send a control instruction to the intelligent lock to generate the steps of the intelligent lock control parameters specifically as follows:

[0034] S501: Based on the judgment result of the rejection signal, calculate the matching degree between multiple regional feature values and a preset feature reference value according to the multiple local region features in the user's face image, and generate a local feature matching degree value;

[0035] S502: Invoke the local feature matching degree value, analyze the key point distribution of the face contour in the image, and in combination with the three-dimensional structure features, calculate the matching degree of the face contour and the three-dimensional structure features, and generate a contour and structure matching degree value;

[0036] S503: Call the contour and structure matching degree value, calculate the matching degree score of the face image, adjust the control parameters of the door lock, and generate the intelligent lock control parameters.

[0037] As a further solution of the present invention, the specific formula for calculating the matching degree score of the face image is:

[0038]

[0039] Calculate the matching degree score, adjust the door lock control parameters, and generate the intelligent lock control parameters;

[0040] Among them, M 89 represents the matching degree of the i'-th local area, C face represents the face contour matching degree, C 3D represents the three-dimensional structure feature matching degree, w 89 represents the weight of the i'-th local feature matching degree, w 1 represents the weight of the local feature matching degree, w 2 represents the weight of the face contour matching degree, w < represents the weight of the three-dimensional structure feature matching degree, i' represents the number of the local area involved in the local feature matching degree, n' represents the total number of local features, and S' represents the matching degree score.

[0041] On the other hand, an intelligent lock unlocking system based on face image processing is provided. This system is applied to the intelligent lock unlocking method based on face image processing. The system includes:

[0042] The multi-angle image acquisition module uses an image acquisition device to capture the user's face image from multiple directions, extract image features, and construct a facial feature data set;

[0043] The feature weight optimization module calls the facial feature data set, analyzes the differences in features from multiple perspectives, evaluates the consistency and richness of features, assigns weights to multiple data sets and performs data fusion to form a data set fusion result;

[0044] The facial depth analysis module extracts facial geometric structure information from the multi-angle images according to the data set fusion result, calculates the depth values of multiple regions of the face, and obtains the facial three-dimensional feature data;

[0045] The ambient light adaptation module uses the facial three-dimensional feature data, combines with a light sensor to monitor the ambient light conditions, evaluates the influence of various light intensities on the face recognition effect, sets a light tolerance threshold, and generates a rejection signal judgment result;

[0046] The unlocking decision execution module evaluates the matching degree of facial local features, contours, and three-dimensional structures based on the recognition rejection signal judgment result, calculates the matching degree score, and sends a control instruction to the smart lock to generate smart lock control parameters.

[0047] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0048] By collecting multi-angle facial images, constructing a multi-dimensional feature data set, realizing feature fusion through difference degree evaluation and weight assignment, enhancing the stability and anti-interference ability of biometric information, extracting three-dimensional geometric information and depth data, breaking through the limitations of traditional two-dimensional plane recognition, combining with a dynamic light threshold monitoring mechanism, suppressing the influence of environmental variables on the recognition accuracy, and improving the resistance to planar attacks, so that the identity verification system can achieve a double improvement in security level and scene adaptability while maintaining convenience. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0050] Figure 1 It is a schematic diagram of the working process of the present invention;

[0051] Figure 2 It is a detailed flowchart of S1 of the present invention;

[0052] Figure 3 It is a detailed flowchart of S2 of the present invention;

[0053] Figure 4 It is a detailed flowchart of S3 of the present invention;

[0054] Figure 5 It is a detailed flowchart of S4 of the present invention;

[0055] Figure 6 It is a detailed flowchart of S5 of the present invention;

[0056] Figure 7 It is a system flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] The following will describe the technical solutions in the present invention with reference to the drawings.

[0058] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design described as an "example" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0059] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "Of", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0060] In the embodiments of the present invention, sometimes a subscript such as W 1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0061] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0062] Please refer to Figure 1 , the present invention provides a technical solution, an intelligent lock unlocking method based on face image processing, including the following steps:

[0063] S1: Through an image acquisition device, obtain facial images from multiple angles, extract the key facial features of the user, including facial contour, facial symmetry, and the positions of facial features, and generate a facial feature dataset;

[0064] S2: Based on the facial feature dataset, calculate the difference degree of each perspective facial image, evaluate the stability and richness of features in multiple perspective images, assign weights to multiple datasets and perform fusion to generate a dataset fusion result;

[0065] S3: Through the dataset fusion result, extract the geometric information of the face from facial images of multiple perspectives, calculate the depth information of multiple positions on the face, and generate facial three-dimensional feature data;

[0066] S4: Call the facial three-dimensional feature data, combine it with a light sensor to monitor the ambient light intensity in real time, set a light tolerance threshold and send a rejection signal by analyzing the face recognition accuracy under various light intensities, and generate a rejection signal judgment result;

[0067] S5: Based on the recognition rejection signal judgment result, analyze the matching degrees of multiple local feature regions, facial contours, and three-dimensional structural features in the image, calculate the matching degree score of the image, send a control instruction to the intelligent lock, and generate intelligent lock control parameters.

[0068] The facial feature dataset includes facial images from multiple perspectives, extracted facial feature data, and facial contour information. The dataset fusion result includes feature difference degree, stability evaluation result, and dataset weight allocation result. The facial three-dimensional feature data includes depth change data, curvature feature information, and marked results of convex and concave regions. The recognition rejection signal judgment result includes light adaptability information, recognition accuracy data, and recognition rejection signal sending records. The intelligent lock control parameters include unlocking instructions, recognition rejection instructions, and authentication status.

[0069] Please refer to Figure 2 , to obtain facial images from multiple angles through an image acquisition device, extract the key facial features of the user, including facial contour, facial symmetry, and facial feature positions, the steps for generating the facial feature dataset are specifically as follows:

[0070] S101: Use the image acquisition device to obtain facial images of the user from multiple angles, and record the shooting angle and distance information of each image to generate facial image data;

[0071] When obtaining facial images of the user from multiple angles, first place the image acquisition device at different shooting angle positions, and obtain the user's facial images at different distances to ensure that the collected facial images cover all angles, and record the shooting angle and distance information of each image. The image acquisition device usually uses a high-definition camera, combined with a stable tripod device to avoid image jitter. In the actual implementation process, a fixed shooting angle range can be set, such as a horizontal rotation range between 30° and 150°, or a vertical range between 30° and 90°, and ensure that the distance between the device and the shooting object is adjustable, usually the distance is set in the range of 1 meter to 3 meters. Assume the shooting distance is d = 2 meters, and the angle range is θ 1 = 30° to θ 2 = 150°, through this range, different-angle images can be ensured to be obtained. If the shooting angle of the device and the distance from the person are fixed, the shooting point position at each angle can be calculated. For example, assume that the relationship between the shooting point position and the angle change can be represented by trigonometric functions. Assume that the camera is always at a distance of 2 meters. When the shooting angle is θ = 90°, the position change on its x-axis is:

[0072] x = d·cos(θ) = 2·cos(90°) = 0;

[0073] That is, the position does not shift. When θ = 30°, the shift on the x-axis is:

[0074] x = d·cos(30°) = 2·cos(30°) ≈ 1.732;

[0075] After image acquisition, record these angle and distance information for subsequent facial data processing and analysis, and finally generate facial image data. Through this process, facial images at multiple angles and different distances are obtained, providing complete data support for subsequent analysis.

[0076] S102: Call the facial image data and preprocess the image, including removing noise and performing image normalization, unifying the image size and color space, and generating a preprocessed image;

[0077] When preprocessing the acquired image data, first perform denoising on the image, which is carried out in multiple ways, including using a Gaussian filter for image smoothing to reduce noise in the image. First, find the noise area in the acquired image, and perform weighted averaging on each pixel in the image through Gaussian convolution operation, based on the Gaussian kernel function. The filtered image will reduce the interference of noise. The formula for the Gaussian filter is:

[0078]

[0079] Where G(x,y) represents the filter value at the coordinate position x,y, σ is the standard deviation, which controls the width of the Gaussian kernel, and exp is the natural exponential function. Assuming σ = 1, for a pixel point located at (2,3), substituting into the formula gives:

[0080]

[0081] Next, the image normalization process needs to ensure that all images have the same size and are converted to a unified color space for subsequent unified processing. During the size normalization process, first unify images of different sizes to a specified size, using bilinear interpolation method for adjustment to keep the image ratio unchanged. When converting the color space, usually convert the image from the RGB color space to the standard YUV color space, and the conversion formula is:

[0082] Y = 0.299R + 0.587G + 0.114B;

[0083] U = -0.169R - 0.331G + 0.499B;

[0084] V = 0.499R - 0.419G - 0.081B;

[0085] Assuming there is a pixel point with R = 255, G = 100, B = 50, then when converting to the YUV color space:

[0086] Y = 0.299×255 + 0.587×100 + 0.114×50 = 76.245 + 58.7 + 5.7 = 140.645;

[0087] U = -0.169×255 - 0.331×100 + 0.499×50 = -43.095 - 33.1 +

[0088] 24.95 = -51.245;

[0089] V = 0.499×255 - 0.419×100 - 0.081×50 = 127.245 - 41.9 - 4.05 = 81.295;

[0090] The processed image can eliminate the differences in size and color, making the subsequent facial feature extraction process more accurate.

[0091] S103: Call the preprocessed image, extract multiple key feature points on the face, including eyes, nose, mouth, and chin, identify the facial features, and analyze the facial contour and symmetry of the face to generate a facial feature dataset;

[0092] When extracting multiple key feature points on the face, first identify the facial region in the image. Usually, a deep learning-based face detection algorithm is used to complete this task, such as the Haar cascade classifier or the MTCNN algorithm. Detect the image through this algorithm to locate the position of the face in the image. Subsequently, use a fine key point detection algorithm to identify key points such as eyes, nose, mouth, and chin, and obtain a set of coordinate values to represent these key positions. When further analyzing the facial contour and symmetry, first calculate the relative distances between feature points such as eyes, nose, and mouth, and analyze the symmetry of the face by calculating the Euclidean distance between these feature points. For example, the Euclidean distance formula for calculating the distance between the left eye and the right eye is:

[0093]

[0094] Assume the coordinates of the left eye are (x 1 , y 1 ) = (50, 70), and the coordinates of the right eye are (x 2 , y 2 ) = (150, 70), then its Euclidean distance is:

[0095]

[0096] Evaluate the balance of the face according to the symmetry of the parameters. When performing symmetry analysis, a threshold can be set. For example, an image with a difference in Euclidean distance less than 5% is considered symmetric, otherwise it is regarded as an asymmetric image. Compare facial images at different angles, and evaluate the changes of the face at different angles through the changes in distance and angle, and generate a facial feature dataset.

[0097] Please refer to Figure 3 , based on the facial feature dataset, calculate the difference degree of facial images at each perspective, evaluate the stability and richness of features in multiple perspective images, assign weights to multiple datasets and perform fusion. The specific steps for generating the dataset fusion result are as follows:

[0098] S201: Call the facial feature dataset, and calculate the difference degree of facial images at each perspective according to the facial image features at multiple perspectives to obtain perspective difference degree data;

[0099] When calling the facial feature dataset, first calculate the difference degree of facial images at each perspective according to the facial image features at multiple perspectives. The calculation of the difference degree requires quantifying the difference of feature points in each image. Usually, the difference between feature points can be quantified by calculating the Euclidean distance. The specific calculation method is as follows:

[0100]

[0101] where D represents the difference degree of facial feature points, x 8 , y 8 are the coordinates of the feature points in the i-th perspective image, x 8 ′, y 8 ′ are the coordinates of the feature points in another perspective image, and n is the total number of feature points. By calculating the difference degrees between all perspectives, the difference degree data of facial images at each perspective can be obtained. The difference degree data helps to further evaluate the changes and stability of facial features at different perspectives. For example, assume that the coordinates of the eye positions are (x 1 , y 1 ) = (1, 2) and (x 1 ′, y 1 ′) = (1.1, 2.1), and the coordinates of the mouth positions are (x 2 , y 2 ) = (3, 4) and (x 2 ′, y 2 ′0 = (3.2, 4.10. Then the difference degrees of the eyes and the mouth are respectively:

[0102]

[0103] Based on the difference data of multiple perspective images, the change degree of feature points in each perspective image at different shooting angles can be analyzed, providing data support for subsequent analysis.

[0104] S202: Based on the perspective difference data, evaluate the stability and richness of features in multiple perspective facial images, and generate a feature stability evaluation result;

[0105] Based on the perspective difference data, it is necessary to evaluate the stability and richness of features in multiple perspective facial images. The stability evaluation is mainly based on the difference in the difference degree of the same feature points under different perspectives. By comparing the difference degrees of the same feature points in different perspective images, its stability is judged. If the difference degree is less than the set threshold, it is considered that the stability of this feature point is higher under multiple perspectives. For example, set the threshold as T = 0.2. If the difference degree of a certain feature point is calculated to be less than T for many consecutive times, it is considered that its stability is higher. Suppose the difference degree of the eye position is D eye = 0.1, and the difference degree of the mouth position is D mouth = 0.25, then the eyes have higher stability and the mouth has lower stability. The richness evaluation is mainly determined based on the diversity of feature points. By counting the number of feature points that can be accurately extracted in each perspective and calculating the coverage of feature points, if the number of feature points is large, it is considered that the image has a high richness. The richness formula is:

[0106]

[0107] where R represents richness, N features represents the number of feature points extracted in one perspective, and N max represents the maximum number of feature points that should be extracted theoretically. Suppose 12 feature points are extracted in a certain perspective and the maximum number of feature points is 15, then the richness of this perspective is:

[0108]

[0109] Through the above evaluations of difference degree, stability and richness, a feature stability evaluation result is obtained, providing a basis for subsequent data fusion processing.

[0110] S203: According to the feature stability evaluation result, assign fusion weights to the feature datasets of multiple perspective images, and perform fusion processing on multiple datasets to generate a dataset fusion result;

[0111] According to the feature stability evaluation result, it is necessary to assign fusion weights to the feature datasets of multiple perspective images. The specific method is to set weights according to the stability and richness of each feature point. The weight assignment can be determined by the following formula:

[0112] W feature= α × S + β × R;

[0113] Among them, W feature represents the fusion weight of the feature, S represents the stability score, R represents the richness score, α and β are the weight coefficients of stability and richness respectively, and are usually set to α = 0.7 and β = 0.3, which indicates that stability dominates in weight allocation. Suppose the stability score of a certain feature point is 0.9 and the richness score is 0.8, then the fusion weight of this feature is:

[0114] W feature = 0.7 × 0.9 + 0.3 × 0.8 = 0.87;

[0115] After weight allocation for all feature points according to this method, multiple datasets can be fused to generate the dataset fusion result. The fusion process is carried out by weighted average, combining the weighted values of all feature points to obtain the final fused dataset.

[0116] Please refer to Figure 4 , and through the dataset fusion result, from facial images in multiple perspectives, the steps to extract the geometric information of the face, calculate the depth information of multiple positions on the face, and generate the three-dimensional feature data of the face are specifically as follows:

[0117] S301: Based on the dataset fusion result, according to the facial images in multiple perspectives, calculate the depth information of multiple positions on the face to generate facial depth data;

[0118] Based on the dataset fusion result, calculating the depth information of multiple positions on the face requires extracting the three-dimensional coordinates of each feature point from multiple perspective images, and then calculating the depth of each position under different perspectives. The basic idea of depth calculation is to obtain the depth value of the target point through triangulation. Suppose the coordinates of a certain feature point P(x 1 , y 1 ) in the first perspective and the position parameters of the camera are known, and at the same time, the coordinates of this point in the second perspective P(x 2 , y 2 ) are also known. Through triangulation, the depth Z of this point can be calculated, and the depth value calculation formula is:

[0119]

[0120] Among them, f is the focal length of the camera, b is the baseline distance between the two cameras, |x 1 - x 2| is the horizontal displacement of the same point in different perspectives. The depth value of the feature point in each perspective is calculated through this formula, and combined with the information of images in different perspectives, facial depth data is generated. In practical applications, assuming the focal length f = 500 pixels and the baseline distance b = 0.5 meters, if the displacement of a certain feature point is 0.2 pixels, then its depth is:

[0121]

[0122] Through this method, corresponding depth data can be generated for each facial position.

[0123] S302: Based on the facial depth data, calculate the curvature information of multiple positions on the face to generate facial curvature data;

[0124] Based on the facial depth data, calculate the curvature information of multiple positions on the face. First, analyze the degree of bending of the facial area through the depth data of each point. The calculation of curvature is usually based on the depth information of the neighborhood of the point. Assuming that the depth values of a certain point P(x, y, Z) and its neighborhood points are known, then the curvature K can be calculated by the following formula:

[0125]

[0126] Among them, R is the radius, and are the depth gradients of this point in the x and y directions respectively. By calculating the curvature of each point through this formula, the curvature information of each area on the facial surface can be obtained. In practical applications, assuming that the depth of a certain point is Z = 5 cm, and the depth change rates near this point are and then the curvature K is:

[0127]

[0128] According to the calculation results, generate the curvature data of the face.

[0129] S303: According to the facial curvature data, identify multiple convex and concave areas on the user's face to generate facial three-dimensional feature data;

[0130] Based on the facial curvature data, multiple convex and concave regions in the user's face are identified, mainly by performing threshold classification on the curvature data. First, a threshold T of the curvature value needs to be set. When the curvature K of a certain region is greater than T, that region is considered a convex region. When K is less than -T, that region is considered a concave region. The specific process includes extracting the curvature information of all feature points from the facial image and calibrating each region according to the curvature value. Suppose the threshold T = 0.8 is set. If the curvature of a certain region is K = 0.9, then that region is considered a convex region. If the curvature is K = -1.2, then it is a concave region. Further process these curvature data, classify the convex regions and the concave regions, and generate the three-dimensional facial feature data. In practical applications, suppose the curvature data extracted from the facial region are multiple values: K 1 = 0.9, K 2 = -1.2, K < = 0.5. After setting the threshold T = 0.8, it can be recognized that region 1 is convex, region 2 is concave, and region 3 is a flat region. Finally, the three-dimensional facial feature data is generated and output.

[0131] Please refer to Figure 5 , call the three-dimensional facial feature data, combine with the light sensor, and monitor the ambient light intensity in real time. By analyzing the face recognition accuracy under various light intensities, set the light tolerance threshold and send a rejection signal. The steps to generate the judgment result of the rejection signal are specifically as follows:

[0132] S401: Obtain the three-dimensional facial feature data, combine with the light sensor, and collect the ambient light intensity data in real time to generate the ambient light data;

[0133] After obtaining the three-dimensional facial feature data, the process of collecting the ambient light intensity data in real time by combining with the light sensor requires using the light sensor to detect and record the light intensity in the environment in real time. This process usually obtains the light intensity through the output voltage or signal value of the light sensor. The relationship between the light intensity value and the voltage output by the sensor can be expressed by a linear formula:

[0134] I = k × V;

[0135] where I is the ambient light intensity, k is the sensitivity constant of the sensor, and V is the voltage value output by the sensor. Suppose k = 1.5 mV / lux and the sensor output voltage is V = 2 mV, then the light intensity is:

[0136] I = 1.5 × 2 = 3 lux;

[0137] Through this formula, the real-time light intensity data can be calculated.

[0138] S402: Calculate the face recognition accuracy under multiple light intensities based on the environmental light data, analyze the recognition effects under multiple light conditions, and generate light recognition accuracy data;

[0139] The specific formula for calculating the face recognition accuracy under multiple light intensities is:

[0140]

[0141] Calculate the recognition accuracy;

[0142] Among them, P(I) represents the recognition accuracy under the current light intensity I in the face recognition test, N correct represents the number of correctly recognized faces under this light intensity, N total represents the total number of face recognitions performed under this light intensity, I is the current experimental light intensity, I ideal represents the optimal ideal light intensity, α is the weight coefficient used to adjust the influence of light error on the recognition accuracy, β is the coefficient used to adjust the influence of light intensity change on the recognition accuracy, I j is the light intensity value tested during the experiment, I j-1 is the previous light intensity value, m represents the number of light intensity data points, and j represents the index of the light intensity data test.

[0143] Formula:

[0144]

[0145] Detailed explanation of the formula and the derivation process of the formula calculation:

[0146] The formula is used to calculate the recognition accuracy under multiple light intensities in the face recognition test. The result is used to evaluate the influence of light intensity on the face recognition accuracy and analyze the recognition effect according to its change;

[0147] Meaning and setting values of parameters:

[0148] N correct is the number of correctly recognized faces under the given light intensity, N total is the total number of recognitions. Assume that 100 recognitions are performed under the condition of I = 100 lux, and 80 of them are recognized correctly, that is, N correct = 80, N total = 100;

[0149] α is the weight coefficient of light error. Assume that α is 0.02;

[0150] I ideal is the ideal light intensity. Assume that I ideal = 200 lux;

[0151] β is the influence coefficient of light change, which is used to adjust the influence of light intensity change on the recognition accuracy. Assume β is 0.05;

[0152] I j represents the light values under multiple test light intensities. I j-1 is the light value of the previous test point, and m is the number of test light intensity points. Assume that in an experiment, 5 different light intensity values are tested, where I 1 = 100 lux, I 2 = 150 lux, I < = 200 lux, I r = 250 lux, I a = 300 lux, m = 5;

[0153] Substitute the parameters into the formula for calculation:

[0154]

[0155] |I - I ideal | = |100 - 200| = 100 lux;

[0156] α × |I - I ideal | = 0.02 × 100 = 2;

[0157] 1 + α × |I - I ideal | = 1 + 2 = 3;

[0158] (I 1 - I g ) 2 = (100 - 0) 2 = 10000;

[0159] (I 2 - I 1 ) 2 = (150 - 100) 2 = 2500;

[0160] (I < - I 2 ) 2 = (200 - 150) 2 = 2500;

[0161] (I r - I < 0 2 = (250 - 200) 2 = 2500;

[0162] (I a - I r ) 2 = (300 - 250)2 = 2500;

[0163]

[0164]

[0165] P(I) = 0.8×(3 + 7.07) = 0.8×10.07 = 8.056;

[0166] The result 8.056 indicates that the accuracy of face recognition under the target lighting conditions is 80.56%, and the result is used to set the lighting tolerance threshold and determine whether to send a rejection signal.

[0167] S403: According to the lighting recognition accuracy data, set the lighting tolerance threshold and compare it with the real-time lighting intensity. When the lighting conditions do not meet the image recognition requirements, send a rejection signal and generate a judgment result of the rejection signal;

[0168] According to the lighting recognition accuracy data, set the lighting tolerance threshold and compare it with the real-time lighting intensity. First, set a lighting tolerance threshold T. When the real-time lighting intensity I real is lower than this threshold, the system determines that the lighting conditions are insufficient and sends a rejection signal. The process of setting the lighting tolerance threshold can be calculated based on experimental data. Assume that the lighting tolerance threshold T needs to be selected from the recognition accuracy decline under different lighting intensities. When the real-time lighting intensity I real is lower than the threshold, a rejection signal is generated. The calculation of the rejection signal is based on the relationship between the real-time lighting intensity and the threshold, and the calculation formula is:

[0169] ΔI = T - I real ;

[0170] where ΔI represents the difference between the lighting intensity and the threshold. When ΔI < 0, it means that the real-time lighting intensity is lower than the set threshold, and the system sends a rejection signal. Assume that the set lighting tolerance threshold T = 50 lux and the real-time lighting intensity I real = 40 lux, then the calculation is as follows:

[0171] ΔI = 50 - 40 = 10;

[0172] Since ΔI > 0, it means that the real-time lighting intensity is greater than the threshold, and the system will not send a rejection signal. Assume that the real-time lighting intensity I real = 60 lux, then:

[0173] ΔI = 50 - 60 = -10;

[0174] Since ΔI < 0, the system will send a rejection signal.

[0175] Please refer to Figure 6, based on the result of the rejection signal judgment, analyze the matching degrees of multiple local feature regions, face contours, and three-dimensional structural features in the image, calculate the matching degree score of the image, and send a control instruction to the intelligent lock to generate the intelligent lock control parameters. The specific steps are as follows:

[0176] S501: Based on the result of the rejection signal judgment, calculate the matching degrees of multiple regional feature values and preset feature reference values according to the multiple local region features in the user's face image, and generate local feature matching degree values;

[0177] After judging according to the rejection signal result, first calculate the matching degree between multiple local region features and the preset feature reference value. For the features of each local region in the face image, first extract the local region features of the image, such as the texture, shape, edges, etc. of parts such as eyes, nose, mouth, and chin. These features are usually extracted by means of pixel intensity, edge detection, or color histogram, etc., and then compared with the preset reference features. The reference features represent a standard template or previously known ideal features. When calculating the matching degree, first compare the feature value of each region with the corresponding reference value through a formula, calculate the difference degree of each local region, and generate a matching degree value according to the difference degree. The calculation formula is as follows:

[0178]

[0179] where M is the matching degree of the local feature, f 8 is the feature value of the i-th local region, b 8 is the preset feature reference value of this region, and n is the number of local regions. By calculating this formula, the matching degree values of each local region can be obtained, and these values represent the similarity degree of this region to the preset features. For example, if the feature value of the eye region is f 1 =0.80 and the reference value is b 1 =0.85, the matching degree calculation is:

[0180]

[0181] In this way, the matching degree of the features of each region with the reference value can be obtained, and the feature matching degree of the overall face image can be further judged.

[0182] S502: Call the local feature matching degree value, analyze the key point distribution of the face contour in the image, and combine with the three-dimensional structural features to calculate the matching degree of the face contour and the three-dimensional structural features, and generate the contour and structure matching degree value;

[0183] Next, based on the local feature matching degree value, analyze the distribution of key points of the face contour in the image, and combine the three-dimensional structure features to calculate the matching degree. First, extract the positions of the key points in the image. The key points are usually the coordinates of the significant features of the face, such as the corners of the eyes, eyebrows, the tip of the nose, and the chin. The two-dimensional positions of these points need to be compared with the corresponding three-dimensional structure features. The three-dimensional structure features include the three-dimensional shape information of the face, which describes the depth changes of the facial features when observing the face from different angles. To calculate the matching degree, the two-dimensional coordinates of each key point need to be mapped into the three-dimensional space to obtain the corresponding depth value. Then, calculate the difference degree of each point in the three-dimensional space, and combine these difference degrees to obtain two parts: "face contour matching degree" and "three-dimensional structure feature matching degree". The specific calculation formulas are as follows:

[0184] Formula for calculating the face contour matching degree:

[0185]

[0186] Among them, C face is the face contour matching degree, p 8 is the two-dimensional coordinate of the i-th key point, t 8 is the three-dimensional preset coordinate value, and n is the number of key points.

[0187] Formula for calculating the three-dimensional structure feature matching degree:

[0188]

[0189] Among them, C 3D is the matching degree of the three-dimensional structure feature, d 8 is the depth value of the i-th key point, d′ 8 is the preset depth value of the corresponding three-dimensional structure of this point, and n is the number of key points.

[0190] Suppose there are 3 key points, and the extracted two-dimensional coordinates are p 1 =(1.0, 2.5), p 2 =(2.0, 3.5), p < =(3.0, 4.5), and the corresponding three-dimensional preset coordinates are t 1 =(1.1, 2.4), t 2 =(2.1, 3.4), t < =(3.1, 4.6), and the corresponding depth values are d 1 =1.2, d 2 =2.2, d < =3.2, and the preset depth value is d′ 1 =1.1, d′ 2 =2.1, d′ < =3.1, then the calculated matching degree is:

[0191]

[0192] Calculate the matching degree between the facial contour and the three-dimensional structure through a formula.

[0193] S503: Call the matching degree value between the contour and the structure, calculate the matching degree score of the face image, adjust the control parameters of the door lock, and generate the intelligent lock control parameters;

[0194] The specific formula for calculating the matching degree score of the face image is:

[0195]

[0196] Calculate the matching degree score, adjust the door lock control parameters, and generate the intelligent lock control parameters;

[0197] Among them, M 89 represents the matching degree of the i'-th local area, C face represents the facial contour matching degree, C 3D represents the three-dimensional structure feature matching degree, w 89 represents the weight of the matching degree of the i'-th local feature, w 1 represents the weight of the local feature matching degree, w 2 represents the weight of the facial contour matching degree, w < represents the weight of the three-dimensional structure feature matching degree, i' represents the number of the local area involved in the local feature matching degree, n' represents the total number of local features, and S' represents the matching degree score.

[0198] Formula:

[0199]

[0200] Detailed explanation of the formula and the derivation process of the formula calculation:

[0201] The formula is used to calculate the matching degree score of the facial image, and the score is used to adjust the control parameters of the intelligent lock;

[0202] Meaning and setting values of the parameters:

[0203] M 89 : The matching degree of the i'-th local area. The matching degree of the local area is calculated based on the difference between the local features extracted from the image and the preset reference value, reflecting the matching degree of the local area features. Assume M 1 = 0.05, M 2 = 0.04, M < = 0.07;

[0204] w 89: The weight of the matching degree of the i'-th local area, indicating the influence degree of the matching degree of this local area on the overall matching degree. Assume w 1 = 0.3, w 2 = 0.2, w < = 0.5;

[0205] C face : The face contour matching degree, indicating the matching degree between the two-dimensional coordinates of the face contour and the preset three-dimensional contour coordinates. Assume C face = 0.1;

[0206] C 3D : The three-dimensional structure feature matching degree, indicating the matching degree between the three-dimensional depth information of the facial features and the preset three-dimensional structure features when observing the face from different angles. Assume C 3D = 0.2;

[0207] w 1 : The weight of the local feature matching degree, indicating the weight of the local feature matching degree in the calculation of the total matching degree. Assume w 1 = 0.3;

[0208] w 2 : The weight of the face contour matching degree, indicating the importance of the face contour in the total matching degree score. Assume w 2 = 0.4;

[0209] w < : The weight of the three-dimensional structure feature matching degree, indicating the contribution of the three-dimensional structure features to the total matching degree score. Assume w < = 0.3;

[0210] n': The total number of local features, indicating the number of local areas extracted from the face image. Assume n' = 3, corresponding to three feature areas at different positions;

[0211] Substitute the parameters into the formula for calculation:

[0212]

[0213] The result 0.158 indicates that the overall matching degree of the face image is relatively low. The value is used to further adjust the control parameters of the intelligent lock to determine whether to unlock the door lock. The value reflects the comprehensive matching situation of multiple local features, contours, and three-dimensional structure features, providing a decision-making basis for the intelligent lock control.

[0214] Please refer to Figure 7 , the intelligent lock unlocking system based on face image processing. The intelligent lock unlocking system based on face image processing is used to execute the above-mentioned intelligent lock unlocking method based on face image processing. The system includes:

[0215] The multi-angle image acquisition module uses an image acquisition device to capture the user's facial images from multiple directions, extract image features, and construct a facial feature dataset;

[0216] The feature weight optimization module calls the facial feature dataset, analyzes the differences in features from multiple perspectives, evaluates the consistency and richness of features, assigns weights to multiple datasets and performs data fusion to form a dataset fusion result;

[0217] The facial depth analysis module extracts facial geometric structure information from multi-angle images based on the dataset fusion result, calculates the depth values of multiple facial regions, and obtains facial three-dimensional feature data;

[0218] The ambient light adaptation module uses the facial three-dimensional feature data, combines with a light sensor to monitor the ambient light conditions, evaluates the influence of various light intensities on the face recognition effect, sets a light tolerance threshold, and generates a rejection signal judgment result;

[0219] The unlocking decision execution module evaluates the matching degree of facial local features, contours, and three-dimensional structures according to the rejection signal judgment result, calculates a matching degree score, and sends a control instruction to the smart lock to generate smart lock control parameters.

[0220] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0221] It should be understood that the term "and / or" in this text is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. Additionally, the character " / " in this text generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship. Specifically, it can be understood by referring to the context before and after.

[0222] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following items" or similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0223] It should be understood that in various embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0224] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0225] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0226] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in an electrical, mechanical, or other form.

[0227] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed over multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0228] In addition, each functional unit in various embodiments of the present invention may be integrated in a processing unit, may exist separately physically for each unit, or two or more units may be integrated in one unit.

[0229] If the above-mentioned function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0230] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A smart lock unlocking method based on face image processing, characterized in that: The method comprises: S1: Obtain facial images from multiple angles through image acquisition equipment, extract key facial features of the user, including facial contour, facial symmetry, and facial features, and generate a facial feature dataset; S2: Based on the facial feature dataset, calculate the difference of facial images of each perspective, evaluate the stability and richness of features in multiple perspective images, assign weights to multiple datasets and fuse them to generate a dataset fusion result; S3: extracting facial geometric information through the data set fusion results and facial images from multiple perspectives, calculating depth information at multiple locations of the face, and generating facial three-dimensional feature data; S4: calling the facial three-dimensional feature data, combining with the light sensor, monitoring the ambient light intensity in real time, analyzing the face recognition accuracy under various light intensities, setting the light tolerance threshold and sending a rejection signal, and generating a rejection signal judgment result; S5: Based on the rejection signal judgment result, analyze the matching degree of multiple local feature areas, facial contours and three-dimensional structural features in the image, calculate the image matching score, send a control instruction to the smart lock, and generate smart lock control parameters.

2. The smart lock unlocking method based on face image processing according to claim 1 is characterized in that: The facial feature dataset includes facial images from multiple perspectives, extracted facial feature data, and facial contour information; the dataset fusion result includes feature difference, stability evaluation result, and dataset weight distribution result; the facial three-dimensional feature data includes depth change data, curvature feature information, and convex and concave area marking results; the rejection signal judgment result includes lighting adaptability information, recognition accuracy data, and rejection signal sending records; the smart lock control parameters include unlocking instructions, rejection instructions, and identity authentication status.

3. The smart lock unlocking method based on face image processing according to claim 1 is characterized in that: Through the image acquisition device, facial images from multiple angles are obtained to extract the user's key facial features, including facial contour, facial symmetry, and facial features. The steps to generate the facial feature data set are as follows: S101: using an image acquisition device to obtain facial images of a user at multiple angles, and recording shooting angle and distance information of each image to generate facial image data; S102: calling the facial image data, preprocessing the image, including removing noise and performing image standardization, unifying the image size and color space, and generating a preprocessed image; S103: calling the preprocessed image, extracting multiple key feature points of the face, including eyes, nose, mouth, and chin, identifying the facial features, analyzing the facial contour and the symmetry of the face, and generating a facial feature data set.

4. The smart lock unlocking method based on face image processing according to claim 1, characterized in that: Based on the facial feature dataset, the difference of facial images of each view is calculated, the stability and richness of features in multiple view images are evaluated, weights are assigned to multiple datasets and fused, and the steps of generating dataset fusion results are specifically as follows: S201: calling the facial feature data set, and calculating the difference of the facial image at each viewing angle according to the facial image features at multiple viewing angles, to obtain viewing angle difference data; S202: Based on the perspective difference data, evaluating the stability and richness of features in the facial images of multiple perspectives, and generating a feature stability evaluation result; S203: According to the feature stability evaluation result, fusion weights are assigned to the feature data sets of multiple viewpoint images, and the multiple data sets are fused to generate a data set fusion result.

5. The smart lock unlocking method based on face image processing according to claim 1 is characterized in that: The steps of extracting facial geometric information through the data set fusion result, using facial images from multiple perspectives, calculating depth information at multiple positions of the face, and generating facial three-dimensional feature data are as follows: S301: Based on the data set fusion result, the depth information of multiple positions of the face is calculated according to the facial images of multiple perspectives to generate facial depth data; S302: Calculating curvature information of multiple facial positions based on the facial depth data to generate facial curvature data; S303: Identify multiple convex and concave areas in the user's face based on the facial curvature data, and generate facial three-dimensional feature data.

6. The smart lock unlocking method based on face image processing according to claim 1 is characterized in that: The steps of calling the facial three-dimensional feature data, combining with the light sensor, monitoring the ambient light intensity in real time, analyzing the face recognition accuracy under various light intensities, setting the light tolerance threshold and sending the rejection signal, and generating the rejection signal judgment result are as follows: S401: Acquire the facial three-dimensional feature data, and collect ambient light intensity data in real time in combination with a light sensor to generate ambient light data; S402: Calculating face recognition accuracy under various light intensities based on the ambient light data, analyzing recognition effects under various light conditions, and generating light recognition accuracy data; S403: According to the illumination recognition accuracy data, an illumination tolerance threshold is set and compared with the real-time illumination intensity, and a rejection signal is sent when the illumination condition does not meet the image recognition requirement, thereby generating a rejection signal judgment result.

7. The smart lock unlocking method based on face image processing according to claim 6 is characterized in that: The specific formula for calculating the face recognition accuracy under various light intensities is: Calculate recognition accuracy; Among them, P(I) represents the recognition accuracy under the current light intensity I in the face recognition test, N correct Represents the number of correct face recognition times under this light intensity, N total represents the total number of face recognitions performed under this light intensity, I is the current experimental light intensity, and I ideal represents the optimal ideal illumination intensity, α is the weight coefficient used to adjust the influence of illumination error on recognition accuracy, β is the coefficient used to adjust the influence of illumination intensity change on recognition accuracy, and I j is the light intensity value tested during the experiment, I j-1 is the previous light intensity value, m represents the number of light intensity data points, and j represents the index of the light intensity data test.

8. The smart lock unlocking method based on face image processing according to claim 1, characterized in that: Based on the rejection signal judgment result, the matching degree of multiple local feature areas, face contours and three-dimensional structural features in the image is analyzed, the matching score of the image is calculated, and a control instruction is sent to the smart lock to generate the smart lock control parameters. Specifically, the steps are as follows: S501: Based on the rejection signal judgment result, according to multiple local area features in the user's face image, calculate the matching degree of multiple area feature values ​​and preset feature reference values ​​to generate a local feature matching degree value; S502: calling the local feature matching value, analyzing the key point distribution of the face contour in the image, combining the three-dimensional structure feature, calculating the matching degree of the face contour and the three-dimensional structure feature, and generating a contour and structure matching value; S503: Call the contour and structure matching value, calculate the matching score of the face image, adjust the control parameters of the door lock, and generate smart lock control parameters.

9. The smart lock unlocking method based on face image processing according to claim 8, characterized in that: The specific formula for calculating the matching score of the face image is: Calculate the matching score, adjust the door lock control parameters, and generate smart lock control parameters; Among them, M 89 represents the matching degree of the i′th local area, C face Represents the facial contour matching degree, C 3D represents the matching degree of three-dimensional structural features, w 89 represents the weight of the i′th local feature matching, w1 represents the weight of the local feature matching, w2 represents the weight of the face contour matching, and w < represents the weight of the three-dimensional structural feature matching, i′ represents the number of the local area involved in the local feature matching, n′ represents the total number of local features, and S′ represents the matching score.

10. The intelligent lock unlocking system based on face image processing is characterized by: According to the smart lock unlocking method based on face image processing according to any one of claims 1 to 9, the system comprises: The multi-angle image acquisition module uses image acquisition equipment to capture user facial images from multiple directions, extract image features, and construct a facial feature data set; The feature weight optimization module calls the facial feature dataset, analyzes the differences of features under multiple viewing angles, evaluates the consistency and richness of features, performs weight assignment and data fusion on multiple datasets, and forms a dataset fusion result; The facial depth analysis module extracts facial geometric structure information from multi-angle images based on the data set fusion results, calculates the depth values ​​of multiple facial regions, and obtains facial three-dimensional feature data; The ambient light adaptation module uses the facial three-dimensional feature data and combines with the light sensor to monitor the ambient light conditions, evaluates the impact of various light intensities on the face recognition effect, sets the light tolerance threshold, and generates a rejection signal judgment result; The unlocking decision execution module evaluates the matching degree of local facial features, contours, and three-dimensional structures according to the judgment result of the rejection signal, calculates the matching score, sends a control instruction to the smart lock, and generates smart lock control parameters.

Images