Intelligent office management method and system based on face recognition function

Abstract

The invention discloses an intelligent office management method and system based on a face recognition function, and particularly relates to the technical field of computer vision, and the method comprises the following steps: collecting face image data of a user, and carrying out the feature extraction and analysis based on the collected image data, the credibility of the current identity verification result is obtained according to the feature analysis result; when a preset re-recognition mechanism is triggered, performing identity verification again to obtain a feature analysis result, the credibility of an identity verification result and the number of times of performing identity verification again under a preset time window, performing optimization analysis, and determining whether to optimize the current camera collection environment condition according to an optimization analysis result; by analyzing the environmental condition influence index and the facial posture influence index, the rationality of the environment collected by the camera can be evaluated in real time, optimization is carried out when needed, and a guarantee is provided for the accuracy and stability of recognition.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer vision, and more particularly, to an intelligent office management method and system based on face recognition function. BACKGROUND

[0002] With the rapid development of information technology, intelligent office management systems have gradually become widely used technologies in enterprises and public institutions. Especially in the field of face recognition, with the continuous progress of deep learning and computer vision technology, face recognition as an important identity verification method has been widely applied. Through face recognition technology, the system can quickly and accurately verify the identity of personnel, improving the security and convenience of the office environment.

[0003] Existing face recognition systems still face some challenges in actual application. Especially in complex environments (such as insufficient light, user face occlusion, posture change, etc.), the decline in recognition accuracy often leads to a decrease in system recognition rate, affecting the stability and security of the system. How to maintain high recognition accuracy in these complex environmental conditions and dynamically adapt to environmental changes is a major problem. Therefore, the present application proposes an intelligent office management method and system based on face recognition function to solve the above problems. SUMMARY

[0004] To achieve the above purpose, the present application provides the following technical scheme: An intelligent office management method based on face recognition function, comprising the following steps: Collecting the face image data of the user through the camera and using the preset face recognition algorithm to perform identity verification on the collected image data, recognizing and recording the identity information of the user; Based on the collected image data, performing feature extraction and analysis, and then inputting the feature analysis result into the pre-trained machine learning model to output the credibility of the current identity verification result; When the credibility of the current identity verification result triggers the preset re-recognition mechanism, re-performing identity verification; Obtaining the feature analysis result, the credibility of the identity verification result, and the number of times of re-performing identity verification in a preset time window, and performing optimization analysis, and deciding whether to optimize the current camera collection environment condition according to the optimization analysis result.

[0005] In a preferred embodiment, feature extraction refers to extracting the collection environment features when the camera collects the face image data of the user to obtain an environment feature group, and extracting the user face features when the camera collects the face image data of the user to obtain a face feature group.

[0006] In a preferred embodiment, the feature analysis refers to: performing environment analysis based on the environment feature group to generate an environment condition influence index representing the influence of the lighting condition of the image on the recognition accuracy, and performing face analysis based on the face feature group to generate a face posture influence index representing the influence of the face condition of the user on the recognition accuracy.

[0007] In a preferred embodiment, the acquisition logic of the environment condition influence index is: The environment feature group comprises a lighting intensity feature, a contrast feature, a shadow feature, and a backlight feature, and the environment condition influence index calculation formula is: ; is the environment feature type corresponding to a preset non-zero proportion coefficient, is the environment feature type corresponding to an influence function, the environment feature type is 1, 2, 3, or 4, respectively corresponding to the lighting intensity feature, the contrast feature, the shadow feature, and the backlight feature, is the environment condition influence index; is the environment feature type corresponding to an influence function, the environment feature type ; represents the average brightness of the image, represents the contrast between the face and the background, represents a preset optimal contrast critical value, represents a preset contrast sensitivity adjustment coefficient, represents the pixel ratio of the area of the shadow region in the image to the area of the entire image region, represents a preset non-zero constant one, represents the proportion value of the area of the backlight region in the image to the area of the entire image region.

[0008] In a preferred embodiment, the acquisition logic of the face posture influence index is: The face feature group comprises an angle deviation feature, an occlusion degree feature, and an expression change feature, and the face posture influence index calculation formula is: ; represents an angle deviation function, represents an occlusion function, represents an expression change function, represents the face posture influence index; The angle deviation function expression is: ; represents an angle deviation value, represents a preset non-zero constant two, represents the maximum allowed deviation under ideal identification conditions; The shielding function calculation logic is: predefining each type of face recognition area of the face, obtaining the ratio of the total amount of shielding pixels of each type of face recognition area to the total amount of pixels of the area, then multiplying each type of face recognition area by the corresponding preset weight coefficient to obtain the shielding influence value of each type of face recognition area and summing them up. The expression change function calculation logic is: predefining n initial points of the initial expression of the face, then obtaining the actual coordinates corresponding to the n initial points in the image, respectively calculating the Euclidean distance between the actual coordinates corresponding to each initial point and the initial coordinates, and then taking the average value.

[0009] In a preferred embodiment, the angle deviation value acquisition logic is: Substitute the yaw angle, pitch angle and roll angle into the exponential decay function, and the result is the angle deviation value.

[0010] In a preferred embodiment, the machine learning model is a convolutional neural network model.

[0011] In a preferred embodiment, fuzzy logic reasoning is used in optimization analysis, and the reasoning result is that the type corresponding to the current camera acquisition environment condition is a reasonable acquisition environment or an unreasonable sampling environment.

[0012] In a preferred embodiment, the use logic of fuzzy logic reasoning is: Obtain the feature analysis result, the identity verification result, the confidence of the identity verification result, and the number of times of re-identity verification in a preset time window, then calculate the average value of the environment condition influence index, the average value of the face posture influence index, the average value of the identity verification result, and the frequency of re-identity verification, and together as the input variables of the fuzzy logic, the type corresponding to the current camera acquisition environment condition as the output variable of the fuzzy logic, the input variables are fuzzy processed, the values of the input variables are converted into fuzzy sets, the output variables are fuzzy processed, the output variables are converted into fuzzy sets, the fuzzy rules are formulated, the adaptation degree of the current camera acquisition environment condition corresponding to each type under different data category combinations is described, the fuzzy input variables are inferred through the fuzzy rules, and the type corresponding to the current camera acquisition environment condition is obtained.

[0013] In a preferred embodiment, an intelligent office management system based on face recognition function comprises: An image acquisition module acquires face image data of a user through a camera; An identity verification module uses a preset face recognition algorithm to perform identity verification on the acquired image data, identifies and records the identity information of the user; A feature processing module performs feature extraction and analysis based on the collected image data; A credibility evaluation module inputs the feature analysis results into a pre-trained machine learning model and outputs the credibility of the current identity verification result; A re-identification module re-performs identity verification when the credibility of the current identity verification result triggers a pre-set re-identification mechanism; An optimization analysis module obtains the feature analysis results, the credibility of the identity verification result, and the number of times of re-performing identity verification within a pre-set time window, and performs optimization analysis to determine whether to optimize the current camera capturing environment conditions based on the optimization analysis results.

[0014] The technical effects and advantages of the present application are as follows: The present application can evaluate the rationality of the camera capturing environment in real time by analyzing the environmental condition influence index and the facial pose influence index, and optimize when necessary. This automatic adjustment can effectively reduce manual intervention, improve the intelligent level of the system, and ensure the best recognition effect under different environmental conditions. By combining the analysis of environmental conditions and facial features, the present application can more comprehensively evaluate various factors that affect face recognition accuracy, especially in complex environments (such as low light, angle changes, or occlusion), the system can adjust the recognition process to improve the accuracy and stability of recognition.

[0015] The present application can reduce misidentification and missed identification caused by environmental factors (such as poor lighting, angle problems, etc.) through intelligent optimization and adaptive adjustment. Especially in the face of dynamic changes in the environment, the system can more accurately identify and optimize, significantly improving the credibility of identity verification. The present application combines the influence of environmental conditions and facial features with the credibility of identity verification results for real-time analysis through fuzzy logic reasoning. According to real-time feedback, the system can adjust the environment to continuously maintain high recognition accuracy and system stability, meeting the efficient and safe needs of daily office management. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to facilitate understanding by those skilled in the art, the present application will be further described below with reference to the accompanying drawings; Figure 1 A schematic diagram of an intelligent office management method based on face recognition function in the present application.

[0017] Figure 2 A schematic diagram of an intelligent office management system based on face recognition function in the present application. DETAILED DESCRIPTION

[0018] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0019] With reference to Figure 1 - Figure 2 The following examples are obtained: Example 1: With the improvement of modern office automation level, especially the rapid development of intelligent office technology, the traditional manual identity verification means (such as access card, password, etc.) gradually exposes its security, convenience and management deficiencies. Especially in some occasions with high security requirements, the traditional identity verification means is easy to be misused or cracked, which brings hidden dangers. In addition, with the development and maturity of face recognition technology, more and more intelligent management systems use this technology to improve efficiency and security.

[0020] However, the existing face recognition technology still faces some challenges in actual application, especially the decline of recognition accuracy in complex environments (such as poor lighting conditions, poor user posture, occlusion, etc.). These factors will directly affect the stability and accuracy of the system. Therefore, how to improve the accuracy and robustness of face recognition, and how to automatically optimize the recognition accuracy according to the environment and user facial features, have become urgent problems to be solved.

[0021] The present application proposes an intelligent office management method based on face recognition function, aiming to solve the deficiencies in the prior art through the following improvements: Improve recognition accuracy and robustness: By combining environmental features and facial features, the influence of factors such as light, angle, and occlusion on recognition accuracy is analyzed and optimized, thereby improving the accuracy and robustness of face recognition, especially in complex environmental application scenarios.

[0022] Intelligent optimization mechanism: By introducing credibility scoring and re-recognition mechanism, the present application can dynamically adjust and optimize the recognition process of the system, ensuring the credibility and accuracy of the recognition results. Through intelligent optimization based on collected environmental data, the system can automatically adjust the working state under different environmental conditions, ensuring the stability and efficiency of the system.

[0023] Improve user experience and office efficiency: Through the automatic identity verification process, the present application can identify user identity and record information in real time and accurately without human intervention, improving the intelligent level of the office environment. In addition, the optimized intelligent recognition system can reduce the false recognition rate and improve the efficiency of the overall work flow.

[0024] Stronger security: Through machine learning-based adaptive optimization analysis, the system can dynamically adjust the recognition strategy according to user facial features, environmental changes, etc., thereby enhancing the protection capability of the system and reducing potential security risks.

[0025] An intelligent office management method based on face recognition function, comprising the following steps: Collect the user's facial image data through the camera, and use the pre-set face recognition algorithm to verify the collected image data, identify and record the user's identity information; Collect the user's facial image data in real time through the camera, and use the pre-set face recognition algorithm to verify the collected image data. This is the core step of the intelligent office management method, the purpose is to identify and verify the user's identity in an efficient and secure way, to avoid the potential security risks of traditional identity authentication methods (such as access cards, passwords, etc.). Through biometric technology (face recognition), the accuracy and security of identity verification are improved, avoiding problems such as password theft or loss of access cards. After verifying the user's identity, the system automatically records his identity information, supporting subsequent management operations (such as recording attendance, access control, etc.). The face recognition algorithm belongs to the existing technology and is maturely applied in the technical field of the present invention, so it will not be described again.

[0026] Based on the collected image data, feature extraction and analysis are performed, and then the feature analysis results are input into the pre-trained machine learning model to output the credibility of the current identity verification result; This step extracts various features (such as the relative positions of eyes, nose, mouth, facial angle, expression, etc.) in the user's facial image. These features help improve the accuracy of identity verification, especially in dynamic environments. By inputting the extracted features into the pre-trained machine learning model, a deep analysis can be performed based on historical data and features to output a more accurate credibility score. This learning-based model can continuously optimize to improve the accuracy and intelligence level of recognition. According to the recognition results and features, the credibility score of the current identity verification result is calculated to help judge the accuracy of the current verification. If the credibility of the current verification is lower than the pre-set threshold (for example, due to environmental factors such as light, angle or obstruction, leading to recognition errors), the system will trigger a re-recognition mechanism. At this time, the system will re-collect images and perform identity verification. Improve verification accuracy: This mechanism can dynamically optimize the verification process based on real-time recognition accuracy feedback, reducing misidentification or missed identification, ensuring the accuracy of user identity verification. Enhance user experience: When there is an identity verification error, the system automatically adjusts and re-verifies, avoiding manual intervention and improving user experience.

[0027] When the credibility of the current identity verification result triggers the preset re-identification mechanism, re-identity verification is performed. It should be noted that the application scenario of the present application is the credibility verification after the current face recognition passes, and the purpose is to ensure the reliability of the face recognition pass result. Specifically, when the credibility of the current identity verification result is lower than the preset threshold, the preset re-identification mechanism is triggered, and the identity verification of face recognition is required. When the credibility of the current identity verification result is greater than or equal to the preset threshold, the preset re-identification mechanism is not triggered, and the result of the identity verification of face recognition is determined to be reliable.

[0028] The credibility of the feature analysis result and the identity verification result in the preset time window, and the number of times of re-identity verification are obtained, and optimization analysis is performed, and whether the current camera acquisition environment condition is optimized is determined according to the optimization analysis result. By analyzing the feature result, the credibility score and the number of re-verification, the performance of the system in a certain time window is obtained. Through the analysis of these data, potential environmental problems (such as poor lighting, angle problems, etc.) can be identified and optimized. Environmental optimization decision: according to the result of optimization analysis, the system can automatically decide whether to adjust the camera acquisition environment condition (for example, improve the lighting, adjust the camera angle, etc.). This adaptive optimization mechanism can ensure that the face recognition process always maintains high accuracy and robustness. This step automatically adjusts the environmental conditions by analyzing the changes in recognition effect, avoids the influence of environmental factors on the recognition effect, and ensures the stability of the system.

[0029] Feature extraction refers to extracting environmental features when the camera collects user facial image data, obtaining an environmental feature group, and extracting user facial features when the camera collects user facial image data, obtaining a facial feature group. Environmental conditions have a great impact on the accuracy of face recognition. For example, too strong or too dark light, backlight, shadow, and other factors may cause the recognition accuracy to decrease. By extracting environmental features (such as light intensity, contrast, shadow, backlight, etc.), the system can understand the current environmental conditions and make corresponding adjustments and optimizations based on the environmental features. The environmental feature group provides data support for subsequent analysis. Based on these features, the system can determine whether environmental optimization is needed, such as adjusting the camera's light, adjusting the angle, etc., to ensure the accuracy of the recognition. The facial feature group extraction is performed after the preliminary face recognition passes, and the main purpose is to further analyze and judge the credibility of the recognition result. Even if the preliminary face recognition passes, the system still needs to extract more facial features to confirm the accuracy of the recognition result to ensure that the final authentication result is reliable. By extracting the user's facial angle, eye, mouth relative position, expression change, etc., the system can obtain more detailed information. These information helps to judge whether there is an error in face recognition, avoid recognition errors caused by environmental factors such as light and angle, and improve the accuracy of recognition. The extraction of facial features helps to make more detailed evaluation of the recognition result in specific situations (such as user expression change, slight occlusion, etc.). Even if the preliminary recognition passes, the data provided after the facial feature group extraction can further verify the reliability of the result.

[0030] Feature analysis refers to: Based on the environmental feature group, the environmental analysis is performed to generate an environmental condition impact index representing the influence of the light conditions of the image on the recognition accuracy, and based on the facial feature group, the facial analysis is performed to generate a facial pose impact index representing the influence of the user's facial conditions on the recognition accuracy.

[0031] The environmental condition impact index (ECII) integrates the influence of four key environmental features (light intensity, contrast, shadow, and backlight) on image recognition accuracy. It can evaluate the influence of these changes on the recognition result based on the changes in the image collection environment, so as to further determine whether optimization is needed to improve the accuracy and robustness of the face recognition system. The logic for obtaining the environmental condition impact index is: The environmental feature group includes light intensity features, contrast features, shadow features, and backlight features, and the environmental condition impact index calculation formula is: ; The environmental feature type is corresponding preset non-zero proportion coefficient, which is used to represent the influence degree of each environmental feature on the recognition accuracy. Through different weight values, the system can determine which environmental factors have a greater influence on recognition, thereby adjusting and optimizing the strategy. For example, the light intensity may be more important than the shadow, so the light feature can be given a higher weight, environmental feature type corresponding influence function, environmental feature type When i = 1, 2, 3, 4, it respectively corresponds to the light intensity feature, the contrast feature, the shadow feature, and the backlight feature, environmental condition influence index; the influence function of the environmental feature type is used to calculate the influence of each environmental feature on the recognition accuracy. Different environmental features (light, contrast, shadow, and backlight) have different influence functions, and the environmental feature type The corresponding influence function is: ; represents the average brightness of the image, represents the contrast between the face and the background, represents a preset optimal contrast threshold, represents a preset contrast sensitivity adjustment coefficient, represents the pixel ratio of the shadow area in the image to the entire image area, represents a preset non-zero constant one, represents the ratio of the backlight area in the image to the entire image area. The logarithmic function is used to calculate the influence of light intensity on the image, which is suitable for a large range of light changes. The Sigmoid function is used to represent the influence of contrast on recognition accuracy, which is highly adaptable and can handle relatively smooth contrast changes. The exponential decay function is used to represent the influence of shadow, and the recognition accuracy decreases rapidly with the increase of the shadow area. The square root function is used to represent the influence of the backlight area on the recognition accuracy. The square root function is used to model the influence of backlight on the image, reflecting the influence of the area size of the backlight area on the recognition accuracy. The environmental feature type i ∈ {1, 2, 3, 4} respectively represents the light intensity, contrast, shadow, and backlight four environmental features. These features are different according to the environment during image acquisition, and have different influences on the recognition result, so it is necessary to design the corresponding influence function for each feature.

[0032] The face pose influence index (FPII) comprehensively considers the influence of face angle deviation, occlusion degree, and expression change on face recognition accuracy. The face pose influence index can quantify the influence of these factors on the recognition accuracy, thereby helping the system to optimize and verify the recognition result. The acquisition logic of the face pose influence index is: The facial feature group includes an angle deviation feature, an occlusion degree feature, and an expression change feature, and the facial pose influence index calculation formula is: ; represents an angle deviation function, reflecting the influence of the angle deviation of the user's face relative to the camera on the recognition accuracy, represents an occlusion function, reflecting the influence of the degree of occlusion of the facial region on the recognition, represents an expression change function, reflecting the influence of the facial expression change on the recognition accuracy, represents the facial pose influence index; The angle deviation function is used to express the influence of the angle deviation of the user's face on the recognition result. Generally, the larger the facial angle deviation, the lower the recognition accuracy. The specific form of the formula is as follows: ; represents the angle deviation value, representing the angle error of the user's face relative to the camera, represents a preset non-zero constant two, used to adjust the influence intensity of the angle deviation on the recognition accuracy, represents the maximum allowed deviation under ideal recognition conditions; The occlusion function calculation logic is: predefine each facial recognition region of the face, obtain the ratio of the total amount of occluded pixels in each facial recognition region to the total amount of pixels in the region, then multiply each facial recognition region by the corresponding preset weight coefficient to obtain the occlusion influence value of each facial recognition region and sum it up; the occlusion function quantifies the influence of the occlusion degree of each region of the face on the recognition accuracy. The system will evaluate the influence of occlusion on the recognition result according to whether different facial regions (such as eyes, nose, and mouth) are occluded. Calculate the ratio of the total amount of occluded pixels in each region to the total amount of pixels in the region, and then weight and sum to obtain the total occlusion influence value. The preset weight coefficients corresponding to each facial recognition region can be adjusted and assigned according to their importance, and expert assignment method or other assignment methods can be used.

[0033] The expression change function calculation logic is: predefine n initial points of the initial expression of the face, then obtain the actual coordinates corresponding to the n initial points in the image, calculate the Euclidean distance between the actual coordinates corresponding to each initial point and the initial coordinates respectively, and then take the average. The influence of expression change on recognition accuracy is usually dynamic, and as the facial expression changes, the facial features will change. Calculating the Euclidean distance between the actual coordinates corresponding to each initial point and the initial coordinates is equivalent to calculating the offset degree of the expression trajectory of each initial point, and taking the average is to calculate the offset degree of the overall expression trajectory.

[0034] The angle deviation value acquisition logic is: The yaw angle, pitch angle, and roll angle are substituted into the exponential decay function to calculate the angle deviation value. The yaw angle represents the angle of left and right rotation of the face, i.e., the rotation of the face along the vertical axis. The pitch angle represents the angle of up and down tilt of the face, i.e., the rotation of the face along the horizontal axis. The roll angle represents the rotation angle of the face around its own axis, usually the rotation along the front and back axis of the face. For example, square and sum the three angle deviations (yaw angle, pitch angle, and roll angle), and then take the square root to get the total angle deviation value. This value is used to measure the difference between the user's face posture and the relative angle of the camera. The larger the deviation, the more obvious the decline in recognition accuracy. Combined with the exponential decay function, the formula can be Formula 1: ; or Formula 2: ; is the angle deviation value calculated from the yaw angle, pitch angle, and roll angle, is a preset non-zero deviation coefficient one, controlling the speed of the influence of the deviation on the recognition accuracy, The larger the value, the more the angle deviation will affect the recognition, is a preset non-zero deviation coefficient two, used to control the quadratic influence of the deviation. By introducing the quadratic term, the influence of large deviations on recognition is enhanced, is a preset non-zero amplification coefficient. The additional quadratic term can more accurately control the sensitivity in the recognition process, especially when the angle deviation is large, to ensure the robustness of the system in extreme cases. It can be used when the angle deviation value exceeds the preset range. Formula 1 uses a simple exponential decay form, which is suitable for most cases of angle deviation processing. When the angle change is not large, it can provide effective recognition accuracy prediction. Formula 2 introduces a quadratic term, making it more sensitive when facing large angle deviations. If the user's face angle deviation changes greatly during the recognition process, Formula 2 can provide more accurate control, especially in the case of large angle deviation, which can strengthen the system's response. The two formulas provide different flexibility. Formula 1 is a basic exponential decay model, suitable for simpler environments and application scenarios, while Formula 2 introduces more control through the quadratic term, suitable for complex situations or more extreme angle changes. This allows the system to choose the appropriate formula according to the actual situation, or to choose the appropriate decay method according to the characteristics of the angle change.

[0035] The machine learning model is a convolutional neural network model. In the present application, the role of the convolutional neural network model is: input data: environmental condition impact index and facial posture impact index, as input features of the convolutional neural network. The system passes the feature information extracted from these data to the network for credibility prediction. Learning and training: the network is trained through a large amount of historical data to learn the influence law of environmental conditions and facial postures on the credibility of identity verification. During the training process, the convolutional neural network automatically adjusts the network parameters through the back propagation algorithm, thereby optimizing the output results. Output results: the convolutional neural network will output a credibility score according to the input feature data, which represents the reliability of the current identity verification result. If the credibility is low, the re-identity verification mechanism may be triggered.

[0036] Optimization analysis uses fuzzy logic reasoning, and the reasoning result is that the type corresponding to the current camera acquisition environment condition is a reasonable acquisition environment or an unreasonable sampling environment.

[0037] The use logic of fuzzy logic reasoning is: The feature analysis result, the credibility of the identity verification result, and the number of re-identity verification in a preset time window are obtained, and then the environmental condition impact index average, the facial posture impact index average, the credibility of the identity verification result average, and the frequency of re-identity verification are calculated, which are used as input variables of fuzzy logic together, the type corresponding to the current camera acquisition environment condition is used as the output variable of fuzzy logic, the input variables are fuzzified, the values of the input variables are converted into fuzzy sets, the output variables are fuzzified, the output variables are converted into fuzzy sets, fuzzy rules are formulated to describe the adaptation degree of each type corresponding to the current camera acquisition environment condition under different data category combinations, the fuzzified input variables are inferred through the fuzzy rules to obtain the type corresponding to the current camera acquisition environment condition.

[0038] In the present application, fuzzy logic reasoning is used to determine whether the current camera acquisition environment condition is reasonable. Through analysis of different environments and user recognition results, fuzzy logic reasoning can output a rationality evaluation result of the camera acquisition environment according to the fuzzification of the input variables.

[0039] The specific steps of the inference logic are: obtaining data: first, collect the feature analysis results in a preset time window, including environmental condition impact index, facial posture impact index, credibility of identity verification result, and number of re-identity verification.

[0040] Computing the average values: For the data collected above, compute their average values within this time window. These average values represent the overall performance of the current environment, facial features, and identity verification. The environmental condition impact index average value: reflects the average impact of environmental conditions (such as lighting, contrast, etc.) on recognition accuracy. The facial pose impact index average value: integrates the impact of user facial angle, expression, and other factors on recognition accuracy. The identity verification result reliability average value: indicates the overall reliability of identity verification. The frequency of re-verification: reflects the number of times re-verification was triggered due to recognition failure within this time window.

[0041] Fuzzification of input variables: The above average values are taken as input variables for fuzzy logic, and through fuzzification, these numerical values are converted into fuzzy sets. Fuzzy sets allow the processing of data with uncertainty, so that different data values can be represented as a range rather than an exact value.

[0042] Fuzzification of output variables: The rationality of the camera-captured environment (rational or non-rational camera-captured environment) is taken as the output variable. The output variable is also subjected to fuzzification, representing the degree of adaptation of the camera-captured environmental conditions.

[0043] Formulation of fuzzy rules: Fuzzy rules are formulated to describe how different combinations of input variables affect the adaptability of the camera-captured environment. For example: if the environmental condition impact index is high and the facial pose impact index is low, and the reliability is high, then the camera-captured environment may be rational. If the environmental condition impact index is high and re-verification is frequently performed, then the camera-captured environment may be non-rational.

[0044] Fuzzy inference: Once all input variables are fuzzified and fuzzy rules are formulated, the system performs fuzzy inference. This is a process of obtaining output results by weighting fuzzy sets and combining fuzzy rules. The inference process can be carried out in the following ways: Use of trigonometric functions: Trigonometric functions are often used to describe the relative relationship between variables. Combine weighted rules: Different input variables are given different weights to reflect their degree of influence on environmental assessment. Fuzzified input variables are inferred through the set fuzzy rules, and finally the type of camera-captured environmental conditions (rational camera-captured environment or non-rational camera-captured environment) is obtained.

[0045] Fuzzy logic reasoning can automatically assess the rationality of camera environment collection according to the changes of the current environment, and give corresponding feedback. This automatic adjustment helps to improve the robustness of the recognition system under different environmental conditions. Through fuzzy processing, the system can tolerate uncertainty and ambiguity, and still make reasonable decisions when the data is incomplete or insufficient in accuracy, ensuring the reliability and adaptability of the system. By accurately assessing the rationality of environmental conditions, the system can make adjustments in cases of insufficient light, poor angle, etc., to ensure high recognition accuracy in different situations.

[0046] For example: Suppose the invention has obtained the following data within a time window: the average value of the environmental condition influence index is 70 (indicating that the environmental conditions have a greater impact on recognition); the average value of the facial pose influence index is 30 (indicating that the facial pose has a smaller impact); the average value of the credibility of the identity verification result is 90% (indicating that the recognition accuracy is high); the frequency of re-identifying is 2 times (there are 2 re-identifications within this time window). After fuzzy processing, the input variables (environmental condition influence index, facial pose influence index, credibility, and re-verification frequency) are classified into different categories, such as "high impact", "low impact", "high credibility", "low frequency", etc. According to the set fuzzy rules: if the environmental condition influence index is high, and the facial pose influence is small, and the identity verification credibility is high, then output "rational collection of environment". If the re-verification frequency is high and the environmental condition influence index is high, then output "non-rational collection of environment". The final evaluation shows that the environmental conditions of this time window are "rational collection of environment", indicating that it can continue to operate normally without adjusting the collection conditions.

[0047] Example 2: An intelligent office management system based on face recognition function, comprising: An image collection module collects facial image data of users through a camera; An identity verification module uses a pre-set face recognition algorithm to verify the collected image data, identifies and records the identity information of the user; A feature processing module extracts and analyzes features based on the collected image data; A credibility evaluation module inputs the feature analysis results into a pre-trained machine learning model and outputs the credibility of the current identity verification result; A re-identification module re-performs identity verification when the credibility of the current identity verification result triggers a pre-set re-identification mechanism; The optimization analysis module obtains the feature analysis result, the identity verification result, the credibility of the identity verification result, and the number of times of re-identity verification in a preset time window, and performs optimization analysis, and determines whether to optimize the current camera acquisition environment condition according to the optimization analysis result. When optimizing, the current camera acquisition environment condition such as the light condition and the camera shooting angle can be automatically optimized by using the existing genetic algorithm and the like.

[0048] The above formulas are dimensionless values calculated, and the formulas are obtained by collecting a large amount of data to simulate a formula of the nearest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0049] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0050] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0051] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0052] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent office management method based on face recognition function, characterized in that: The following steps are involved: Collect the user's facial image data, extract and analyze features based on the collected image data, and then input the feature analysis results into a pre-trained machine learning model to output the credibility of the current identity verification result; When the credibility of the current identity verification result triggers the preset re-identification mechanism, identity verification is performed again; Obtain the feature analysis results within the preset time window, the credibility of the identity authentication results, and the number of re-identification times, and perform optimization analysis. Based on the optimization analysis results, decide whether to optimize the current camera acquisition environment conditions.

2. The intelligent office management method based on face recognition function according to claim 1, characterized in that: Feature extraction refers to extracting the collection environment features when the camera collects the user's facial image data to obtain an environment feature group, and extracting the user's facial features when the camera collects the user's facial image data to obtain a facial feature group.

3. The intelligent office management method based on face recognition function according to claim 2, characterized in that: Feature analysis refers to: Based on the environmental feature group, environmental analysis is performed to generate an environmental condition impact index that represents the impact of the image's lighting conditions on recognition accuracy. Based on the facial feature group, facial analysis is performed to generate a facial posture impact index that represents the impact of the user's facial conditions on recognition accuracy.

4. The intelligent office management method based on face recognition function according to claim 3 is characterized in that: The logic for obtaining the environmental condition impact index is: The environmental feature group includes light intensity feature, contrast feature, shadow feature, and backlight feature. The environmental condition impact index calculation formula is: ; Environmental feature type The corresponding preset non-zero scale factor, Environmental feature type Corresponding influence function, environmental feature type When it is 1, 2, 3, and 4, it corresponds to the light intensity feature, contrast feature, shadow feature, and backlight feature, respectively. is the environmental conditions impact index.

5. The intelligent office management method based on face recognition function according to claim 4 is characterized in that: The logic for obtaining the facial posture influence index is: The facial feature group includes angle deviation features, occlusion features, and expression change features. The facial posture influence index calculation formula is: ; represents the angle deviation function, represents the occlusion function, represents the expression change function, Represents the facial posture influence index.

6. The intelligent office management method based on face recognition function according to claim 5, characterized in that: The logic for obtaining the angle deviation value is: Substitute the yaw angle, pitch angle, and roll angle into the exponential decay function, and the result is the angle deviation value.

7. The intelligent office management method based on face recognition function according to claim 6, characterized in that: The machine learning model is a convolutional neural network model.

8. The intelligent office management method based on face recognition function according to claim 7, characterized in that: Fuzzy logic reasoning is used in the optimization analysis, and the reasoning result is that the corresponding type of the current camera acquisition environment condition is a reasonable acquisition environment or an unreasonable sampling environment.

9. The intelligent office management method based on face recognition function according to claim 8, characterized in that: The logic used in fuzzy logic reasoning is: Obtain the feature analysis results, the credibility of the identity authentication results, and the number of re-authentication times within the preset time window, then calculate the average value of the environmental condition influence index, the average value of the facial posture influence index, the average value of the credibility of the identity authentication results, and the frequency of re-authentication, and use them together as input variables of the fuzzy logic. Use the corresponding type of the current camera acquisition environment condition as the output variable of the fuzzy logic, fuzzify the input variables, and convert the values ​​of the input variables into fuzzy sets. Fuzzy process the output variables and convert the output variables into fuzzy sets. Formulate fuzzy rules to describe the adaptability of the corresponding types of the current camera acquisition environment conditions under different data type combinations. Infer the fuzzified input variables through fuzzy rules to obtain the corresponding type of the current camera acquisition environment condition.

10. An intelligent office management system based on face recognition function, used to implement the intelligent office management method based on face recognition function according to any one of claims 1 to 9, characterized in that: include: Image acquisition module, which collects user's facial image data through a camera; The identity authentication module uses a preset face recognition algorithm to authenticate the collected image data, identify and record the user's identity information; Feature processing module, which performs feature extraction and analysis based on the collected image data; The credibility assessment module inputs the feature analysis results into a pre-trained machine learning model and outputs the credibility of the current authentication result; Re-identification module, which re-performs identity authentication when the credibility of the current identity authentication result triggers the preset re-identification mechanism; The optimization analysis module obtains the feature analysis results under the preset time window, the credibility of the identity authentication results, and the number of re-identification times, and performs optimization analysis. Based on the optimization analysis results, it decides whether to optimize the current camera acquisition environment conditions.

Images