Intelligent campus identity management system and method based on face recognition

Abstract

The invention belongs to the technical field of face recognition, and discloses an intelligent campus identity management system and method based on face recognition, and the method comprises the steps: collecting a face image sequence and corresponding identity information, carrying out the dynamic feature tracking of the face image sequence, and obtaining a feature marking graph sequence; performing skin state analysis on the feature marking graph sequence to obtain a skin state feature set; performing expression analysis on the skin state feature set to obtain static features and micro-expression features, and performing micro-deformation compensation on the micro-expression features to obtain micro-motion compensation features; an identity recognition model is constructed based on the compensation features, the static features and the identity information, and accurate recognition of the personnel is achieved based on the identity recognition model; and the accuracy and efficiency of identity recognition are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of face recognition technology, and more specifically, to a smart campus identity management system and method based on face recognition. Background Art

[0002] With the rapid development of artificial intelligence and biometric technology, face recognition has been widely used in security verification, identity recognition and behavior analysis. In practical applications, traditional face recognition systems still face a series of challenges. For example, in dynamic scenes such as talking and laughing, due to the continuous deformation of facial muscles, feature points are prone to drift. In this case, it is difficult for traditional face recognition systems to accurately extract facial features, resulting in feature point matching errors, thus affecting the accuracy of identity recognition. In addition, for special cases such as local scars and temporary skin lesions on the face, the adaptability of existing systems is poor. Abnormal facial deformation (such as acne, scars or other short-term skin problems) often significantly reduces the accuracy of feature point extraction and matching, which may cause misidentification or rejection in large-scale campus environments. Most existing face recognition algorithms are based on feature extraction of static images, and rarely consider the influence of dynamic factors such as expressions and postures. However, in practical applications, changes in facial posture, closing eyes, squinting, smiling and other expression changes will introduce errors in the recognition process, especially in complex environments, which can easily lead to a significant decrease in system recognition stability.

[0003] In view of this, the present invention proposes a smart campus identity management system and method based on face recognition to solve the above problems. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a smart campus identity management system based on face recognition, comprising: Data acquisition and processing module: collects facial image sequences and corresponding identity information, performs dynamic feature tracking on facial image sequences, and obtains feature marker image sequences; Skin analysis module: performs skin condition analysis on the feature marker image sequence to obtain a skin condition feature set; Expression analysis module: including a modality analysis unit and a modality compensation unit. The modality analysis unit performs expression analysis on the skin state feature set to obtain static features and micro-expression features. The modality compensation unit performs micro-deformation compensation on the micro-expression features to obtain micro-motion compensation features. Model building module: build an identity recognition model based on compensation features, static features and identity information, and realize accurate identification of personnel based on the identity recognition model.

[0005] Furthermore, the facial image sequence is collected in the following manner: A high-resolution camera is used in conjunction with a high-speed camera. The camera is equipped with an automatic focus function and is set to high frame rate acquisition. The acquisition frame rate is 60fps or higher. The acquisition time is one minute for people who appear in the camera's field of view, and image acquisition is performed at an interval of one frame per second. All acquired images are sorted from early to late based on the acquisition time to obtain a facial image sequence.

[0006] Furthermore, the method of performing dynamic feature tracking on the facial image sequence includes: Preset standard pixel coordinates, and mark the coordinates of the pixels of each frame in the face image sequence based on the standard pixel coordinates to obtain pixel coordinates; use the first frame in the face image sequence as the starting tracking image, and use the other images in the face image sequence except the first frame as the feature exploration image, and use the facial feature point detection algorithm to mark the feature point serial number of the starting tracking image to obtain a marked feature image; use the marked pixels in the marked feature image as contour feature points, preset a displacement threshold, and the displacement threshold is an integer greater than zero, use the pixels at the same position as the pixel coordinates of the contour feature point in the feature exploration image as the center of the circle, and use the displacement threshold as the radius to draw a circle, use the pixels within the circle as feature track pixels, and perform feature evaluation on each feature track pixel to obtain a feature evaluation value; use the feature track pixel with the largest feature evaluation value as the contour pixel, and encode and mark the contour pixel with the code marked in the contour feature point to obtain a search mark image, and all search mark images and the starting tracking image constitute a feature mark image sequence.

[0007] Furthermore, the formula for evaluating the feature of each feature track pixel is: ;in, represents the feature evaluation value, represents the pixel value of the contour feature point, represents the pixel value of the feature trajectory pixel, represents the pixel value of the feature trajectory pixel, The horizontal axis value represents the pixel coordinate of the contour feature point. The vertical axis value represents the pixel coordinate of the contour feature point. The horizontal axis value represents the pixel coordinate of the feature trajectory pixel, The vertical axis value representing the pixel coordinate of the feature trajectory pixel, Represents the displacement threshold.

[0008] Furthermore, the method of performing skin condition analysis on the characteristic marker sequence includes: The marked pixels in each frame image in the feature marking image sequence are used as pixel analysis points, a filter group is preset, the filter size, filter direction angle, filter wavelength, filter bandwidth and filter phase of each filter are initialized, and texture evaluation is performed on the pixel analysis point based on the filter group to obtain skin texture value; a gradient analysis window is preset, the pixel analysis point is used as the center of the gradient analysis window, a gradient operator is used to perform gradient analysis on the pixels in the gradient analysis window to obtain the window pixel gradient, and the maximum and minimum normalization algorithm is used to normalize the mean of the window pixel gradient to obtain the pixel elasticity coefficient; the skin texture value, pixel grayscale and pixel elasticity coefficient of the pixel analysis point constitute a skin state description, and skin state mapping is performed on the skin state description to obtain the skin state value; the skin state values ​​of all pixel analysis points in the same image constitute a skin feature matrix, and all skin feature matrices constitute a skin state feature set.

[0009] Furthermore, the method of performing expression analysis on the skin state feature set includes: Each skin feature matrix in the skin state feature set is taken as the group to be analyzed, and the skin state values ​​in the group to be analyzed are clustered by a cluster analysis algorithm to obtain skin state clusters; the skin state cluster with the largest mean of skin state values ​​is taken as the state center, and the skin homogeneity and average pixel coordinates of the state center are taken as the evaluation center, the skin homogeneity is the mean of the skin state values ​​in the state center, and the average pixel coordinates are the mean of the pixel coordinates corresponding to each skin state value in the state center; based on the evaluation center, a local deformation evaluation is performed on the group to be analyzed to obtain a micro-motion deformation field; all micro-motion deformation fields of the same group to be analyzed constitute an expression description; a deformation threshold is preset, and the expression description whose mean of the micro-motion deformation field in the expression description is greater than or equal to the deformation threshold is taken as a micro-expression feature, and the expression description whose mean of the micro-motion deformation field in the expression description is less than the deformation threshold is taken as a static feature.

[0010] Furthermore, the formula for evaluating the local deformation of the group to be analyzed is: ;in, Represents the first The micro-deformation field of the skin state value, Represents the first Skin condition value, Represents the uniformity of the skin at the center of the state, Represents the coordinate distance of the state center, which is the modulus of the average pixel coordinate of the state center. represents the equilibrium parameter, Represents the first The pixel coordinates corresponding to the skin state value, represents the average pixel coordinate of the state center, represents the spatial parameter, Represents the distance measurement function.

[0011] Furthermore, the method of performing micro-deformation compensation on micro-expression features includes: A standard face image is collected, and the left eye corner, right eye corner, left mouth corner, right mouth corner and nose tip center in the standard face image are used as deformation base points. The standard face image is marked with standard pixel coordinates, and the pixel coordinates of the deformation base points are recorded. All deformation base points constitute a base point set. Each group of micro-expression features is used as a group to be compensated. The group to be compensated is evaluated based on the base point set to obtain micro-motion compensation. The micro-motion compensation corresponding to the micro-motion deformation field in the same group to be compensated constitutes a micro-motion compensation vector, and all micro-motion compensation vectors constitute micro-motion compensation features.

[0012] Furthermore, the formula for evaluating compensation for the group to be compensated is: ;in, Represents the group to be compensated Micro-motion compensation of a micro-motion deformation field, Represents the group to be compensated A micro-deformation field, represents the size of the base point set, Representative base point concentration The deformation weight of the deformation base point, represents the distance measurement function, Represents the group to be compensated The pixel coordinates of the pixels corresponding to the micro-deformation field, Representative base point concentration The pixel coordinates of the deformation base points, represents the grayscale control parameter, Representative The grayscale gradient of the pixel corresponding to each deformation base point, Represents the group to be compensated The gray gradient of the pixel corresponding to the micro-deformation field, Represents the group to be compensated The micromotion gradient of the micromotion deformation field.

[0013] A smart campus identity management method based on face recognition, comprising: S1, collecting a facial image sequence and corresponding identity information, performing dynamic feature tracking on the facial image sequence, and obtaining a feature marker sequence; S2, performing skin condition analysis on the feature marker image sequence to obtain a skin condition feature set; S3, performing expression analysis on the skin state feature set to obtain static features and micro-expression features, performing micro-deformation compensation on the micro-expression features to obtain micro-motion compensation features; S4. Construct an identity recognition model based on compensation features, static features and identity information, and achieve accurate identification of personnel based on the identity recognition model.

[0014] The technical effects and advantages of the smart campus identity management system and method based on face recognition of the present invention are as follows: The present invention solves the shortcomings of traditional static feature point detection methods in feature point drift by performing dynamic feature tracking on facial image sequences, and realizes accurate evaluation of feature trajectory pixels of facial image sequences, thereby ensuring the consistency and accuracy of feature point trajectories; by performing skin condition analysis on feature marker image sequences, temporary changes such as facial swelling or eye bags caused by factors such as fatigue and staying up late can be captured, so that the system can still maintain a high recognition accuracy when dealing with non-persistent facial changes; through the expression analysis module, the system can recognize emotional expressions such as subtle eye movements or changes in mouth corners, effectively reducing recognition errors caused by expression changes, and improving the stability of the system in dynamic expressions and complex scenes; by constructing an identity recognition model, not only the recognition ability of the system is enhanced, but also the robustness and adaptability of the system are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a smart campus identity management system based on face recognition according to the present invention; Figure 2 This is a schematic diagram of a smart campus identity management method based on face recognition according to the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Example 1 See also Figure 1 As shown, this embodiment is a smart campus identity management system based on face recognition, including: Data acquisition and processing module: collects facial image sequences and corresponding identity information, performs dynamic feature tracking on facial image sequences, and obtains feature marker image sequences; Skin analysis module: performs skin condition analysis on the feature marker image sequence to obtain a skin condition feature set; Expression analysis module: including a modality analysis unit and a modality compensation unit. The modality analysis unit performs expression analysis on the skin state feature set to obtain static features and micro-expression features. The modality compensation unit performs micro-deformation compensation on the micro-expression features to obtain micro-motion compensation features. Model building module: builds an identity recognition model based on compensation features, static features and identity information, and realizes accurate identification of personnel based on the identity recognition model; Each module is connected by wired and / or wireless means to achieve data transmission between modules; The collection methods of face image sequences include: A high-resolution camera is used in conjunction with a high-speed camera to ensure that rapidly changing facial expressions and gestures can be captured. The camera is set to high frame rate acquisition, with an acquisition frame rate of 60fps or higher, to ensure that sufficient data can be captured even in fast dynamic scenes. The camera is equipped with an automatic focus function to ensure that the camera can always focus clearly even during movement, ensuring that facial details are not distorted. For people who appear in the camera's field of view, the acquisition time is one minute, and image acquisition is performed at an interval of one frame per second. All acquired images are sorted from early to late based on the acquisition time to obtain a facial image sequence.

[0018] In the case of rapid changes in facial posture (such as looking up, lowering, turning the head) or dynamic scenes (such as continuous talking, laughing), facial feature points are prone to drift. Traditional static feature point detection methods are difficult to track the changes of feature points in real time. Dynamic feature tracking of facial image sequences solves the problem of feature point drift caused by rapid posture changes and facial expression movements, ensuring the consistency and accuracy of feature point trajectories, thereby providing stable input for subsequent processing. Specifically: The standard pixel coordinates are preset, and the horizontal and vertical coordinates in the standard pixel coordinates are both marked with one pixel as the coordinate scale; based on the standard pixel coordinates, the pixels of each frame in the face image sequence are marked with the lower left corner of each frame in the face image sequence as the mapping point of the coordinate origin in the standard pixel coordinates, and the horizontal axis value of the coordinate is increased by one for each pixel added to the right, and the vertical axis value of the coordinate is increased by one for each pixel added to the left, so as to obtain the pixel coordinates; the first frame in the face image sequence is used as the starting tracking map, and the images other than the first frame in the face image sequence are used as feature exploration maps, and the facial feature point detection algorithm is used to mark the feature point serial number of the starting tracking map to obtain the marked feature point. Feature map, common facial feature point detection algorithms include Dlib68 point facial marking method and OpenCV facial feature point detection algorithm. The marked feature map contains the marking of important points (i.e., pixel points) of the face contour, and the marking is accompanied by the encoding of the feature point pixels to distinguish the feature points; the marked pixels in the marked feature map are used as contour feature points, and the displacement threshold is preset. The displacement threshold is an integer greater than zero. The pixel at the same position as the pixel coordinate of the contour feature point in the feature exploration map is used as the center of the circle, and the displacement threshold is used as the radius to draw a circle. The pixels in the circle are used as feature track pixels. The feature evaluation is performed on each feature track pixel. The formula for evaluating the feature of each feature track pixel is as follows: ;in, represents the feature evaluation value, represents the pixel value of the contour feature point, represents the pixel value of the feature trajectory pixel, represents the pixel value of the feature trajectory pixel, The horizontal axis value represents the pixel coordinate of the contour feature point. The vertical axis value represents the pixel coordinate of the contour feature point. The horizontal axis value represents the pixel coordinate of the feature trajectory pixel, The vertical axis value representing the pixel coordinate of the feature trajectory pixel, represents the displacement threshold; the feature trajectory pixel with the largest feature evaluation value is used as the contour pixel, and the contour pixel is encoded and marked with the code marked in the contour feature point to obtain the exploration mark map. All exploration mark maps and the starting tracking map constitute the feature mark map sequence.

[0019] Traditional face recognition systems have poor recognition effects when dealing with temporary facial changes such as facial swelling and eye bags caused by lack of sleep, staying up late, fatigue, etc. The reason is that facial swelling and eye bags will cause distortion of facial features, making it difficult for the system to match accurate identity information. By performing skin state analysis on the feature marker sequence, changes in skin features caused by facial swelling or other physiological changes can be captured, allowing the system to adapt to these temporary changes instead of relying on the original static image features. Specifically: The marked pixels in each frame of the feature mark image sequence are used as pixel analysis points, a filter group is preset, the filter size, filter direction angle, filter wavelength, filter bandwidth and filter phase of each filter are initialized, and texture evaluation is performed on the pixel analysis points based on the filter group. The formula for texture evaluation of the pixel analysis points is: ;in, Represents the skin texture value, represents the pixel coordinates of the pixel analysis point, Representative filters, represents the filter convolution operation, Represents the number of filters, which is used to multiply and sum the filter with the local area of ​​the image divided by the local window centered on the pixel coordinates of the pixel analysis point and the filter size, and output the texture response in a specific direction; commonly used filters include Gabor filters, which can extract directional characteristics of textures, such as the fineness and texture direction of the skin; A gradient analysis window is preset, and the pixel analysis point is taken as the center of the gradient analysis window. A gradient operator is used to perform gradient analysis on the pixels in the gradient analysis window to obtain the window pixel gradient. The maximum and minimum normalization algorithm is used to normalize the mean of the window pixel gradient to obtain the pixel elasticity coefficient. Commonly used gradient operators include the Sobel operator and the Prewitt operator. The skin texture value, pixel grayscale and pixel elasticity coefficient of the pixel analysis point constitute the skin state description, and the skin state description is mapped to the skin state. The formula for mapping the skin state description to the skin state is: ;in, Represents the skin condition value, represents the state weight matrix, Represents the skin condition description, Represents the bias parameter, which is used to adjust the balance of the mapping. represents the hyperbolic tangent function; the skin condition values ​​of all pixel analysis points in the same image constitute a skin feature matrix, and all skin feature matrices constitute a skin condition feature set.

[0020] Micro-expressions (such as slight eye movements, slightly raised corners of the mouth, etc.) are relatively hidden and rapid changes for face recognition systems. Traditional recognition technology is difficult to accurately capture these subtle facial changes. Micro-expressions reflect the emotional state and physiological changes of an individual, and these subtle changes will affect the accuracy of the recognition system. By performing expression analysis on the skin state feature set and extracting micro-expression features, the system can better adapt to the rapid dynamic changes of the face, especially when the user's expression changes greatly. Specifically: Each skin feature matrix in the skin state feature set is taken as the group to be analyzed, and the skin state values ​​in the group to be analyzed are clustered using a cluster analysis algorithm to obtain skin state clusters. Commonly used cluster analysis algorithms include K-Means clustering algorithm and hierarchical clustering algorithm; the skin state cluster with the largest mean of skin state values ​​is taken as the state center, and the skin homogeneity and average pixel coordinates of the state center are taken as the evaluation center. The skin homogeneity is the mean of the skin state values ​​in the state center, and the average pixel coordinates are the mean of the pixel coordinates corresponding to each skin state value in the state center; local deformation evaluation is performed on the group to be analyzed based on the evaluation center, and the formula for local deformation evaluation is: ;in, Represents the first The micro-deformation field of the skin state value, Represents the first Skin condition value, Represents the uniformity of the skin at the center of the state, Represents the coordinate distance of the state center, which is the modulus of the average pixel coordinate of the state center. Represents the balance parameter, which is used to control the balance between the difference in skin state values ​​and the spatial distance. Represents the first The pixel coordinates corresponding to the skin state value, represents the average pixel coordinate of the state center, Represents the spatial parameter, which is used to balance the effect of spatial distance on the micro-deformation field. represents a distance measurement function, and common distance measurement functions include the Euclidean distance function and the Manhattan distance function; all micro-motion deformation fields of the same group to be analyzed constitute an expression description; a deformation threshold is preset, and the deformation threshold is set by technical personnel in this field according to actual conditions, and the expression description whose mean value of the micro-motion deformation field in the expression description is greater than or equal to the deformation threshold is taken as a micro-expression feature, and the expression description whose mean value of the micro-motion deformation field in the expression description is less than the deformation threshold is taken as a static feature.

[0021] Rapid deformation of facial muscles and changes in expression will cause the position of feature points to drift, affecting the stability and accuracy of facial recognition, especially in situations where there are changes in expression or dynamic scenes, where the geometric shape of the face changes. By performing micro-deformation compensation on micro-expression features, the changes in facial shape can be adjusted at multiple levels to reduce the impact of expression changes on recognition results, so that the system can stably handle facial deformation in dynamic environments and more accurately reflect the actual facial state. Specifically: Collect standard face images. The standard face images are grayscale images with uniform face position, scale and orientation after face detection and alignment processing, which reduces the interference of color on recognition and retains the shape and structural information of the face. The size and resolution of the standard face images are consistent with those in the face image sequence; the left eye corner, right eye corner, left mouth corner, right mouth corner and nose tip center in the standard face image are used as deformation base points, the standard face image is marked with standard pixel coordinates, and the pixel coordinates of the deformation base points are recorded. All deformation base points constitute a base point set, and each group of micro-expression features is used as a group to be compensated. Based on the base point set, compensation evaluation is performed on the group to be compensated. The formula for compensation evaluation of the group to be compensated is as follows: ;in, Represents the group to be compensated Micro-motion compensation of a micro-motion deformation field, Represents the group to be compensated A micro-deformation field, represents the size of the base point set, Representative base point concentration The deformation weight of the deformation base point, Represents the distance measurement function. Common distance measurement functions include Euclidean distance function and Manhattan distance function. Represents the group to be compensated The pixel coordinates of the pixels corresponding to the micro-deformation field, Representative base point concentration The pixel coordinates of the deformation base points, Represents the grayscale control parameter, which is used to control the sensitivity of grayscale contrast. Representative The grayscale gradient of the pixel corresponding to each deformation base point, Represents the group to be compensated The gray gradient of the pixel corresponding to the micro-deformation field, Represents the group to be compensated The micro-motion gradient of each micro-motion deformation field; the micro-motion compensation corresponding to the micro-motion deformation field in the same group to be compensated constitutes a micro-motion compensation vector, and all micro-motion compensation vectors constitute a micro-motion compensation feature.

[0022] Based on compensating features, static features and identity information, the CNN model is used as the initial model of the identity recognition model, the compensating features, static features and identity information are used as training data, the training data is used as a training sample set, the CNN model is trained using the training sample set, the compensating features, static features and identity information are used as input data of the identity recognition model, and the predicted identity label is used as output data of the identity recognition model; minimizing the error between the actual identity information and the identity label predicted by the identity recognition model is used as the training goal, the recall rate function is used as the loss function of the identity recognition model, and when the loss function converges, the training is stopped to obtain the identity recognition model.

[0023] This embodiment solves the shortcomings of traditional static feature point detection methods in feature point drift by performing dynamic feature tracking on facial image sequences, and realizes accurate evaluation of feature trajectory pixels of facial image sequences, thereby ensuring the consistency and accuracy of feature point trajectories; by performing skin condition analysis on feature marker image sequences, it is able to capture temporary changes such as facial swelling or eye bags caused by factors such as fatigue and staying up late, so that the system can still maintain a high recognition accuracy when dealing with non-persistent facial changes; through the expression analysis module, the system can recognize emotional expressions such as subtle eye movements or changes in mouth corners, effectively reducing recognition errors caused by expression changes, and improving the stability of the system in dynamic expressions and complex scenes; by constructing an identity recognition model, not only the recognition ability of the system is enhanced, but also the robustness and adaptability of the system are improved.

[0024] Example 2 See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a smart campus identity management method based on face recognition is provided, including: S1, collecting a facial image sequence and corresponding identity information, performing dynamic feature tracking on the facial image sequence, and obtaining a feature marker sequence; S2, performing skin condition analysis on the feature marker image sequence to obtain a skin condition feature set; S3, performing expression analysis on the skin state feature set to obtain static features and micro-expression features, performing micro-deformation compensation on the micro-expression features to obtain micro-motion compensation features; S4. Construct an identity recognition model based on compensation features, static features and identity information, and achieve accurate identification of personnel based on the identity recognition model.

[0025] Example 3 This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the smart campus identity management method based on face recognition provided above is implemented.

[0026] Since the electronic device introduced in this embodiment is an electronic device used to implement a smart campus identity management method based on face recognition in the embodiment of this application, based on the smart campus identity management method based on face recognition introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not introduced in detail here. As long as the technical personnel of this field implement the electronic device used in the smart campus identity management method based on face recognition in the embodiment of this application, it belongs to the scope of protection of this application.

[0027] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0028] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A smart campus identity management system based on face recognition, characterized in that: include: Data acquisition and processing module: collects facial image sequences and corresponding identity information, performs dynamic feature tracking on facial image sequences, and obtains feature marker image sequences; Skin analysis module: performs skin condition analysis on the feature marker image sequence to obtain a skin condition feature set; Expression analysis module: including a modality analysis unit and a modality compensation unit. The modality analysis unit performs expression analysis on the skin state feature set to obtain static features and micro-expression features. The modality compensation unit performs micro-deformation compensation on the micro-expression features to obtain micro-motion compensation features. Model building module: build an identity recognition model based on compensation features, static features and identity information, and realize accurate identification of personnel based on the identity recognition model.

2. The smart campus identity management system based on face recognition according to claim 1 is characterized in that: The acquisition method of the face image sequence includes: A high-resolution camera is used in conjunction with a high-speed camera. The camera is equipped with an automatic focus function and is set to high frame rate acquisition. The acquisition frame rate is 60fps or higher. The acquisition time is one minute for people who appear in the camera's field of view, and image acquisition is performed at an interval of one frame per second. All acquired images are sorted from early to late based on the acquisition time to obtain a facial image sequence.

3. The smart campus identity management system based on face recognition according to claim 2 is characterized in that: The method of performing dynamic feature tracking on a facial image sequence includes: Preset standard pixel coordinates, and mark the coordinates of the pixels of each frame in the face image sequence based on the standard pixel coordinates to obtain pixel coordinates; use the first frame in the face image sequence as the starting tracking image, and use the other images in the face image sequence except the first frame as the feature exploration image, and use the facial feature point detection algorithm to mark the feature point serial number of the starting tracking image to obtain a marked feature image; use the marked pixels in the marked feature image as contour feature points, preset a displacement threshold, and the displacement threshold is an integer greater than zero, use the pixels at the same position as the pixel coordinates of the contour feature point in the feature exploration image as the center of the circle, and use the displacement threshold as the radius to draw a circle, use the pixels within the circle as feature track pixels, and perform feature evaluation on each feature track pixel to obtain a feature evaluation value; use the feature track pixel with the largest feature evaluation value as the contour pixel, and encode and mark the contour pixel with the code marked in the contour feature point to obtain a search mark image, and all search mark images and the starting tracking image constitute a feature mark image sequence.

4. The smart campus identity management system based on face recognition according to claim 3 is characterized in that: The formula for evaluating the feature of each feature track pixel is: ;in, represents the feature evaluation value, represents the pixel value of the contour feature point, represents the pixel value of the feature trajectory pixel, represents the pixel value of the feature trajectory pixel, The horizontal axis value represents the pixel coordinate of the contour feature point. The vertical axis value represents the pixel coordinate of the contour feature point. The horizontal axis value represents the pixel coordinate of the feature trajectory pixel, The vertical axis value representing the pixel coordinate of the feature trajectory pixel, Represents the displacement threshold.

5. The smart campus identity management system based on face recognition according to claim 4 is characterized in that: The method of performing skin condition analysis on the characteristic marker sequence comprises: The marked pixels in each frame image in the feature marking image sequence are used as pixel analysis points, a filter group is preset, the filter size, filter direction angle, filter wavelength, filter bandwidth and filter phase of each filter are initialized, and texture evaluation is performed on the pixel analysis point based on the filter group to obtain skin texture value; a gradient analysis window is preset, the pixel analysis point is used as the center of the gradient analysis window, a gradient operator is used to perform gradient analysis on the pixels in the gradient analysis window to obtain the window pixel gradient, and the maximum and minimum normalization algorithm is used to normalize the mean of the window pixel gradient to obtain the pixel elasticity coefficient; the skin texture value, pixel grayscale and pixel elasticity coefficient of the pixel analysis point constitute a skin state description, and skin state mapping is performed on the skin state description to obtain the skin state value; the skin state values ​​of all pixel analysis points in the same image constitute a skin feature matrix, and all skin feature matrices constitute a skin state feature set.

6. The smart campus identity management system based on face recognition according to claim 5 is characterized in that: The method of performing expression analysis on the skin state feature set includes: Each skin feature matrix in the skin state feature set is taken as the group to be analyzed, and the skin state values ​​in the group to be analyzed are clustered by a cluster analysis algorithm to obtain skin state clusters; the skin state cluster with the largest mean of skin state values ​​is taken as the state center, and the skin homogeneity and average pixel coordinates of the state center are taken as the evaluation center, the skin homogeneity is the mean of the skin state values ​​in the state center, and the average pixel coordinates are the mean of the pixel coordinates corresponding to each skin state value in the state center; based on the evaluation center, a local deformation evaluation is performed on the group to be analyzed to obtain a micro-motion deformation field; all micro-motion deformation fields of the same group to be analyzed constitute an expression description; a deformation threshold is preset, and the expression description whose mean of the micro-motion deformation field in the expression description is greater than or equal to the deformation threshold is taken as a micro-expression feature, and the expression description whose mean of the micro-motion deformation field in the expression description is less than the deformation threshold is taken as a static feature.

7. The smart campus identity management system based on face recognition according to claim 6 is characterized in that: The formula for evaluating the local deformation of the group to be analyzed is: ;in, Represents the first The micro-deformation field of the skin state value, Represents the first Skin condition value, Represents the uniformity of the skin at the center of the state, Represents the coordinate distance of the state center, which is the modulus of the average pixel coordinate of the state center. represents the equilibrium parameter, Represents the first The pixel coordinates corresponding to the skin state value, represents the average pixel coordinate of the state center, represents the spatial parameter, Represents the distance measurement function.

8. The smart campus identity management system based on face recognition according to claim 7 is characterized in that: The method of performing micro deformation compensation on micro expression features includes: A standard face image is collected, and the left eye corner, right eye corner, left mouth corner, right mouth corner and nose tip center in the standard face image are used as deformation base points. The standard face image is marked with standard pixel coordinates, and the pixel coordinates of the deformation base points are recorded. All deformation base points constitute a base point set. Each group of micro-expression features is used as a group to be compensated. The group to be compensated is evaluated based on the base point set to obtain micro-motion compensation. The micro-motion compensation corresponding to the micro-motion deformation field in the same group to be compensated constitutes a micro-motion compensation vector, and all micro-motion compensation vectors constitute micro-motion compensation features.

9. The smart campus identity management system based on face recognition according to claim 8 is characterized in that: The formula for evaluating compensation for the group to be compensated is: ;in, Represents the group to be compensated Micro-motion compensation of a micro-motion deformation field, Represents the group to be compensated A micro-deformation field, represents the size of the base point set, Representative base point concentration The deformation weight of the deformation base point, represents the distance measurement function, Represents the group to be compensated The pixel coordinates of the pixels corresponding to the micro-deformation field, Representative base point concentration The pixel coordinates of the deformation base points, represents the grayscale control parameter, Representative The grayscale gradient of the pixel corresponding to each deformation base point, Represents the group to be compensated The gray gradient of the pixel corresponding to the micro-deformation field, Represents the group to be compensated The micromotion gradient of the micromotion deformation field.

10. A smart campus identity management method based on face recognition, which is implemented based on the smart campus identity management system based on face recognition according to any one of claims 1 to 9, characterized in that: include: S1, collecting a facial image sequence and corresponding identity information, performing dynamic feature tracking on the facial image sequence, and obtaining a feature marker sequence; S2, performing skin condition analysis on the feature marker image sequence to obtain a skin condition feature set; S3, performing expression analysis on the skin state feature set to obtain static features and micro-expression features, performing micro-deformation compensation on the micro-expression features to obtain micro-motion compensation features; S4. Construct an identity recognition model based on compensation features, static features and identity information, and achieve accurate identification of personnel based on the identity recognition model.

Images