Face recognition method based on artificial intelligence model

Abstract

The invention relates to the technical field of face recognition, and particularly discloses a face recognition method based on an artificial intelligence model, and the method comprises the following steps: S1, photographing a face image, generating a face model, presetting a collection point in the face model, and setting a living body attribute for the collection point; s2, screening a living body model and a non-living body model based on living body attributes, and calculating comparison data of collection points in the feature region; s3, sorting feature regions in the living body model, calculating a difference value and screening a first similar model; s4, screening a second similar model based on the comparison points; and S5, calculating similarity rates of the living body model and the second similar models, and selecting the second similar model with the maximum similarity rate. The human face recognition method based on the artificial intelligence model is used for improving the accuracy of an artificial intelligence model living body detection technology and distinguishing a real human face from an attack medium, so that the accuracy of human face recognition is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of face recognition, and particularly relates to a face recognition method based on an artificial intelligence model. Background Art

[0002] An artificial intelligence model is a mathematical model that uses machine learning algorithms to learn from training data and predict new data. It is a simplified and abstract description of a certain aspect or specific problem in the real world to achieve specific goals or tasks. The components of an artificial intelligence model include data, algorithms, and computing resources. Data is the basis for model training and operation, including training data, validation data, and test data. The data needs to be rich, accurate, and cover various possible situations. Algorithms are the core methods for learning features and patterns from data.

[0003] In the face recognition method using an artificial intelligence model, liveness detection is the ability to distinguish a real face from an attack medium (such as a photo, video, etc.). In a face recognition system, it is a key step to prevent fraudulent attacks. Although some liveness detection technologies have been proposed and applied, there are still some limitations. For example, some liveness detection methods may be affected by environmental conditions, resulting in a decrease in detection accuracy; other methods may be easily deceived or bypassed.

[0004] In the prior art, it is necessary to design a face recognition method based on an artificial intelligence model to improve the accuracy of the liveness detection technology of the artificial intelligence model, distinguish a real face from an attack medium, and thus improve the accuracy of face recognition. Summary of the Invention

[0005] The purpose of the present invention is to provide a face recognition method based on an artificial intelligence model to solve the above technical problems.

[0006] The purpose of the present invention can be achieved by the following technical solutions: A face recognition method based on an artificial intelligence model includes the following steps: S1: Take a face image and generate a facial model. Preset N collection points in the facial model, obtain the distance L between the collection points and the camera and the temperature W at the collection points, and set liveness attributes for the collection points based on the distance L and the temperature W; S2: Screen the liveness model and the non-liveness model based on the liveness attributes, mark the feature regions of the liveness model and obtain the collection points within the feature regions. The feature regions refer to the five sense organs in the liveness model, and calculate the comparison data S of the collection points within the feature regions; S3: Sort the feature regions in the liveness model in a preset order and obtain {M1, M2,..., M n}, where M nRepresents the number of acquisition points within the nth feature region, and calculates the difference , if the difference C = 0, then mark the corresponding face model in the database as the first similar model, where I represents the number of feature regions, and MS i represents the number of acquisition points within the ith feature region of the face model in the database; S4: Establish a spatial coordinate system and generate a live model and the first similar model. Mark the acquisition points within the feature region as comparison points, and screen for the second similar model based on the comparison points; S5: Calculate the similarity rate K between the live model and the second similar model, screen for the second similar models with a similarity rate K greater than 90%, sort the second similar models in descending order of the similarity rate K, and the first second similar model in the sorting is the face corresponding to the live model.

[0007] As a further solution of the present invention: In the step S1, setting the live attribute for the acquisition points based on the distance L and the temperature W includes: Obtain the current external air temperature Ws, calculate the temperature difference ΔW = W - Ws at the acquisition point and the average value ΔW of the temperature difference ave , and calculate the temperature difference ratio BW = ΔW / ΔW ave ; Obtain the average value L of the distance between the acquisition point and the camera ave , calculate the distance ratio BL = L / L ave , and set the live attribute (BW: BL) for the acquisition point.

[0008] As a further solution of the present invention: The steps of screening the live model and the non-live model based on the live attribute include: Place the ultrasonic transmitter at the camera position, make the ultrasonic transmitter operate at a preset power and receive the intensity Q of the rebounded ultrasonic wave, and calculate the theoretical intensity of the rebounded ultrasonic wave , where BW n represents the temperature difference ratio of the nth acquisition point, and BL n represents the distance ratio of the nth acquisition point. If the intensity Q of the rebounded ultrasonic wave is within the interval [QL - λ, QL + λ], then mark it as the live model, and λ represents the preset fluctuation value.

[0009] As a further solution of the present invention: In the step S2, comparing the live model with the database model to obtain the comparison data includes: Obtain the acquisition points in the feature region. The feature region refers to the facial features and cheeks in the facial model. Obtain the red channel intensity R, green channel intensity G, and blue channel intensity B of the acquisition points in the feature region, and calculate the comparison data , where H represents the grayscale value of the corresponding acquisition point.

[0010] As a further solution of the present invention: in the step 2, if it is a non-living model, prompt the staff to check the object photographed by the camera.

[0011] As a further solution of the present invention: in the step S3, if the first similar model does not exist in the database, mark it as an outsider and conduct identity verification.

[0012] As a further solution of the present invention: in the step S4, the specific process of screening the second similar model based on the comparison points includes: Record the comparison points of the living model as reference points, calculate the distance S between the acquisition points and the reference points in the first similar model within the same feature region, and calculate the average value S of the distances ave , if the average value S of the distances ave is less than the preset distance judgment value S, record the corresponding first similar model as the second similar model.

[0013] As a further solution of the present invention: in the step S4, the specific process of calculating the similarity rate K between the living model and the second similar model includes: Select the comparison points of the same part in the living model and the second similar model, connect the comparison points with the origin of the space coordinate system to generate feature vectors, and the direction of the feature vectors is from the origin to the comparison points, and calculate the similarity rate , where represents the i-th feature vector in the living model, represents the i-th feature vector in the second similar model.

[0014] The beneficial effects of the present invention: The present invention screens out the living model by obtaining the distance between the acquisition point and the camera and the temperature at the acquisition point, and compares the attributes of the acquisition points in the feature region of the living model with the existing data in the database, so as to obtain the face model with the highest similarity, thereby realizing a face recognition method based on an artificial intelligence model to improve the accuracy of the living detection technology of the artificial intelligence model, distinguish the real face from the attack medium, and thus improve the accuracy of face recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below with reference to the accompanying drawings.

[0016] Figure 1 is a schematic flow chart of a face recognition method based on an artificial intelligence model of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0018] Please refer to Figure 1 as shown, the present invention is a face recognition method based on an artificial intelligence model, including the following steps: S1: Shoot a face image and generate a facial model. Preset N collection points in the facial model, obtain the distance L between the collection points and the camera and the temperature W at the collection points, and set a living body attribute for the collection points based on the distance L and the temperature W; S2: Screen the living body model and the non-living body model based on the living body attribute, mark the feature areas of the living body model and obtain the collection points within the feature areas. The feature areas refer to the five sense organs in the living body model, and calculate the comparison data S of the collection points within the feature areas; S3: Sort the feature areas in the living body model in a preset order to obtain {M1, M2,..., M n}, where M n represents the number of collection points in the nth feature area, calculate the difference , if the difference C = 0, then record the corresponding face model in the database as the first similar model, where I represents the number of feature areas, and MS i represents the number of collection points in the ith feature area of the face model in the database; S4: Establish a spatial coordinate system and generate a living body model and a first similar model, record the collection points within the feature areas as comparison points, and screen the second similar model based on the comparison points; S5: Calculate the similarity rate K between the living body model and the second similar model, screen the second similar models with the similarity rate K greater than 90%, sort the second similar models in descending order of the similarity rate K, and the first second similar model in the sorting is the face corresponding to the living body model.

[0019] It should be noted that to effectively determine the difference between a living body and non-living bodies such as a photo or a model, a crucial point is to utilize the element of the distance between the camera and the collection points. The face has feature areas with rich information. Due to the existence of the five sense organs such as eyes, nose, mouth, and eyebrows, the surface of the face naturally presents a concave-convex contour.

[0020] This unique physiological structure results in the fact that when collecting face images, the distance between the camera and the collection point is not fixed. When factors such as the collection angle, light conditions, and face pose change, the camera needs to flexibly adjust the distance from the collection point according to the actual situation in order to clearly capture all the detailed features of the face.

[0021] As a living body, the internal metabolic activities of the human body will generate heat. This is one of the basic characteristics of life activities, and the generation of this heat is relatively stable and detectable. Through specialized thermal imaging equipment or other temperature sensors, the heat distribution emitted by the human body can be accurately sensed.

[0022] However, in sharp contrast to the human body, a photo is merely a flat record of a real scene. It is essentially a two-dimensional image information and does not have the attributes of a real physical entity, so it will not generate heat. Similarly, whether it is an art model for exhibition or a simulation model for other purposes, most of them are made of various non-living materials and do not have the metabolic function of a living body, so they will not generate heat either.

[0023] Based on these significant differences, it is possible to rely on the variation law of the distance between the camera and the collection point and the temperature information at the collection point. Thus, these live data can be used to accurately identify and distinguish between live and non-live objects, effectively ensuring the security and reliability of the system and avoiding potential risks and losses caused by problems such as identity impersonation.

[0024] Next, based on the unique attributes of the live object, the live models are screened out. Through this operation, not only can the live models with life characteristics be accurately selected, but also they can be clearly distinguished from the non-live models that do not have life attributes.

[0025] After successfully screening out the live models, focus on each collection point within the feature area. By calculating the information contained in these collection points, valuable comparison data can be obtained. When coordinating between collection points, the comparison data can help us more accurately judge the relevance and differences between different collection points, making each collection point no longer exist in isolation but rather be interconnected and mutually verified, jointly constituting a complete recognition system.

[0026] In terms of optimization, the comparison data can provide us with detailed information about the collection points. Through the analysis of these data, we can targetedly adjust and optimize these collection points, thereby improving the accuracy and reliability of the entire face recognition system. In order to gain a deeper understanding of the uniqueness of each collection point, two important image feature parameters, namely the three primary colors and the grayscale value, are used for precise calculation.

[0027] As the basic elements of color composition, the three primary colors can reflect the characteristics of the acquisition points in the color dimension. Different factors such as skin color and lighting conditions will affect the three-primary-color values of the acquisition points. By accurately measuring and analyzing these values, combining the three primary colors with the grayscale values, the special features of each acquisition point can be calculated. These special features are like the unique "fingerprints" of each face and are important bases for judging the similarity of faces. In the actual operation process, the feature information of different acquisition points is compared one by one to quantify the similarity between them. This comparison method based on multi-dimensional feature information can more accurately reflect the similarity between two faces, thus providing more reliable and accurate results for face recognition technology.

[0028] Subsequently, a spatial coordinate system needs to be constructed, which is used to depict the morphological and positional characteristics of the living model and the first similar model. With the help of this spatial coordinate system, the living model and the first similar model can be generated efficiently, and the comparison and analysis between the models can be intuitively carried out under the same coordinate framework. This simplifies the traditional recognition process. By transforming the complex model comparison problem into data operation and analysis in the coordinate system, the processing efficiency is significantly improved.

[0029] Particularly crucial is that during the process of establishing the acquisition points in the facial feature regions, the comparison points should cover key facial feature regions such as eyes, nose, mouth, and eyebrows. These regions contain rich facial feature information and play a decisive role in distinguishing the faces of different individuals. By focusing on the acquisition points in these facial feature regions, we can not only dig deeper into the subtle differences of the face but also provide strong data support for the subsequent screening of the second similar models.

[0030] After that, through the described algorithms and models, the similarity law K between the living model and the second similar model is calculated. The calculation of this similarity law K is based on in-depth comparison and analysis of the two models in multiple dimensions, including but not limited to the shape, position, proportion of facial features, and the overall characteristics of the facial contour. To ensure the basic success rate, a threshold of 90% is first set, which means that only when the similarity rate K between the living model and the second similar model is greater than 90% will the second similar model be included in the further screening range.

[0031] This step ensures the second similar models that have a high similarity with the living model, thus effectively reducing the risk of misidentification. After the preliminary screening, the second similar models are sorted in descending order of the similarity rate K. This sorting process is based on the concept that the higher the similarity rate of the second similar model, the higher its matching degree with the face.

[0032] Therefore, the second most similar model in the sorting, that is, the model with the highest similarity rate to the live model, is identified as the face corresponding to the live model. This selection method not only improves the success rate of face recognition, but also further enhances the accuracy and reliability of recognition by ensuring a high degree of similarity between the selected model and the live model in multiple dimensions.

[0033] In another preferred embodiment of the present invention, setting the live body attribute for the acquisition point based on the distance L and the temperature W includes: Obtain the current external air temperature Ws, calculate the temperature difference ΔW = W - Ws at the acquisition point and the average value ΔW of the temperature difference ave , and calculate the temperature difference ratio BW = ΔW / ΔW ave ; Obtain the average value L of the distance between the acquisition point and the camera ave , calculate the distance ratio BL = L / L of the acquisition point ave , and set the live body attribute (BW: BL) for the acquisition point.

[0034] It is worth noting that to effectively judge the difference between a live body and non-live bodies such as a photo and a model, the face has characteristic regions with rich information. Due to the presence of the five sense organs such as eyes, nose, mouth, and eyebrows, the surface of the face naturally presents a concave-convex contour.

[0035] This unique physiological structure results in the fact that the distance between the camera and the acquisition point is not fixed when collecting face images. When factors such as the acquisition angle, light conditions, and face pose change, the camera needs to flexibly adjust the distance from the acquisition point according to the actual situation in order to clearly capture each detailed feature of the face.

[0036] The metabolic activities of the human body will generate heat, and the heat distribution emitted by the human body can be sensed by a thermal imaging device. However, in sharp contrast to the human body, a photo is only a planar record of a real scene, so it will not generate heat.

[0037] Based on these significant differences above, it is possible to rely on the change law of the distance between the camera and the acquisition point and the temperature information at the acquisition point. Thus, it is possible to accurately identify and distinguish between a live body and non-live bodies with the help of these live body data, effectively ensuring the security and reliability of the system and avoiding potential risks and losses caused by problems such as identity impersonation.

[0038] In another preferred embodiment of the present invention, the steps of screening the live model and the non-live model based on the live body attribute include: Place the ultrasonic transmitter at the camera position, make the ultrasonic transmitter operate at a preset power and receive the ultrasonic intensity Q after reflection, and calculate the theoretical intensity after the ultrasonic wave rebounds , where BW n represents the temperature difference ratio of the nth acquisition point, and BL n represents the distance ratio of the nth acquisition point. If the ultrasonic intensity Q after reflection is within the interval [QL - λ, QL + λ], it is recorded as a living body model, and λ represents a preset fluctuation value.

[0039] It can be understood that when ultrasonic waves are used to detect a living body, due to the specific acoustic properties of living tissues, ultrasonic waves will encounter different interfaces and tissue structures during propagation, resulting in phenomena such as reflection, refraction, and scattering. However, due to the relatively uniform internal structure of non-living objects and the lack of characteristics of life activities, the characteristics of the echo signals are also different. By comparing the characteristics of the echo signals of living and non-living objects, the two can be distinguished, and through the analysis and processing of these echo signals, the effective distinction and identification of living and non-living objects can be achieved.

[0040] In another preferred embodiment of the present invention, comparing the living body model with the database model to obtain comparison data includes: Obtaining the acquisition points in the feature region, where the feature region refers to the facial features and cheeks in the facial model, obtaining the red channel intensity R, green channel intensity G, and blue channel intensity B of the acquisition points in the feature region, and calculating the comparison data , where H represents the gray value of the corresponding acquisition point.

[0041] It should be noted that the three primary colors (RGB) contain the color information of the image. The skin color, eye color, lip color, etc. of different ethnic groups have unique distribution characteristics in the three-primary-color space. The gray value reflects the brightness information of the image. The facial features, wrinkles, and other details of the human face can be better presented in the gray-scale image. Combining the three primary colors and the gray value can not only use the color information for preliminary screening but also analyze the detailed features such as facial features with the help of gray-scale information, realizing the comprehensive integration of information and improving the accuracy of recognition.

[0042] In another preferred embodiment of the present invention, if it is a non-living body model, the staff is prompted to check the object photographed by the camera.

[0043] It should be noted that the prompt content explains that the detected object is not a living body, and the possible potential problems or abnormal situations, such as the object may be a replica, a prop, or a substitute that does not meet the work requirements, etc. At the same time, it is necessary to emphasize the adverse effects that such a non-living body model may have on subsequent work links, data analysis, and the final result, and emphasize the necessity and urgency of checking it.

[0044] In another preferred embodiment of the present invention, if the first similar model does not exist in the database, it is marked as an outsider and identity verification is performed.

[0045] It is understandable that the system will record the relevant information of the foreign personnel and the process of identity verification for subsequent query and analysis. If the identity verification is passed, it will be decided whether to release the person or conduct further management and monitoring according to the specific situation; if the identity verification fails or there are suspicious situations, further measures will be taken.

[0046] In another preferred embodiment of the present invention, screening the second similar model based on the comparison points specifically includes: Denote the comparison points of the live model as reference points, calculate the distance S between the acquisition points in the first similar model and the reference points within the same feature region, and calculate the average value S of the distances ave , if the average value S of the distances ave is less than the preset distance judgment value S, denote the corresponding first similar model as the second similar model.

[0047] It should be noted that during the face comparison process, the comparison points of the live model are denoted as reference points. For the first similar model within the same feature region, calculate the distance between the acquisition points and the reference points therein. Specifically, for each acquisition point, measure the distance between it and the corresponding reference point respectively to obtain multiple distance values. Then calculate the average value of these distance values. If the average value of the distances is less than the preset distance judgment value, denote the corresponding first similar model as the second similar model. The model similar to the live model can be screened out more precisely, improving the accuracy and reliability of recognition. At the same time, by setting the preset distance judgment value, the similarity standard can be adjusted according to the specific situation to adapt to different application scenarios and requirements.

[0048] In another preferred embodiment of the present invention, calculating the similarity rate K between the live model and the second similar model specifically includes: Select the comparison points of the same part in the live model and the second similar model, connect the comparison points with the origin of the space coordinate system to generate feature vectors, and the direction of the feature vector points from the origin to the comparison point, and calculate the similarity rate , where represents the i-th feature vector in the live model, represents the i-th feature vector in the second similar model.

[0049] It should be noted that vectors can precisely represent various features of the human face. By converting the key information of the human face, such as the shape, position, and size of the facial features, into vector form, the details of the human face can be depicted more meticulously. Compared with the traditional method based on intuitive observation or simple comparison, vector representation can capture richer information, thereby improving the accuracy of similarity calculation.

[0050] The calculation of face similarity can consider features in multiple dimensions. The vector can not only include the geometric features of the facial features, but also incorporate dynamic features such as skin texture, skin color, and expressions, so as to more comprehensively evaluate the similarity of faces. And the vector calculation method has strong robustness to noise and interference factors.

[0051] The above has described in detail one embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equal changes and improvements made according to the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. A face recognition method based on an artificial intelligence model, characterized in that: The following steps are involved: S1: Take a face image and generate a face model, preset N collection points in the face model, obtain the distance L between the collection point and the camera and the temperature W at the collection point, and set the liveness attribute for the collection point based on the distance L and the temperature W; S2: Screening living models and non-living models based on living body attributes, marking characteristic regions of the living models and obtaining acquisition points within the characteristic regions, wherein the characteristic regions refer to the facial features of the living models, and calculating comparison data S of the acquisition points within the characteristic regions; S3: Sort the feature regions in the living body model in a preset order and obtain {M1, M2, ..., M n }, where M n Represents the number of collection points in the nth feature area, and calculates the difference , if the difference C=0, the corresponding face model in the database is recorded as the first similarity model, where I represents the number of feature regions, MS i Represents the number of acquisition points in the i-th feature area of ​​the face model in the database; S4: establishing a spatial coordinate system and generating a living body model and a first similar model, recording the collection points in the feature area as comparison points, and selecting a second similar model based on the comparison points; S5: Calculate the similarity K between the living body model and the second similarity model, select the second similarity model with a similarity K greater than 90%, and sort the second similarity models in descending order according to the similarity K. The first second similarity model in the sorting is the face corresponding to the living body model.

2. The face recognition method based on artificial intelligence model according to claim 1, characterized in that: In the step S1, setting the living body attribute for the collection point based on the distance L and the temperature W includes: Get the current outside temperature Ws, calculate the temperature difference ΔW=W-Ws at the collection point and the mean value of the temperature difference ΔW ave , and calculate the temperature difference ratio BW=ΔW / ΔW ave ; Get the mean value L of the distance between the acquisition point and the camera ave , calculate the distance ratio of the collection points BL=L / L ave , set the liveness attribute (BW:BL) for the collection point.

3. The face recognition method based on artificial intelligence model according to claim 1, characterized in that: In the step S2, the step of screening the living body model and the non-living body model based on the living body attributes includes: Place the ultrasonic transmitter at the camera position, operate it at the preset power, receive the ultrasonic intensity Q after rebound, and calculate the theoretical intensity of the ultrasonic wave after rebound. , where BW n Represents the temperature difference ratio of the nth collection point, BL n Represents the distance ratio of the nth collection point. If the ultrasonic intensity Q after rebound is in the interval [QL-λ, QL+λ], it is recorded as a living body model, and λ represents the preset fluctuation value.

4. The face recognition method based on artificial intelligence model according to claim 1, characterized in that: In the step S2, comparing the living body model with the database model to obtain comparison data includes: Get the collection points in the feature area, the feature area refers to the facial features and cheeks in the face model, get the red channel intensity R, green channel intensity G and blue channel intensity B of the collection points in the feature area, and calculate the comparison data , where H represents the grayscale value of the corresponding acquisition point.

5. The face recognition method based on artificial intelligence model according to claim 1, characterized in that: In the step 2, if the model is not a living body, the staff is prompted to check the object photographed by the camera.

6. The face recognition method based on artificial intelligence model according to claim 1, characterized in that: In step S3, if the first similar model does not exist in the database, the person is marked as an outsider and identity verification is performed.

7. The face recognition method based on artificial intelligence model according to claim 1, characterized in that: In the step S4, screening the second similar model based on the comparison point specifically includes: The comparison point of the living model is recorded as the reference point, and the distance S between the collection point in the first similar model and the reference point in the same feature area is calculated, and the mean value S of the distance is calculated. ave , if the mean distance S ave If the distance is less than a preset distance judgment value S, the corresponding first similarity model is recorded as the second similarity model.

8. The face recognition method based on artificial intelligence model according to claim 1, characterized in that: In the step S4, calculating the similarity K between the living body model and the second similar model specifically includes: Select the comparison points of the same part in the living model and the second similar model, connect the comparison points with the origin of the spatial coordinate system and generate a feature vector, the direction of the feature vector points from the origin to the comparison point, and calculate the similarity ,in, represents the i-th eigenvector in the living model, Represents the i-th eigenvector in the second similarity model.

Images