Face image matching method and device, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a face image matching method and device, equipment and a medium. According to the method and the device, the local face features of each face image in the to-be-matched face image and the screened face images are extracted, and the first feature vector of the to-be-matched face image and the second feature vector of each screened face image are constructed according to the local face features and the attitude angle recognition result; and comparing the first feature vector with the second feature vector, and determining a target face image similar to the to-be-matched face image in a preset face image database, so as to determine identity information of a person corresponding to the to-be-matched face image according to the target face image, and the corresponding target face image is determined directly according to the face local features, so that the matching efficiency of the to-be-matched face image is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and in particular, to a face image matching method, apparatus, device, and medium. Background Art

[0002] Wanted person search is an important part of security management, aiming to improve security management and social security levels and provide better support for protecting the safety of the people. Wanted person search takes person identification as the core, analyzes person characteristics, and focuses on monitoring the areas where the wanted persons abscond. In the prior art, face recognition is generally used to identify persons. During the face recognition process, generally, the face image of a person is matched with the face images in a face image database containing preset persons, and the recognition result of the person is determined according to the matching result. During the process of matching the face image of a person with the face images in the face image database containing preset persons, it mainly depends on the matching of full-face features, but the matching efficiency for persons wearing masks or having occlusions is relatively low. Therefore, how to improve the matching efficiency of the face images of persons with facial occlusions has become an urgent problem to be solved. Summary of the Invention

[0003] In view of this, the embodiments of the present invention provide a face image matching method, apparatus, device, and medium to solve the problem of relatively low matching efficiency of the face images of persons with facial occlusions during the face matching process.

[0004] In a first aspect, the embodiments of the present invention provide a face image matching method, and the face image matching method includes:

[0005] Obtain the pose angle recognition result of the face image to be matched, and filter the face images in the preset face image database according to the pose angle recognition result to obtain the filtered face images;

[0006] Determine N local features required for matching according to the pose angle recognition result, where N is an integer greater than zero;

[0007] Extract the N local features of the face image to be matched, and extract the feature results corresponding to the N local features of each face image in the filtered face images;

[0008] Construct a feature vector corresponding to each image according to the N local features of each extracted image, to obtain the first feature vector of the face image to be matched and the second feature vectors of each filtered face image;

[0009] Calculate the similarity between the first feature vector and each second feature vector to obtain a similarity value corresponding to each filtered face image, and determine the filtered face image corresponding to the similarity value greater than the preset similarity threshold as the target face image that matches the face image to be matched.

[0010] In a second aspect, an embodiment of the present invention provides a face image matching device, which includes:

[0011] A screening module, configured to obtain the pose angle recognition result of the face image to be matched, and screen the face images in the preset face image database according to the pose angle recognition result to obtain the filtered face images;

[0012] A determination module, configured to determine N local features required for matching according to the pose angle recognition result, where N is an integer greater than zero;

[0013] An extraction module, configured to extract the N local features of the face image to be matched, and extract the N local features of each face image in the filtered face images;

[0014] A construction module, configured to construct a feature vector corresponding to each image according to the N local features of each extracted image, and obtain the first feature vector of the face image to be matched and the second feature vectors of each filtered face image;

[0015] A calculation module, configured to calculate the similarity between the first feature vector and each second feature vector to obtain a similarity value corresponding to each filtered face image, and determine the filtered face image corresponding to the similarity value greater than the preset similarity threshold as the target face image that matches the face image to be matched.

[0016] In a third aspect, an embodiment of the present invention provides a computer device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the face image matching method described in the first aspect is implemented.

[0017] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the face image matching method described in the first aspect is implemented.

[0018] The beneficial effects of the present invention compared with the prior art are:

[0019] Obtain the pose angle recognition result of the face image to be matched. According to the pose angle recognition result, screen the face images in the preset face image database to obtain the screened face images. According to the pose angle recognition result, determine N local features required for matching, where N is an integer greater than zero. Extract the N local features of the face image to be matched, and extract the N local features of each face image in the screened face images. According to the N local features of each extracted image, construct a feature vector corresponding to each image, obtain the first feature vector of the face image to be matched and the second feature vector of each screened face image. Calculate the similarity between the first feature vector and each second feature vector to obtain the similarity value corresponding to each screened face image. Determine the screened face image corresponding to the similarity value greater than the preset similarity threshold as the target face image that matches the face image to be matched. In this application, by extracting the local face features of each face image in the face image to be matched and the screened face images, constructing the first feature vector of the face image to be matched and the second feature vector of each screened face image according to the local face features and the pose angle recognition result, and comparing the first feature vector with the second feature vector, determine the target face image in the preset face image database that is similar to the face image to be matched, so as to determine the identity information of the corresponding person of the face image to be matched according to the target face image, and directly determine the corresponding target face image according to the local face features, thereby improving the matching efficiency of the face image to be matched. Brief Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to these drawings.

[0021] Figure 1 It is a schematic diagram of an application environment of a face image matching method provided by an embodiment of the present invention;

[0022] Figure 2 It is a schematic flowchart of a face image matching method provided by an embodiment of the present invention;

[0023] Figure 3 It is a schematic structural diagram of a face image matching device provided by an embodiment of the present invention;

[0024] Figure 4 It is a schematic structural diagram of a computer device provided by an embodiment of the present invention. Detailed Description of the Embodiments

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system architectures and technologies are presented in order to thoroughly understand the embodiments of the present invention. However, those skilled in the art should understand that the present invention can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present invention.

[0027] It should be understood that when used in the specification and claims of the present invention, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0028] It should also be understood that the term "and / or" as used in the specification and claims of the present invention refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0029] As used in the specification and claims of the present invention, the term "if" may be interpreted as "when", "once", "in response to determining", or "in response to detecting" according to the context. Similarly, the phrase "if determined" or "if [the described condition or event] is detected" may be interpreted as meaning "once determined", "in response to determining", "once [the described condition or event] is detected", or "in response to detecting [the described condition or event]" according to the context.

[0030] In addition, in the description of the specification and claims of the present invention, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0031] References to "one embodiment" or "some embodiments" etc. described in the specification of the present invention mean that specific features, structures, or characteristics described in connection with that embodiment are included in one or more embodiments of the present invention. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized.

[0032] Embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results.

[0033] Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technologies mainly include several major directions such as computer vision technology, robotics, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0034] It should be understood that the magnitudes of the sequence numbers of the steps in the following embodiments do not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0035] In order to illustrate the technical solution of the present invention, specific embodiments will be used for illustration below.

[0036] A face image matching method provided by an embodiment of the present invention can be applied, for example, in Figure 1In the application environment, the client communicates with the server. The client includes, but is not limited to, computer devices such as a personal digital assistant (PDA), a tablet computer, a desktop computer, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a cloud computer device, etc. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0037] See Figure 2 , which is a schematic flowchart of a face image matching method provided by an embodiment of the present invention. The above face image matching method can be applied to Figure 1 the server in Figure 2 As shown in

[0038] S201: Obtain the pose angle recognition result of the face image to be matched. According to the pose angle recognition result, screen the face images in the preset face image database to obtain the screened face images.

[0039] In step S201, the face image to be matched is the face image of a randomly captured person. The pose angle recognition result of the face image to be matched includes the recognition result of the pitch angle, the recognition result of the yaw angle, and the recognition result of the roll angle. According to the pose angle recognition result, screen the face images in the preset face image database, and screen out the images with a small angle difference from the pose angle of the face image to be matched to obtain the screened face images. Among them, the face images in the preset face image database are the face images of preset persons, and the face images in the preset face image database are frontal face images without occlusion.

[0040] In this embodiment, a face image to be matched is obtained, and pose angle recognition is performed on the face image to be matched to obtain a pose angle recognition result, that is, the pitch angle, yaw angle, and roll angle of the face in the face image to be matched are recognized. When performing pose angle recognition on the face image to be matched, a pre-trained pose angle recognition model can be used for recognition. Among them, the pre-trained pose angle recognition model can be a convolutional neural network trained according to a preset first sample input set and a first sample output set. The pose angle recognition model can extract multiple feature maps of the face image to be matched, and then determine the pose angle of the face in the face image to be matched according to the multiple feature maps, and output the pose angle to obtain the pose angle recognition result of the face image to be matched. Through the pose angle recognition model, the pose angle can be obtained quickly and accurately.

[0041] It should be noted that the above pose angle recognition model can be a convolutional neural network, for example, it can include a convolutional layer, a feedback layer, a fully connected layer, and an output layer. First, the face image to be matched is input into the convolutional layer, and convolutional layer features are extracted from the face image to be matched through the convolutional layer, that is, a preset number of feature maps of the face image to be matched. Then, through the feedback layer, combining the previous feedback layer features and the next feedback layer features, the current feedback layer features are extracted from the preset number of feature maps output by the convolutional layer. After that, the fully connected layer performs an abstraction process on the feedback layer features to achieve feature fusion of the preset number of feature maps, so as to obtain the pose angle in the face image to be matched.

[0042] According to the pose angle recognition result, the face images in the preset face image database are screened to obtain the screened face images. When screening the face images in the preset face image database, images with a small difference in pose angle from the face image to be matched can be screened out. In this embodiment, when screening the face images in the preset face image database, the pose angle recognition result of the face image to be matched can be compared with the pose angle of each face image in the preset face image database, and the face image with a small difference between the pose angle of the face image in the preset face image database and the pose angle recognition result of the face image to be matched is selected as the screened face image. For example, when calculating the difference between the pose angle of the face image in the preset face image database and the pose angle recognition result of the face image to be matched, the pitch angle, yaw angle, and roll angle in the pose angle can be used as a three-dimensional coordinate, and the difference between the first three-dimensional coordinate of the pose angle recognition result and the second three-dimensional coordinate of the pose angle of the face image in the preset face image database is calculated. According to the difference, the face images in the preset face image database are screened to obtain the screened face images.

[0043] It should be noted that when obtaining the pose angle of each face image in the preset face image database, a pre-trained pose angle recognition model can also be used for pose angle recognition.

[0044] In this embodiment, the pose angle recognition result of the face image to be matched is obtained. According to the pose angle recognition result, the face images in the preset face image database are screened to obtain the screened face images. Screening the face images in the preset face image database through the pose angle recognition result of the face image to be matched is convenient for screening out images with pose angles not much different from those of the face image to be matched, thereby improving the accuracy of subsequent matching.

[0045] Optionally, after obtaining the pose angle recognition result of the face image to be matched, screening the face images in the preset face image database according to the pose angle recognition result to obtain the screened face images, further including:

[0046] Obtaining the age recognition result and gender recognition result of the face image to be matched;

[0047] According to the age recognition result and gender recognition result, screening the screened face images to obtain the finally screened face images, and determining the finally screened face images as the screened face images.

[0048] In this embodiment, when screening the face images in the preset face image database, the gender and age of the face image to be matched are also screened to obtain the screened face images. After screening according to the pose angle recognition result, screening the screened face images according to the gender recognition result and age recognition result of the face image to be matched. When screening the screened face images according to the gender recognition result of the face image to be matched, face images with the same gender as the face image to be matched can be screened out. When screening the screened face images according to the age recognition result of the face image to be matched, face images with ages close to that of the face image to be matched can be screened out.

[0049] It should be noted that before screening the filtered face images according to the gender recognition result and age recognition result of the face image to be matched, the gender and age of the face image to be matched are recognized. When recognizing the gender of the face image to be matched, a trained gender classification network model can be used for recognition. For example, the trained gender classification network model can be a gender classification network constructed based on the FaceNet network, adding a classification layer and designing a triple sample generation rule. The input of the trained gender classification network model is the face image to be matched. The network completes the mapping, obtains the face embedding feature, and outputs the corresponding gender. In this embodiment, a classification layer is added at the end of the FaceNet network to construct a gender classification network model. Among them, the classification layer can include one convolutional layer, one pooling layer, and one fully connected layer, so as to convert the 128-dimensional face embedding feature output by the FaceNet network into a gender classification result, and obtain the gender recognition result of the face image to be matched.

[0050] After screening the filtered face images according to the gender recognition result of the face image to be matched, face images with the same gender recognition result as the face image to be matched are obtained. Then, according to the age recognition result of the face image to be matched, the filtered face images are screened to select face images with an age recognition result close to that of the face image to be matched.

[0051] It should be noted that when identifying the age of a face image to be matched, the feature information of the face image to be matched is extracted. The feature information is the information that can reflect the characteristics of the face image to be matched. The feature information of different face images to be matched is different, and the unique face image to be matched can be determined through the feature information. Exemplarily, the feature information can be a feature vector, and the feature vector is extracted from the face image to be matched through a certain data processing algorithm or with the help of a machine learning model. According to the feature information, the first age corresponding to the male of the face image to be matched and the second age corresponding to the female of the face image to be matched are determined respectively. Considering the influence of gender on determining the age of the face image to be matched, the embodiment of the present application can consider the gender factor when determining the age of the face image to be matched. When the gender of the face image to be matched is taken as male, the first age is obtained, and when the gender of the face image to be matched is taken as female, the second age is obtained. By determining the first age and the second age, the gender factor is introduced when calculating the age, providing a data basis for the subsequent age prediction of the face image to be matched. According to the first age corresponding to the male of the face image to be matched and the second age corresponding to the female of the face image to be matched, the age corresponding to the face image to be matched is determined, and the age recognition result of the face image to be matched is obtained. According to the gender corresponding to the face image to be matched, the first age corresponding to the male of the face image to be matched, and the second age corresponding to the female of the face image to be matched, the age corresponding to the face image to be matched is determined. By using the gender corresponding to the face image to be matched to assist in determining the age corresponding to the face image to be matched, the influence of gender on age recognition is fully considered, thereby improving the accuracy of age recognition.

[0052] Optionally, obtain the pose angle recognition result of the face image to be matched. According to the pose angle recognition result, screen the face images in the preset face image database to obtain the screened face images, including:

[0053] Obtain the recognition results of the first pose angle, the second pose angle, and the third pose angle in the pose angle recognition result, and the angle values of the first pose angle, the second pose angle, and the third pose angle of each face image in the preset face image database;

[0054] Calculate the first difference between the recognition result of the first pose angle and the angle value of the first pose angle of each face image in the preset face image database;

[0055] Calculate the second difference between the recognition result of the second pose angle and the angle value of the second pose angle of each face image in the preset face image database;

[0056] Calculate the third difference between the recognition result of the third pose angle and the angle value of the third pose angle of each face image in the preset face image database;

[0057] In the preset face image database, determine the face images with the first difference less than the first preset threshold, the second difference less than the first preset threshold, and the third difference less than the first preset threshold as the filtered face images.

[0058] In this embodiment, obtain the recognition results of the first pose angle, the second pose angle, and the third pose angle in the pose angle recognition result, where the first pose angle, the second pose angle, and the third pose angle are respectively one of the pitch angle, the yaw angle, and the roll angle. For example, the first pose angle is the pitch angle, the second pose angle is the yaw angle, and the third pose angle is the roll angle.

[0059] When obtaining the recognition results of the first pose angle, the second pose angle, and the third pose angle in the pose angle recognition result, and the angle values of the first pose angle, the second pose angle, and the third pose angle of each face image in the preset face image database, a pre-trained pose angle recognition model can be used for recognition. Calculate the first difference between the recognition result of the first pose angle and the angle value of the first pose angle of each face image in the preset face image database, calculate the second difference between the recognition result of the second pose angle and the angle value of the second pose angle of each face image in the preset face image database, calculate the third difference between the recognition result of the third pose angle and the angle value of the third pose angle of each face image in the preset face image database, and determine whether to filter the corresponding face image in the preset face image database according to the first difference, the second difference, and the third difference. Determine the face images in the preset face image database with the first difference less than the first preset threshold, the second difference less than the first preset threshold, and the third difference less than the first preset threshold as the filtered face images, that is, the pose angles in each direction of the corresponding face image in the preset face image database and the pose angles in the corresponding direction of the face image to be matched are not very different. The first preset threshold in this embodiment can be 10 degrees or other angular values, which is not limited in this embodiment.

[0060] In this embodiment, by comparing the differences between the pose angles in each direction of the corresponding face image in the preset face image database and the pose angles in the corresponding direction of the face image to be matched, determine whether to filter the face image in the preset face image database. Compare multiple pose angles respectively, so as to improve the accuracy of filtering the corresponding face image in the preset face image database, obtain a filtered face image closer to the face image to be matched, and use the filtered face image to match the face image to be matched, avoiding the matching of face images with large pose angle differences and the face image to be matched, thereby improving the subsequent matching efficiency.

[0061] S202: Determine N local features required for matching according to the pose angle recognition result, where N is an integer greater than zero.

[0062] In step S202, according to the pose angle recognition result, determine N local features required for matching. Among them, for different pose angle recognition results, the local features required for matching are different. The N local features are the human face features that can be extracted from the human face image under the condition of facial occlusion. The N local features required for matching are the human face features required when the face image to be matched is matched with the filtered face image.

[0063] In this embodiment, according to the pose angle recognition result, determine N local features required for matching. For different pose angle recognition results, the local features required for matching are different, that is, for different pose angle recognition results, the number or the local features of the local features extracted from the corresponding image are different.

[0064] According to the coordinates of the key points extracted, calculate the feature results corresponding to the local features. For example, according to the key point coordinates of the eye part, the distance between the two eyes can be calculated to obtain the distance value between the two eyes. Among them, the distance between the two eyes is one of the N local features, and the distance value between the two eyes is the feature result corresponding to the local feature.

[0065] In this embodiment, according to the pose angle recognition result, determine N local features required for matching, so that for different pose angle recognition results, different local features are extracted, so that the extracted local features can better represent the features of the human face image under the pose angle recognition result, and thus the similarity calculation can be based on the local feature, improving the accuracy of the similarity calculation.

[0066] Optionally, determining N local features required for matching according to the pose angle recognition result includes:

[0067] If the recognition result of at least one pose angle in the pose angle recognition result is within the first interval range, determine at least one of the local features corresponding to the position of the center point of the ear in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead.

[0068] In this embodiment, the attitude angle recognition result is judged. The attitude angle recognition result includes the recognition result of the first attitude angle, the recognition result of the second attitude angle, and the recognition result of the third attitude angle. The first attitude angle is the pitch angle, the second attitude angle is the yaw angle, and the third attitude angle is the roll angle. One of the attitude angles is judged. If the recognition result of at least one attitude angle in the attitude angle recognition result is within the first interval range, at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead is determined. The first interval range is interval.

[0069] For example, if the yaw angle in the attitude angle recognition result is within the first interval range, and the first interval range is interval, at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead in the face image to be matched is extracted, and for the face images with the yaw angle within the interval in the filtered face images, at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead is extracted. The interval range of the corresponding yaw angle can also be set to other interval ranges, which is not limited in this embodiment.

[0070] It should be noted that the features and the number of local features of the local features extracted from the face image to be matched should be equal to the features and the number of local features of the local features in the face images with the yaw angle within the interval in the filtered face images. For example, if the local features extracted from the face image to be matched are the distance between the Adam's apple and the chin and the distance between the two ears, the local features in the face images with the yaw angle within the interval in the filtered face images are also the distance between the Adam's apple and the chin and the distance between the two ears. This is convenient for calculating the similarity using the same local features.

[0071] It should be noted that the face detection box in the position of the ear center point in the face detection box is the face detection box obtained when performing face detection on each image. The ear center point includes the center point of the left ear and the center point of the right ear. The distance between the Adam's apple and the chin can be determined according to the distance between the detected Adam's apple key point and the chin key point. The distance between the two ears can be determined according to the distance between the key points of the two ear center points. The distance between the two eyes can be determined according to the distance between the key points of the two eyes. The distance from the eyebrows to the forehead can be determined according to the distance between the eyebrow key point and the forehead key point. The distance from the ears to the tip of the nose can be determined according to the distance between the ear key point and the tip of the nose key point.

[0072] It should be noted that when performing local feature extraction on each image, local feature extraction is performed according to the recognition result of the pose angle of the face in each image. When extracting the position of the ear center point in the face detection box, if the face in the face image is a left face, the center point of the left ear is extracted. If the face in the face image is a right face, the center point of the right ear is extracted. If the face in the face image is a front face and the center points of both ears can be detected, the center point of the left ear and the center point of the right ear are extracted respectively.

[0073] It should be noted that when extracting the key points at the corresponding positions in the face image, if there are multiple key points at the corresponding positions, for example, when extracting the key points of the eyes, each eye includes multiple key points. When calculating the distance between the two eyes, the distance between the two eyes can be calculated according to the mean value of the coordinates of the multiple key points in each eye.

[0074] In this embodiment, if the recognition result of at least one pose angle in the pose angle recognition result is within the first interval range, at least one of the local features corresponding to the position of the ear center point in the face detection box, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead is determined. Among them, in the first interval range, the corresponding face image is a front face image, that is, the facial features on the left side of the face image and the facial features on the right side of the face image can be detected respectively. For example, the features of two ears and two eyes can be obtained, and more local features can be obtained. The similarity between the face image to be matched and the corresponding filtered face image is calculated according to the corresponding local features, thereby improving the calculation accuracy of the similarity.

[0075] In another embodiment, if the roll angle in the pose angle recognition result is within the first interval range, where the first interval range is of the interval, at least one of the local features corresponding to the position of the ear center point in the face detection box, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead in the image to be matched is extracted, and for the filtered face image, the roll angle is within Extract at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead within the interval. The interval range of the corresponding roll angle can also be set to other interval ranges, which is not limited in this embodiment.

[0076] Optionally, determining the N local features required for matching according to the pose angle recognition result further includes:

[0077] If the recognition result of at least one pose angle in the pose angle recognition result is within the second interval range, determine at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose.

[0078] In this embodiment, the pose angle recognition result is judged. Among them, the pose angle recognition result includes the recognition result of the first pose angle, the recognition result of the second pose angle, and the recognition result of the third pose angle. Among them, the first pose angle is the pitch angle, the second pose angle is the yaw angle, and the third pose angle is the roll angle. Judging one of the pose angles, if the recognition result of at least one pose angle in the pose angle recognition result is within the second interval range, determine the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose, where the second interval range is of the interval.

[0079] For example, if the yaw angle in the pose angle recognition result is within the second interval range, where the second interval range is of the interval, extract at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose, and for the face images with the yaw angle within of the interval in the filtered face images, extract at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose. The interval range of the corresponding yaw angle can also be set to other interval ranges, which is not limited in this embodiment.

[0080] It should be noted that the features and the number of local features of the face images to be matched extracted should be equal to the features and the number of local features of the face images with the yaw angle within of the interval in the filtered face images. For example, if the local features extracted from the face images to be matched are the distance between the Adam's apple and the chin and the distance from the ear to the tip of the nose, then for the face images with the yaw angle within The local features in the face image within the interval are also the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose. This facilitates calculating the similarity using the same local features.

[0081] It should be noted that the face detection box in the position of the center point of the ear in the face detection box is the face detection box obtained when performing face detection on each image. The distance between the Adam's apple and the chin can be determined according to the distance between the detected Adam's apple key point and the chin key point, and the distance from the ear to the tip of the nose is determined according to the distance between the detected ear key point and the tip of the nose key point.

[0082] In this embodiment, if the recognition result of at least one pose angle in the pose angle recognition result is within the second interval range, at least one of the local features corresponding to the position of the center point of the ear in the face detection box, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose is determined. The face image within the second interval range is a side face image. When extracting the local features of the side face image, the local features contained in the side face image are fully extracted, and a feature vector is constructed according to the local features, thereby improving the accuracy of constructing the feature vector.

[0083] In another embodiment, if the roll angle in the pose angle recognition result is within the second interval range, where the second interval range is the interval, at least one of the local features corresponding to the position of the center point of the ear in the face detection box, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose is extracted, and for the face image with the roll angle within the interval in the filtered face image, at least one of the local features corresponding to the position of the center point of the ear in the face detection box, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose is extracted. The interval range of the corresponding roll angle can also be set to other interval ranges, which is not limited in this embodiment.

[0084] Optionally, determining the N local features required for matching according to the pose angle recognition result further includes:

[0085] If the recognition result of at least one pose angle in the pose angle recognition result is within the third interval range, at least one of the local features corresponding to the position of the center point of the ear in the face detection box, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose is determined.

[0086] In this embodiment, the attitude angle recognition results are judged. The attitude angle recognition results include the recognition results of the first attitude angle, the second attitude angle, and the third attitude angle. Among them, the first attitude angle is the pitch angle, the second attitude angle is the yaw angle, and the third attitude angle is the roll angle. Judge one of the attitude angles. If the recognition result of at least one attitude angle in the attitude angle recognition results is within the third interval range, determine at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose. The third interval range is the interval of.

[0087] For example, if the yaw angle in the attitude angle recognition results is within the third interval range, and the second interval range is the interval of, extract at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose, and for the face images with the yaw angle within the interval of in the filtered face images, extract at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose. The interval range of the corresponding yaw angle can also be set to other interval ranges, which is not limited in this embodiment.

[0088] It should be noted that the features and the number of local features of the local features extracted from the face images to be matched should be equal to the features and the number of local features of the local features in the face images with the yaw angle within the interval of in the filtered face images. For example, if the local features extracted from the face images to be matched are the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose, then the local features in the face images with the yaw angle within the interval of in the filtered face images are also the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose. This is convenient for calculating the similarity using the same local features.

[0089] In this embodiment, if the recognition result of at least one attitude angle in the attitude angle recognition results is within the third interval range, determine at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose. The face images within the third interval range are side face images. When extracting the local features of the side face images, fully extract the local features included in the side face images, and construct a feature vector according to the local features, thereby improving the accuracy of constructing the feature vector.

[0090] In another embodiment, if the roll angle in the attitude angle recognition results is within the second interval range, and the third interval range is In the interval, extract at least one of the local features corresponding to the position of the center point of the ear in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose, and for the face images with the roll angle in the In the interval, extract at least one of the local features corresponding to the position of the center point of the ear in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose. The interval range of the corresponding roll angle can also be set to other interval ranges, which is not limited in this embodiment.

[0091] S203: Extract N local features of the face image to be matched, and extract N local features of each face image in the filtered face images.

[0092] In step S203, extract N local features of the corresponding images in the face image to be matched and the filtered face images respectively.

[0093] In this embodiment, the corresponding local features can be determined according to the key points of each extracted image. When extracting the key points of each image, a pre-trained key point recognition model can be used for extraction. The pre-trained key point recognition model is a convolutional neural network trained according to a pre-set second sample input set and a second sample output set. The key point recognition model can extract multiple feature maps of each image, and then determine the key points of the face in each image according to the multiple feature maps, and output the key points. Among them, the key points can be, for example, the key point coordinates of parts such as eyebrows, eyes, mouth, nose, ears, etc. It should be noted that the above convolutional neural network is only an example of the key point recognition model, and the present disclosure is not limited thereto, and may also include various other neural networks.

[0094] S204: According to the N local features of each extracted image, construct a feature vector corresponding to each image, and obtain the first feature vector of the face image to be matched and the second feature vector of each filtered face image.

[0095] In step S204, according to the N local features of each extracted image, construct a feature vector corresponding to each image. Among them, when constructing the feature vector of each image, it can be constructed according to the pose recognition result of the face image in the corresponding image. Different pose recognition results use different local features for constructing the corresponding feature vector. Construct a feature vector corresponding to each image, and obtain the first feature vector of the face image to be matched and the second feature vector of each filtered face image.

[0096] In this embodiment, according to the N local features of each extracted image, a feature vector corresponding to each image is constructed. When constructing the feature vector, the corresponding local features can be sorted in sequence, that is, each dimensional vector in the feature vector is an equal local feature. For example, if the recognition result of at least one pose angle in the pose angle recognition result is within the first interval range, the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead are determined. The local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead can be sorted in sequence. That is, the first vector dimension in the feature vector is the local feature of the position of the ear center point in the face detection frame, the second vector dimension in the feature vector is the local feature of the distance between the Adam's apple and the chin, the third vector dimension in the feature vector is the local feature of the distance between the two ears, the fourth vector dimension in the feature vector is the local feature of the distance between the two eyes, and the fifth vector dimension in the feature vector is the local feature of the distance from the eyebrows to the forehead. According to the corresponding order, the first feature vector of the image to be matched and the second feature vector of the filtered face image are determined in sequence.

[0097] It should be noted that before constructing the common feature vector and the second feature vector, the corresponding local features can also be normalized to facilitate converting the corresponding local features into values under the same dimension, which is beneficial to calculating the similarity value corresponding to the first feature vector and the second feature vector.

[0098] In this embodiment, a feature vector is constructed according to the corresponding local features to facilitate calculating the similarity between the face image to be matched and the filtered face image, simplifying the calculation process of the similarity between local features, and thus improving the efficiency of calculating the similarity between the face image to be matched and the filtered face image.

[0099] S205: Calculate the similarity between the first feature vector and each second feature vector to obtain the similarity value corresponding to each filtered face image. The filtered face image corresponding to the similarity value greater than the preset similarity threshold in the similarity values is determined as the target face image matching the face image to be matched.

[0100] In step S205, calculate the similarity between the first feature vector and each second feature vector to obtain a similarity value corresponding to each filtered face image. Among them, the larger the similarity value, the greater the similarity between the first feature vector and each second feature vector; the smaller the similarity value, the smaller the similarity between the first feature vector and each second feature vector. Determine the filtered face images corresponding to the similarity values greater than the preset similarity threshold as the target face images that match the face image to be matched.

[0101] In this embodiment, when calculating the similarity between the first feature vector and each second feature vector, the cosine formula can be used to calculate the cosine similarity to obtain the similarity value. Determine the filtered face images corresponding to the similarity values greater than the preset similarity threshold as the target face images that match the face image to be matched. Among them, the preset similarity threshold is the similarity value of two face images of the same person that have been pre-trained.

[0102] It should be noted that the preset similarity threshold is obtained through a trained model. Input different face images of the same person into the model, and output the corresponding similarity threshold results. Among them, the similarity threshold results include the similarity thresholds corresponding to different interval ranges of the pose angle recognition results. For example, if the recognition result of a pose angle in the pose angle recognition result is in the first interval range, output the corresponding first similarity threshold, that is, the first feature vector and the second feature vector constructed based on the corresponding local features. Compare the similarity value between the first feature vector and the second feature vector with the first similarity threshold. If the recognition result of a pose angle in the pose angle recognition result is in the second interval range, output the corresponding second similarity threshold, that is, the first feature vector and the second feature vector constructed based on the corresponding local features. Compare the similarity value between the first feature vector and the second feature vector with the second similarity threshold. If the recognition result of a pose angle in the pose angle recognition result is in the third interval range, output the corresponding third similarity threshold, that is, the first feature vector and the second feature vector constructed based on the corresponding local features. Compare the similarity value between the first feature vector and the second feature vector with the third similarity threshold.

[0103] It should be noted that the corresponding first similarity threshold, second similarity threshold, and third similarity threshold may be equal or not equal. In this embodiment, the first similarity threshold, second similarity threshold, and third similarity threshold are taken to be equal, which is 0.9. That is, determine the filtered face images corresponding to the similarity values greater than 0.9 as the target face images that match the face image to be matched.

[0104] In this embodiment, the similarity value between the first feature vector and the second feature vector is used to determine the target face image that matches the face image to be matched. By using the feature vectors constructed from the local features of the face image for matching, it is not necessary to extract the global face features of the face image for matching, which improves the matching efficiency.

[0105] Obtain the pose angle recognition result of the face image to be matched. According to the pose angle recognition result, screen the face images in the preset face image database to obtain the screened face images. According to the pose angle recognition result, determine N local features required for matching, where N is an integer greater than zero. Extract the N local features of the face image to be matched, and extract the N local features of each face image in the screened face images. According to the N local features of each extracted image, construct the feature vector corresponding to each image, obtain the first feature vector of the face image to be matched and the second feature vector of each screened face image, calculate the similarity between the first feature vector and each second feature vector, obtain the similarity value corresponding to each screened face image, and determine the screened face image corresponding to the similarity value greater than the preset similarity threshold as the target face image that matches the face image to be matched. In this application, by extracting the local face features of each face image in the face image to be matched and the screened face images, constructing the first feature vector of the face image to be matched and the second feature vector of each screened face image according to the local face features and the pose angle recognition result, and comparing the first feature vector with the second feature vector, the target face image in the preset face image database similar to the face image to be matched is determined, so as to determine the identity information of the corresponding person of the face image to be matched according to the target face image, and directly determine the corresponding target face image according to the local face features, thereby improving the matching efficiency of the face image to be matched.

[0106] Optionally, after determining the screened face image corresponding to the similarity value greater than the preset similarity threshold as the target face image that matches the face image to be matched, it further includes:

[0107] According to the target face image, determine the identity information of the person corresponding to the target face image;

[0108] According to the identity information, extract the activity track of the person corresponding to the target face image;

[0109] According to the activity track, determine whether the person corresponding to the face image to be matched is the target person.

[0110] In this embodiment, based on the target face image, the identity information of the person corresponding to the target face image is determined, where the identity information is the identification information of the target face image pre - saved. Based on the identity information, the activity track of the person corresponding to the target face image is extracted. The activity track is the movement track within a specific time range. The location where the person is located within the specific time range can be found according to the corresponding identity information, and then the image of the person corresponding to the target face image is collected by the camera device at the location, and the person corresponding to the target face image is tracked, so as to determine the activity track of the person corresponding to the target face image. Whether the person corresponding to the face image to be matched is the target person is judged according to the activity track.

[0111] In this embodiment, after the target face image is determined, an investigation and monitoring are carried out on the person corresponding to the target face image to determine the activity track of the person corresponding to the target face image. The target face image is a clear face image, and the corresponding identity information can be obtained. The corresponding activity track is extracted according to the corresponding identity information. Whether the person corresponding to the face image to be matched is the target person is judged according to the activity track, which improves the accuracy of determining whether the person corresponding to the face image to be matched is the target person.

[0112] Optionally, judging whether the person corresponding to the face image to be matched is the target person according to the activity track includes:

[0113] Obtain the target movement track of the person corresponding to the face image to be matched;

[0114] Calculate the coincidence rate between the target movement track and the activity track;

[0115] If the coincidence rate is greater than the preset coincidence rate threshold, then determine the person corresponding to the face image to be matched as the target person.

[0116] In this embodiment, when judging whether the person corresponding to the face image to be matched is the target person according to the activity track, it can be determined according to the coincidence rate between the target movement track of the person corresponding to the face image to be matched and the activity track. When calculating the coincidence rate between the target movement track and the activity track, the target movement track and the activity track can be calculated in segments, which can be segmented according to time, and the coincidence rate within each time segment is calculated. If the coincidence rate within each time segment is greater than the preset coincidence rate threshold, then determine the person corresponding to the face image to be matched as the target person.

[0117] In this embodiment, the coincidence rate between the target movement track and the activity track is calculated. If the coincidence rate is greater than the preset coincidence rate threshold, then determine the person corresponding to the face image to be matched as the target person. Using the track coincidence rate to determine whether it is the target person increases the judgment conditions for the target person, thereby improving the judgment accuracy.

[0118] Please refer to Figure 3 , Figure 3It is a schematic structural diagram of a face image matching device provided by an embodiment of the present invention. In this embodiment, each unit included in the terminal is used to execute Figure 2 each step in the corresponding embodiment. For details, please refer to Figure 2 the relevant descriptions in the corresponding embodiment. For the sake of convenience of description, only the parts related to this embodiment are shown. Refer to Figure 3 , the face image matching device 30 includes: a screening module 31, a determination module 32, an extraction module 33, a construction module 34, and a calculation module 35.

[0119] The screening module 31 is used to obtain the pose angle recognition result of the face image to be matched, and screen the face images in the preset face image database according to the pose angle recognition result to obtain the screened face images.

[0120] The determination module 32 is used to determine N local features required for matching according to the pose angle recognition result;

[0121] The extraction module 33 is used to extract N local features of the face image to be matched, and extract N local features of each face image in the screened face images, where N is an integer greater than zero.

[0122] The construction module 34 is used to construct a feature vector corresponding to each image according to the N local features of each extracted image, and obtain the first feature vector of the face image to be matched and the second feature vectors of each screened face image.

[0123] The calculation module 35 is used to calculate the similarity between the first feature vector and each second feature vector, obtain the similarity value corresponding to each screened face image, and determine the screened face image corresponding to the similarity value greater than the preset similarity threshold as the target face image that matches the face image to be matched.

[0124] Optionally, the above screening module 31 includes:

[0125] An acquisition unit, configured to acquire the recognition result of the first pose angle, the recognition result of the second pose angle, and the recognition result of the third pose angle in the pose angle recognition result, as well as the angle values of the first pose angle, the second pose angle, and the third pose angle of each face image in the preset face image database.

[0126] A first calculation unit, configured to calculate a first difference between the recognition result of the first pose angle and the angle value of the first pose angle of each face image in the preset face image database.

[0127] A second calculation unit, configured to calculate a second difference between the recognition result of the second pose angle and the angle value of the second pose angle of each face image in the preset face image database.

[0128] A third calculation unit, configured to calculate a third difference between the recognition result of the third pose angle and the angle value of the third pose angle of each face image in a preset face image database.

[0129] Determine the face images with the first difference less than the first preset threshold, the second difference less than the first preset threshold, and the third difference less than the first preset threshold as the filtered face images.

[0130] Optionally, the above-mentioned face image matching device 30 further includes:

[0131] An acquisition module, configured to acquire the age recognition result and gender recognition result of the face image to be matched;

[0132] A gender and age screening module, configured to screen the filtered face images according to the age recognition result and gender recognition result to obtain the final filtered face images, and determine the final filtered face images as the filtered face images.

[0133] Optionally, the above-mentioned determination module 32 includes:

[0134] A first determination unit, configured to determine at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead if the recognition result of at least one pose angle in the pose angle recognition result is within the first interval range.

[0135] Optionally, the above-mentioned determination module 32 includes:

[0136] A second determination unit, configured to determine at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose if the recognition result of at least one pose angle in the pose angle recognition result is within the second interval range.

[0137] Optionally, the above-mentioned face image matching device 30 further includes:

[0138] A determination module, configured to determine the identity information of the person corresponding to the target face image according to the target face image.

[0139] A trajectory extraction module, configured to extract the activity trajectory of the person corresponding to the target face image according to the identity information.

[0140] A judgment module, configured to judge whether the person corresponding to the face image to be matched is the target person according to the activity trajectory.

[0141] Optionally, the above-mentioned judgment module includes:

[0142] A trajectory acquisition unit, configured to acquire a target motion trajectory of a person corresponding to a face image to be matched.

[0143] A calculation unit, configured to calculate a coincidence rate between the target motion trajectory and an activity trajectory;

[0144] A determination unit, configured to, if the coincidence rate is greater than a preset coincidence rate threshold, determine the person corresponding to the face image to be matched as the target person.

[0145] It should be noted that for the information interaction, execution process, etc. among the above modules, units, and subunits, since they are based on the same concept as the method embodiment of the present invention, their specific functions and the technical effects brought can be specifically referred to the method embodiment part, and will not be elaborated here.

[0146] Figure 4 This is a schematic structural diagram of a computer device provided by an embodiment of the present invention. As Figure 4 shown, the computer device of this embodiment includes: at least one processor ( Figure 4 only one is shown in the figure), a memory, and a computer program stored in the memory and executable on at least one processor. When the processor executes the computer program, it implements the steps in any of the above method embodiments of the face image matching method.

[0147] The computer device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that Figure 4 this is only an example of a computer device and does not constitute a limitation on the computer device. The computer device may include more or fewer components than shown in the figure, or combine some components, or different components.

[0148] The so-called processor may be a CPU, and the processor may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), field-programmable gate arrays (Field-Programmable Gate Array, FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0149] The memory includes a readable storage medium, an internal memory, etc. Among them, the internal memory can be the memory of a computer device, and the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The readable storage medium can be the hard disk of a computer device, and in some other embodiments, it can also be an external storage device of a computer device. For example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device. Further, the memory can also include both the internal storage unit of the computer device and an external storage device. The memory is used to store an operating system, application programs, a BootLoader, data, and other programs, etc. The other programs such as the program code of a computer program, etc. The memory can also be used to temporarily store the data that has been output or will be output.

[0150] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of the present invention. The specific working processes of the units and modules in the above device can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above method embodiments of the present invention, a computer program can be used to instruct the relevant hardware to complete. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can at least include: any entity or device that can carry the computer program code, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disc, etc. In some jurisdictions, according to legislation and patent practice, the computer-readable medium cannot be an electrical carrier signal and a telecommunication signal.

[0151] All or part of the processes in the above method embodiments of the present invention can also be completed by a computer program product. When the computer program product runs on a computer device, the computer device can be made to execute the steps in the above method embodiments.

[0152] In the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0153] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0154] In the embodiments provided by the present invention, it should be understood that the disclosed device / computer equipment and method can be implemented in other ways. For example, the device / computer equipment embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical or other forms.

[0155] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0156] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A face image matching method, characterized in that, The described face image matching method includes: Obtaining the pose angle recognition result of the face image to be matched, and screening the face images in the preset face image database according to the pose angle recognition result to obtain the screened face images; Determining N local features required for matching according to the pose angle recognition result, where N is an integer greater than zero; Extracting the N local features of the face image to be matched, and extracting the N local features of each face image in the screened face images; Constructing a feature vector corresponding to each image according to the N local features of each extracted image, to obtain the first feature vector of the face image to be matched and the second feature vectors of each screened face image; Calculating the similarity between the first feature vector and each second feature vector to obtain the similarity value corresponding to each screened face image, and determining the screened face image corresponding to the similarity value greater than the preset similarity threshold as the target face image that matches the face image to be matched.

2. The face image matching method according to claim 1, characterized in that The obtaining the pose angle recognition result of the face image to be matched, and screening the face images in the preset face image database according to the pose angle recognition result to obtain the screened face images includes: Obtaining the recognition result of the first pose angle, the recognition result of the second pose angle, and the recognition result of the third pose angle in the pose angle recognition result, as well as the angle value of the first pose angle, the angle value of the second pose angle, and the angle value of the third pose angle of each face image in the preset face image database; Calculating a first difference between the recognition result of the first pose angle and the angle value of the first pose angle of each face image in the preset face image database; Calculating a second difference between the recognition result of the second pose angle and the angle value of the second pose angle of each face image in the preset face image database; Calculating a third difference between the recognition result of the third pose angle and the angle value of the third pose angle of each face image in the preset face image database; Determining the face images in the preset face image database for which the first difference is less than the first preset threshold, the second difference is less than the first preset threshold, and the third difference is less than the first preset threshold as the screened face images.

3. The face image matching method according to claim 1, wherein After obtaining the pose angle recognition result of the face image to be matched, and screening the face images in the preset face image database according to the pose angle recognition result to obtain the screened face images, it further includes: Obtaining the age recognition result and gender recognition result of the face image to be matched; Screening the screened face images according to the age recognition result and the gender recognition result to obtain the finally screened face images, and determining the finally screened face images as the screened face images.

4. The face image matching method according to claim 1, wherein The determining N local features required for matching according to the pose angle recognition result includes: If the recognition result of at least one of the pose angles in the pose angle recognition result is within the first interval range, determine at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, the distance between the two ears, the distance between the two eyes, and the distance from the eyebrows to the forehead.

5. The face image matching method according to claim 1, wherein The determining the N local features required for matching according to the pose angle recognition result further includes: If the recognition result of at least one of the pose angles in the pose angle recognition result is within the second interval range or the third interval range, determine at least one of the local features corresponding to the position of the ear center point in the face detection frame, the distance between the Adam's apple and the chin, and the distance from the ear to the tip of the nose.

6. The face image matching method according to claim 1, wherein, After determining the filtered face images corresponding to the similarity values greater than the preset similarity threshold among the similarity values as the target face images matching the to-be-matched face image, it further includes: Determine the identity information of the person corresponding to the target face image according to the target face image; Extract the activity track of the person corresponding to the target face image according to the identity information; Judge whether the person corresponding to the to-be-matched face image is the target person according to the activity track.

7. The face image matching method according to claim 6, wherein, The judging whether the person corresponding to the to-be-matched face image is the target person according to the activity track includes: Obtain the target movement track of the person corresponding to the to-be-matched face image; Calculate the coincidence rate between the target movement track and the activity track; If the coincidence rate is greater than the preset coincidence rate threshold, determine the person corresponding to the to-be-matched face image as the target person.

8. A face image matching device, characterized in that, The face image matching device includes: A screening module, configured to obtain the pose angle recognition result of the to-be-matched face image, and screen the face images in the preset face image database according to the pose angle recognition result to obtain the filtered face images; A determining module, configured to determine the N local features required for matching according to the pose angle recognition result, where N is an integer greater than zero; An extraction module, configured to extract the N local features of the to-be-matched face image, and extract the N local features of each face image in the filtered face images; A construction module, configured to construct a feature vector corresponding to each image according to the N local features of each extracted image, to obtain a first feature vector of the to-be-matched face image and a second feature vector of each filtered face image; A calculation module, configured to calculate the similarity between the first feature vector and each second feature vector, to obtain a similarity value corresponding to each filtered face image, and determine the filtered face images corresponding to the similarity values greater than the preset similarity threshold as the target face images matching the to-be-matched face image.

9. A computer device, characterized in that, The computer device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the face image matching method according to claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the face image matching method according to claims 1 to 7.

Images