WO2020038136A1 - 面部识别方法、装置、电子设备及计算机可读介质 - Google Patents
面部识别方法、装置、电子设备及计算机可读介质 Download PDFInfo
- Publication number
- WO2020038136A1 WO2020038136A1 PCT/CN2019/095286 CN2019095286W WO2020038136A1 WO 2020038136 A1 WO2020038136 A1 WO 2020038136A1 CN 2019095286 W CN2019095286 W CN 2019095286W WO 2020038136 A1 WO2020038136 A1 WO 2020038136A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- similarity
- pictures
- similarities
- database
- facial
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Definitions
- the present application relates to the technical field of computer information processing, and in particular, to a facial recognition method, device, electronic device, and computer-readable medium.
- Face recognition technology is a biometric recognition technology based on human face feature information for identification in recent years. It is also currently the most widely used artificial intelligence technology based on deep learning. Compared with other biometric technologies, face recognition has many advantages such as friendliness, simplicity, accuracy, economy, and good scalability. It can be widely used in many aspects such as security verification, monitoring, and access control. At present, face recognition technology has been Apply to access control attendance, visitor management, patrol and other places.
- face recognition has at least the following problems: When performing face recognition in natural scenes, when the preset face feature database contains a large number of face features (such as millions of faces) Library), affected by the lighting, resolution, blurring, angle and other factors of the collected face image. Due to the large sample space of the face feature comparison, the possibility of high similarity of face features increases, so, The problem of reduced recognition accuracy will occur in the actual face recognition process
- the purpose of the embodiments of the present application is to provide a facial recognition method, device, electronic device, and computer-readable medium, which can quickly and accurately perform facial feature recognition on a human face in the case of massive data, and output a recognition result.
- a face recognition method includes: comparing a face image to be recognized with a plurality of first pictures in a database, and obtaining a plurality of first similarities; When the largest first similarity in the first similarity is within the first similarity threshold, a part of the first similarities in the plurality of first similarities are extracted; a plurality of first similarities corresponding to the partial first similarities are determined to be multiple Two second pictures; comparing a second facial image to be recognized with a plurality of second pictures to obtain a plurality of second similarities; and determining the facial image to be recognized according to the plurality of second similarities Facial feature recognition results.
- a face recognition device includes a first comparison module configured to perform a first similarity comparison between a face image to be recognized and a plurality of first pictures in a database, and obtain a plurality of first A similarity; a threshold module, configured to extract a portion of the first similarity among the multiple first similarities when the largest first similarity among the multiple first similarities is within the first similarity threshold; the second picture A generation module for determining a plurality of second pictures from a plurality of first pictures corresponding to a portion of the first similarity; a second comparison module for comparing a second face image to be recognized with a plurality of second pictures To obtain a plurality of second similarities; and a first result module configured to determine a facial feature recognition result of the facial image to be recognized according to the plurality of second similarities.
- an electronic device includes: one or more processors; a storage device for storing one or more programs; and when one or more programs are processed by one or more processors Execute so that one or more processors implement the method as above.
- a computer-readable medium is provided on which a computer program is stored, and the program implements the method as described above when executed by a processor.
- the face recognition method, device, electronic device, and computer-readable medium of the present disclosure it is possible to quickly and accurately perform facial feature recognition on a human face in the case of massive data, and output a recognition result.
- An embodiment of the present application further provides a computer program product.
- the computer program product includes a computing program stored on a non-volatile computer-readable storage medium.
- the computer program includes program instructions. When the program instructions are executed by the computer, the computer causes the computer to Perform the method described above.
- FIG. 1 is a flowchart of a facial recognition method according to a first embodiment of the present application
- FIG. 2 is a flowchart of a facial recognition method according to a second embodiment of the present application.
- FIG. 3 is a flowchart of a facial recognition method according to a third embodiment of the present application.
- FIG. 4 is a flowchart of a facial recognition method according to a fourth embodiment of the present application.
- FIG. 5 is a flowchart of a facial recognition method according to a fifth embodiment of the present application.
- FIG. 6 is a functional block diagram of a facial recognition device according to a sixth embodiment of the present application.
- FIG. 7 is a functional block diagram of a facial recognition device according to a seventh embodiment of the present application.
- FIG. 8 is a schematic structural diagram of an electronic device according to an eighth embodiment of the present application.
- FIG. 9 is a schematic structural diagram of a computer-readable storage medium according to a ninth embodiment of the present application.
- first, second, third, etc. may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another. Accordingly, the first component discussed below may be referred to as the second component without departing from the teachings of the concepts of the present disclosure. As used herein, the term "and / or” includes any combination of any one and one or more of the associated listed items.
- the area feature analysis algorithm widely used in face recognition technology uses deep learning technology to extract portrait feature points from videos and photos, and uses the principle of biostatistics to analyze and establish mathematical models, that is, faces Feature template.
- the best matching facial feature template is searched, and a person's Identity Information.
- the extracted facial feature data is searched and matched with the feature templates stored in the database, and an optimal similarity threshold is set. When the similarity exceeds the threshold, the matched result is output.
- this application proposes a face recognition method.
- the attribute recognition of age and gender can be added, which can significantly improve the accuracy of face recognition in a massive data search scenario.
- the facial recognition method proposed in this application can be set on a remote server (such as a cloud server) or in a detection device (facial recognition device), which is not limited in this application.
- FIG. 1 is a flowchart of a facial recognition method according to a first embodiment of the present application.
- the facial recognition method 10 includes at least steps S102 to S110.
- a first similarity comparison is performed between a face image to be identified and a plurality of first pictures in a database to obtain a plurality of first similarities.
- the first similarity comparison is a comparison of facial features.
- feature points of a facial image to be identified may be extracted; the feature points may be compared with feature points of a plurality of first pictures in a database; and the plurality of first image points may be obtained according to a comparison result. A similarity.
- face model training and face feature extraction are processes that convert a face image into a series of fixed-length values.
- This numerical string is called “Face Feature” and has the ability to characterize this face.
- the input of the facial feature extraction process is also "a face image” and “coordinates of the key points of the facial features”, and the output is a numerical string (feature) corresponding to the face.
- the model training module uses deep neural networks to train feature extraction models based on big face data. This model is used for subsequent face recognition comparisons.
- Face recognition comparison includes three processes: face image preprocessing, feature extraction, and feature comparison.
- Face preprocessing includes two processes: face detection and face registration.
- Face comparison is an algorithm that measures the similarity between two faces. The input of the face comparison algorithm is two face features (the face features are obtained by the previous face feature extraction algorithm), and the output is the similarity between the two features.
- a part of the first similarities among the plurality of first similarities is extracted.
- Within the first similarity threshold may be, for example, 55% -65%.
- all the first similarities among the plurality of first similarities that are within the first similarity threshold may be extracted.
- a plurality of second pictures are determined from a plurality of first pictures corresponding to a portion of the first similarity.
- determining the user IDs corresponding to a plurality of first pictures corresponding to the first similarity in the database merging the first pictures having the same user ID; and generating a plurality of first pictures according to the merged result.
- Two pictures and corresponding first similarities are weighted and combined.
- TOP1 when the highest similarity (TOP1) appears between the best similarity threshold (e.g., 55%) and the high similarity threshold (65%), the similarity appears in this interval in less certain When the threshold is at the edge, the probability of misidentification is high.
- the face recognition comparison yields TOP N above the optimal similarity threshold
- TOP N detects the corresponding user ID through the face feature vector. If there are multiple feature vectors for the same person, Then, the similarity of multiple face feature vectors of the same person is weighted and accumulated.
- the similarity between the feature vectors of the photo currently collected by a person and the three face photos is a, b, c (the best similarity is 55%, the high similarity threshold is 65%, 55% ⁇ a, b, c ⁇ 65%), then the similarity between the person ’s current photo and the photo in the face database may be, for example: a + n * b + m * c, where n and m are accumulation coefficients (0 ⁇ n, m ⁇ 1).
- the output result is re-ranked according to the user ID (each person's unique ID, which can correspond to multiple photos).
- a second similarity degree comparison is performed on the face image to be recognized and a plurality of second pictures to obtain a plurality of second similarity degrees.
- the second similarity comparison includes an age comparison and a gender comparison.
- age comparison may be performed on the face image to be identified with a plurality of second pictures to generate age similarity
- gender comparison may be performed on the face image to be identified with a plurality of second pictures to generate gender similarity.
- a facial feature recognition result of the facial image to be recognized is determined according to the plurality of second similarities. For example, based on the gender and age recognition of the current face photo, the TOP ranking is re-ranked based on the real gender and age of each person in the TOP.
- the true gender and age attributes of each face in the face database can be obtained from the ID number information.
- the age recognition feature is used to perform secondary recognition on the recognition result, and a way of providing the final recognition result is provided.
- facial features are quickly and accurately recognized on the face, and the recognition results are output.
- FIG. 2 is a flowchart of a facial recognition method according to a second embodiment of the present application.
- the facial feature recognition method 20 shown in FIG. 2 is a detailed description of “when the largest first similarity among a plurality of first similarities is greater than within a first similarity threshold”.
- a user identifier corresponding to the facial image to be identified in the database is determined according to the recognition result.
- the face image to be identified is added to the database according to the first number of pictures corresponding to the user identification in the database.
- the first image in the database is updated with the facial image to be identified according to the number of first pictures corresponding to the user identification in the database.
- the highest threshold of the first similarity may be, for example, 65%.
- the face image to be identified can be added to the database and the user ID
- the same neural network model is used to extract a face feature database, and the new first picture of the user is stored in the database to support one person with multiple faces.
- the first time of the photos in the first picture corresponding to the user in the database may be more than X years (for example, three years).
- a picture is deleted, and then the face image to be identified is added to the database under the user's ID, and the same neural network model is used to extract a face feature database, which is stored in the database as a newly added first picture of the user.
- the facial image to be recognized is not added to the database as the first picture.
- multiple photos corresponding to each user can cover various scenes at the time of face collection, including light and face size, by means of one user corresponding to multiple photos. Face angle, etc., can provide diverse first pictures for subsequent face recognition.
- the facial recognition method of the present disclosure in the database, a method of replacing multiple first photos in the database with multiple photos corresponding to each user and avoiding bad facial recognition caused by changes in user age can be avoided. influences.
- FIG. 3 is a flowchart of a facial recognition method according to a third embodiment of the present application.
- the process 30 shown in FIG. 3 is a detailed description of S102 in the process shown in FIG. 2 "Comparing a first similarity of a face image to be recognized with multiple first pictures in a database to obtain multiple first similarities" .
- the feature points are compared with the feature points of multiple first pictures in the database.
- Face recognition comparison includes three processes: face image preprocessing, feature extraction, and feature comparison.
- Face preprocessing includes two processes: face detection and face registration.
- Face comparison is an algorithm that measures the similarity between two faces. The input of the face comparison algorithm is the similarity between the two facial features and the output. The extracted facial feature data is searched and matched with the feature template stored in the database, and an optimal similarity threshold is set. When the similarity exceeds the threshold, the matched result is output.
- Fig. 4 is a flowchart illustrating a method for facial recognition according to a fourth embodiment of the present application.
- the process 40 shown in FIG. 4 is a detailed description of S208 in the process shown in FIG. 2 "Comparing a second similarity of a face image to be recognized with a plurality of second pictures to obtain a plurality of second similarities".
- age comparison is performed on a face image to be recognized and a plurality of second pictures to generate age similarity.
- the difficulty of face age recognition lies in the recognition of different ages of a single person and the recognition of different ages of multiple people. Face age recognition is often combined with face recognition to more accurately determine whether "is a person within a certain number of years" "The problem.
- the characteristics of age recognition include, but are not limited to, the position of an adult's eyes is generally 1 ⁇ 2 above and below the head. The distance from the corner of the outer eye to the corner of the mouth is equal to the distance from the tragus. Because of the loss of teeth, the facial features of the elderly are slightly shorter than 1/2 under the eyes. The position of the child's facial features is not yet full because of the chin, and the face above the eyes is slightly longer than 1/2.
- the process of face data training is: 1. Extracting facial feature points 2. Constructing proportional features, length features and other features 3. Performing cluster classification training based on samples.
- a first age of the facial image to be identified may be determined; a plurality of second ages of the plurality of second pictures may be determined; and a first distance and a plurality of second ages may be determined by vector distance calculation.
- Age similarity between ages may be determined.
- the vector distance calculation includes: Euclidean distance calculation, variance calculation, and cosine distance calculation.
- a gender comparison is performed on the facial image to be recognized and a plurality of second pictures to generate a gender similarity.
- Face gender classification is a second-class problem. Two key issues that need to be solved for face gender classification are face feature extraction and the choice of a classifier.
- the gender-specific facial features include, but are not limited to, male skulls: sharp edges and corners, straight lines, and eyebrow arches that have more prominent forehead slopes than women. The orbits are smaller than women, and the nasal and mandible bones are more developed.
- Female skull The edges and corners are soft and round. The eye sockets are larger than those of men, and the nose and jaw atrophy, and the entire head shape appears slightly smaller.
- the gender of the facial image to be recognized may be determined through feature recognition, and the gender is determined based on the identity information of the user corresponding to the second picture, and finally a gender comparison is performed.
- the second similarity is determined by the age similarity and the gender similarity.
- the method of facial feature vector extraction in face recognition is to train a deep facial neural network model through a large amount of face sample data.
- This model is used to extract facial image feature vectors in the face base database, and it is also used in the current Feature extraction of face images collected and detected; face gender and age attribute recognition are also trained by a large number of face images of different genders, ages, and gender classification and age classification.
- face gender and age attribute recognition are also trained by a large number of face images of different genders, ages, and gender classification and age classification.
- the neural network used is different from the face recognition feature extraction.
- supervised learning of face features from two other dimensions through face gender and age recognition can assist in the selection and rearrangement of face recognition results, and has high confidence in the face recognition field.
- Fig. 5 is a flowchart illustrating a facial recognition method according to a fifth embodiment of the present application. A detailed description of the entire process of facial feature recognition in the facial recognition method 50 shown in FIG. 5 is shown in FIG. 5.
- S502 an image to be recognized is compared with a first image to obtain a first similarity.
- the first image corresponding to the highest first similarity is used as the recognition result.
- the system By presetting a high-precision similarity threshold, for example, if the similarity threshold is 65%, during the comparison search process, the similarity exceeds the high threshold (65%), the current photo is considered to be the person in the face database.
- the system automatically adds the currently collected faces to the face base database, and uses the same neural network model to extract the face feature database.
- new faces are stored in the library to support multiple faces of one person, if the number of face photos of that person has reached 10, the most recent face photos that have been older than X years (such as three years) are updated. Otherwise, do not add faces or update face photos; 10 photos per person, as much as possible to cover the scene collected by the current face, including light, face size, face angle, etc.
- the update of face photos is to avoid Age changes affect recognition.
- the recognition TOP1 is directly considered to be the target face.
- the highest similarity (TOP1) appears between the best similarity threshold (e.g. 55%) and the high similarity threshold (65%) (when the similarity appears on the edge of a less certain threshold, the probability of misidentification is very high Large)
- TOP N above 1, the corresponding user ID is detected by TOF N through the face feature vector. If there are multiple feature vectors, For the same person, the similarity between multiple face feature vectors of the same person is weighted and accumulated.
- the similarity between the feature vectors of a person's current photo and three face photos is a, b, c (most If the best similarity is 55%, the high similarity threshold is 65%, and 55% ⁇ a, b, c ⁇ 65%), then the similarity between the current photo of the person and the photos in the face database is (including but not limited to) The following calculation method): a + n * b + m * c, where n and m are accumulation coefficients (0 ⁇ n, m ⁇ 1), the similarity of the same person is performed And, re-output according to the user ID (unique per ID, may correspond to multiple pictures) sorting a TOP P, and outputs, this method may improve the recognition rate largely;
- the feature vectors extracted from the currently collected photos are compared with all face features in the face database.
- the highest similarity does not exceed the high similarity threshold, appears between the best similarity and high similarity, and multiple photos of the same person.
- the similarity is merged.
- the gender and age of the currently collected face image are identified, and the actual real gender of the person corresponding to the face of TOP above all the best similarity thresholds.
- New vector distance calculation with age (gender and age attributes with a confidence coefficient, vector distance calculation, including but not limited to using Euclidean distance, variance, cosine distance, etc.) to get a new TOP
- the picture corresponding to the maximum value in TOP M can be used as the recognition result, and for example, all pictures in TOP M can be used as the recognition result, and these pictures can be displayed in the face recognition terminal.
- artificial means which is not limited in this application.
- the face recognition method disclosed in the present disclosure is directed to a scene where a current face image collected by a face collection terminal is (1: N) searched and compared in a base library (more than one million) with a large number of face features, and a method is provided.
- a current face image collected by a face collection terminal is (1: N) searched and compared in a base library (more than one million) with a large number of face features, and a method is provided.
- a method is provided.
- the face recognition method of the present disclosure is a method for significantly improving the accuracy rate of face recognition in a face recognition large database search scenario.
- the program may be stored in a computer-readable storage medium, and the storage medium may be a read-only memory, a magnetic disk, or an optical disk.
- Fig. 6 is a functional block diagram of a facial recognition device according to a sixth embodiment of the present application.
- the facial recognition device 60 includes a first comparison module 602, a threshold module 604, a second picture generation module 606, a second comparison module 608, and a first result module 610.
- the first comparison module 602 is configured to perform a first similarity comparison between a face image to be identified and a plurality of first pictures in a database to obtain a plurality of first similarities; wherein the first similarity comparison is a facial feature comparison.
- the threshold module 604 is configured to extract a portion of the first similarity among the plurality of first similarities when the largest first similarity among the plurality of first similarities is within the first similarity threshold; the first similarity threshold may be within For example, 55% -65%.
- the second picture generating module 606 is configured to determine a plurality of second pictures by using a plurality of first pictures corresponding to a part of the first similarity; for example, combining the first pictures having the same user ID; The second picture and corresponding first similarities.
- the second comparison module 608 is configured to perform a second similarity comparison between the face image to be identified and a plurality of second pictures to obtain a plurality of second similarities; wherein the second similarity comparison includes an age comparison and a gender comparison.
- the first result module 610 is configured to determine a facial feature recognition result of the facial image to be recognized according to the multiple second similarities.
- the age recognition feature is used to perform secondary recognition on the recognition result, and a way of providing a final recognition result can be provided in a large amount.
- facial features are quickly and accurately recognized on the face, and the recognition results are output.
- Fig. 7 is a block diagram of a facial recognition device according to a seventh embodiment of the present application.
- the facial recognition device 70 further includes a second result module 702, a user identification module 704, and a picture update module 706.
- the second result module 702 is configured to use the first picture corresponding to the largest first similarity as the recognition of the facial image to be recognized when the largest first similarity among the plurality of first similarities is greater than the first similarity threshold. result.
- the user identification module 704 is configured to determine a user identification corresponding to a facial image to be identified in a database according to the recognition result;
- the picture update module 706 is configured to determine a subsequent processing mode according to the first picture quantity corresponding to the user identifier in the database.
- FIG. 8 is a schematic structural diagram of an electronic device according to an eighth embodiment of the present application.
- FIG. 8 An electronic device 200 according to this embodiment of the present disclosure is described below with reference to FIG. 8.
- the electronic device 200 shown in FIG. 8 is merely an example, and should not impose any limitation on the functions and scope of use of the embodiments of the present disclosure.
- the electronic device 200 is expressed in the form of a general-purpose computing device.
- the components of the electronic device 200 may include, but are not limited to, at least one processing unit 210, at least one storage unit 220, a bus 230 connecting different system components (including the storage unit 220 and the processing unit 210), a display unit 240, and the like.
- the storage unit stores program code, and the program code can be executed by the processing unit 210, so that the processing unit 210 executes various exemplary embodiments according to the present disclosure described in the above-mentioned electronic prescription circulation processing method section of this specification. Steps of the implementation.
- the processing unit 210 may perform the steps shown in FIG. 1, FIG. 2, FIG. 3, FIG. 4, and FIG. 5.
- the storage unit 220 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 2201 and / or a cache storage unit 2202, and may further include a read-only storage unit (ROM) 2203.
- RAM random access storage unit
- ROM read-only storage unit
- the storage unit 220 may further include a program / utility tool 2204 having a group (at least one) of program modules 2205.
- program modules 2205 include, but are not limited to, an operating system, one or more application programs, other program modules, and programs. Data, each or some combination of these examples may include an implementation of the network environment.
- the bus 230 may be one or more of several types of bus structures, including a memory unit bus or a memory unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local area using any bus structure in a variety of bus structures bus.
- the electronic device 200 may also communicate with one or more external devices 300 (such as a keyboard, pointing device, Bluetooth device, etc.), and may also communicate with one or more devices that enable a user to interact with the electronic device 200, and / or with Any device (eg, router, modem, etc.) that enables the electronic device 200 to communicate with one or more other computing devices. This communication may be performed through an input / output (I / O) interface 250.
- the electronic device 200 may also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 260.
- the network adapter 260 may communicate with other modules of the electronic device 200 through the bus 230.
- the technical solution according to the embodiment of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a U disk, a mobile hard disk, etc.) or on a network Including several instructions to cause a computing device (which may be a personal computer, a server, or a network device, etc.) to execute the above method according to an embodiment of the present disclosure.
- a non-volatile storage medium which may be a CD-ROM, a U disk, a mobile hard disk, etc.
- a computing device which may be a personal computer, a server, or a network device, etc.
- An embodiment of the present application further provides a computer program product 400.
- the computer program product 400 includes a computing program stored on a non-volatile computer-readable storage medium.
- the computer program includes program instructions. When the program instructions are executed by a computer, The computer is caused to execute the above-mentioned robot control method based on the optical fiber communication network.
- Fig. 9 is a schematic structural diagram of a computer-readable storage medium according to a ninth embodiment of the present application.
- a program product 400 for implementing the above method according to an embodiment of the present disclosure is described, which may adopt a portable compact disc read-only memory (CD-ROM) and include program code, and may be implemented in a terminal device. For example running on a personal computer.
- the program product of the present disclosure is not limited thereto.
- the readable storage medium may be any tangible medium containing or storing a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.
- the program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, which include object-oriented programming languages such as Java, C ++, and the like, as well as conventional procedural Programming language—such as "C" or a similar programming language.
- the program code may be executed entirely on the user computing device, partly on the user device, as an independent software package, partly on the user computing device, partly on the remote computing device, or entirely on the remote computing device or server On.
- the remote computing device may be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g. (Commercially connected via the Internet).
- LAN local area network
- WAN wide area network
- an external computing device e.g. (Commercially connected via the Internet).
- the computer-readable medium carries one or more programs, and when the one or more programs are executed by one of the devices, the computer-readable medium implements the following functions: a face image to be identified and a plurality of first
- the pictures are compared for the first similarity to obtain multiple first similarities.
- a part of the multiple first similarities is extracted.
- a similarity determining a plurality of second pictures by using a plurality of first pictures corresponding to a portion of the first similarity; comparing a face image to be recognized with a plurality of second pictures to obtain a plurality of second similarities ; And determining a facial feature recognition result of the facial image to be recognized according to the plurality of second similarities.
- modules in the above embodiments may be distributed in the device according to the description of the embodiment, or corresponding changes may be made in one or more devices that are different from this embodiment.
- the modules in the above embodiments may be combined into one module, or further divided into multiple sub-modules.
- the technical solution according to the embodiment of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a U disk, a mobile hard disk, etc.) or on a network Including instructions to cause a computing device (which may be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to an embodiment of the present disclosure.
- a non-volatile storage medium which may be a CD-ROM, a U disk, a mobile hard disk, etc.
- a computing device which may be a personal computer, a server, a mobile terminal, or a network device, etc.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
本公开涉及一种面部识别方法、装置、电子设备及计算机可读介质。该方法包括:将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度;在多个第一相似度中的最大第一相似度位于第一相似度门限内时,提取多个第一相似度中的部分第一相似度;通过部分第一相似度对应的多个第一图片确定多个第二图片;将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度;以及根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。本公开涉及的面部识别方法、装置、电子设备及计算机可读介质,能够在海量数据的情况下快速准确的对人脸进行面部特征识别,输出识别结果。
Description
本申请涉及计算机信息处理技术领域,特别涉及一种面部识别方法、装置、电子设备及计算机可读介质。
人脸识别技术是近年来出现的一种基于人的脸部特征信息进行身份识别的生物特征识别技术,也是目前在基于深度学习的人工智能技术应用最广泛的。与其他生物识别技术相比较,人脸识别具有友好、简便、准确、经济及可扩展性良好等众多优势,可广泛应用于安全验证、监控、出入口控制等多个方面,目前人脸识别技术已经应用到门禁考勤,访客管理,巡更等场所。
在研究现有人脸识别的场景中,人脸识别至少存在如下问题:在自然场景下进行人脸识别时,在预设人脸特征库包含的人脸特征数量较大时(如百万人脸库),受到采集人脸图的光照,分辨率,模糊,角度等因素影响,由于人脸特征比对的样本空间范围较大,出现人脸特征相似度很高可能性就增大,因此,在实际的人脸识别过程中会出现识别精确率下降的问题
因此,需要一种新的面部识别方法、装置、电子设备及计算机可读介质。
在所述背景技术部分公开的上述信息仅用于加强对本公开的背景的理解,因此它可以包括不构成对本领域普通技术人员已知的现有技术的信息。
发明内容
本申请实施方式的目的在于提供一种面部识别方法、装置、电子设备及计算机可读介质,能够在海量数据的情况下快速准确的对人脸进行面部特征识别,输出识别结果。
本公开的其他特性和优点将通过下面的详细描述变得显然,或部分地通过本公开的实践而习得。
根据本公开的一方面,提出一种面部识别方法,该方法包括:将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度; 在多个第一相似度中的最大第一相似度位于第一相似度门限内时,提取多个第一相似度中的部分第一相似度;通过部分第一相似度对应的多个第一图片确定多个第二图片;将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度;以及根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。
根据本公开的一方面,提出一种面部识别装置,该装置包括:第一比较模块,用于将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度;门限模块,用于在多个第一相似度中的最大第一相似度位于第一相似度门限内时,提取多个第一相似度中的部分第一相似度;第二图片生成模块,用于通过部分第一相似度对应的多个第一图片确定多个第二图片;第二比较模块,用于将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度;以及第一结果模块,用于根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。
根据本公开的一方面,提出一种电子设备,该电子设备包括:一个或多个处理器;存储装置,用于存储一个或多个程序;当一个或多个程序被一个或多个处理器执行,使得一个或多个处理器实现如上文的方法。
根据本公开的一方面,提出一种计算机可读介质,其上存储有计算机程序,该程序被处理器执行时实现如上文中的方法。
根据本公开的面部识别方法、装置、电子设备及计算机可读介质,能够在海量数据的情况下快速准确的对人脸进行面部特征识别,输出识别结果。
本申请的实施方式还提供了一种计算机程序产品,计算机程序产品包括存储在非易失性计算机可读存储介质上的计算程序,计算机程序包括程序指令,当程序指令被计算机执行时,使计算机执行上所述的方法。
应当理解的是,以上的一般描述和后文的细节描述仅是示例性的,并不能限制本公开。
一个或多个实施方式通过与之对应的附图中的图片进行示例性说明,这些示例性说明并不构成对实施方式的限定,附图中具有相同参考数字标号的元件表示为类似的元件,除非有特别申明,附图中的图不构成比例限制。
图1是根据本申请第一实施方式的一种面部识别方法的流程图;
图2是根据本申请第二实施方式的一种面部识别方法的流程图;
图3是根据本申请第三实施方式的一种面部识别方法的流程图;
图4是根据本申请第四实施方式的一种面部识别方法的流程图;
图5是根据本申请第五实施方式的一种面部识别方法的流程图;
图6是根据本申请第六实施方式的一种面部识别装置的功能框图;
图7是根据本申请第七实施方式的一种面部识别装置的功能框图;
图8是根据本申请第八实施方式的一种电子设备的结构示意图;
图9是根据本申请第九实施方式的一种计算机可读存储介质的结构示意图。
现在将参考附图更全面地描述示例实施例。然而,示例实施例能够以多种形式实施,且不应被理解为限于在此阐述的实施例;相反,提供这些实施例使得本公开将全面和完整,并将示例实施例的构思全面地传达给本领域的技术人员。在图中相同的附图标记表示相同或类似的部分,因而将省略对它们的重复描述。
此外,所描述的特征、结构或特性可以以任何合适的方式结合在一个或更多实施例中。在下面的描述中,提供许多具体细节从而给出对本公开的实施例的充分理解。然而,本领域技术人员将意识到,可以实践本公开的技术方案而没有特定细节中的一个或更多,或者可以采用其它的方法、组元、装置、步骤等。在其它情况下,不详细示出或描述公知方法、装置、实现或者操作以避免模糊本公开的各方面。
附图中所示的方框图仅仅是功能实体,不一定必须与物理上独立的实体相对应。即,可以采用软件形式来实现这些功能实体,或在一个或多个硬件模块或集成电路中实现这些功能实体,或在不同网络和/或处理器装置和/或微控制器装置中实现这些功能实体。
附图中所示的流程图仅是示例性说明,不是必须包括所有的内容和操作/步骤,也不是必须按所描述的顺序执行。例如,有的操作/步骤还可以分解,而有的操作/步骤可以合并或部分合并,因此实际执行的顺序有可能根据实际情况改变。
应理解,虽然本文中可能使用术语第一、第二、第三等来描述各种组件,但这些组件不应受这些术语限制。这些术语乃用以区分一组件与另一组件。因此,下文论述的第一组件可称为第二组件而不偏离本公开概念的教示。如本文中所使用,术语“及/或”包括相关联的列出项目中的任一个及一或多者的所有组合。
本领域技术人员可以理解,附图只是示例实施例的示意图,附图中的模块或流程并不一定是实施本公开所必须的,因此不能用于限制本公开的保护范围。
本申请的发明人发现,人脸识别技术中被广泛采用的区域特征分析算法,采用深度学习技术从视频和照片中提取人像特征点,利用生物统计学的原理进行分析建立数学模型,即人脸特征模板。利用已建成的人脸特征模板与被测者的人的面像进行特征分析,根据分析的结果来给出一个相似度值,最终搜索到最佳匹配人脸特征模板,并因此确定一个人的身份信息。
具体可例如,将提取的人脸特征数据与数据库中储存的特征模板进行搜索匹配,设定一个最佳的相似度阈值,当相似度超过该阈值,则输出匹配后的结果。
目前,最佳相似度阈值的决定需要考虑到输出结果的正确率和数量。假设1000张样本图片里,共600张正样本。相似度为0.9的图片一共100张,其中正样本为99张。虽然0.9阈值的正确率很高,为99/100;但是0.9阈值正确输出的数量确很少,只有99/600,在这种情况下在面部特征识别时很容易发生漏识的情况。
其中,精确率(precision):识别为正确的样本数/识别出来的样本数=99/100;召回率(recall):识别为正确的样本数/所有样本中正确的数=99/600。
有鉴于此,本申请提出了一种面部识别方法,在面部特征识别的技术上,附加年龄和性别的属性识别,能够在海量数据搜索场景下的显著提升人脸识别正确率。
本申请提出的面部识别方法,可设置与远端服务器上(例如为云端服务器)或者检测设备(面部识别装置)中,本申请不以此为限。
下面将通过具体的实施例对本申请的内容进行详细说明:
图1是根据本申请的第一实施方式的一种面部识别方法的流程图。面部识别方法10至少包括步骤S102至S110。
如图1所示,在S102中,将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度。其中,第一相似度比较为面部特征比较。
在一个实施例中,可例如提取待识别的面部图像的特征点;将所述特征点与数据库中多个第一图片的特征点进行特征比对;以及根据比对结果获取所述多个第一相似度。
其中,人脸模型训练和人脸特征提取是将一张人脸图像转化为一串固定长度的数值的过程。这个数值串被称为“人脸特征(Face Feature)”,具有表征这个人脸特点的能力。人脸特征提取过程的输入也是“一张人脸图”和“人脸五官关键点坐标”,输出是人脸相应的一个数值串(特征)。模型训练模块基于人脸大数据采用深度神经网络训练特征提取模型,该模型用于后续的人脸识别比对。
人脸识别比对包括人脸图像预处理、特征提取、特征比对三个过程。人脸预处理又包括人脸检测和人脸配准两个过程。人脸比对是衡量两个人脸之间相似度的算法。人脸比对算法的输入是两个人脸特征(人脸特征由前面的人脸特征提取算法获得),输出是两个特征之间的相似度。
在S104中,在多个第一相似度中的最大第一相似度位于第一相似度门限内时,提取多个第一相似度中的部分第一相似度。第一相似度门限内可例如为55%-65%。
在一个实施例中,可例如提取多个第一相似度中的所有在所述第一相似度门限内的第一相似度。
在S106中,通过部分第一相似度对应的多个第一图片确定多个第二图片。
在一个实施例中,可例如:确定部分第一相似度对应的多个第一图片在数据库中对应的用户标识;将具有相同用户标识的第一图片进行合并;以及根据合并结果生成多个第二图片以及对应的多个第一相似度。具体还可例如,将具有相同用户标识的多个第一图片对应的多个第一相似度进行权重累加合并。
在一个实施例中,当最高相似度(TOP1)出现在最佳相似度阈值(如55%)和高相似度阈值(65%)之间时,在这个区间中相似度出现在不太确定的阈值边缘时,即识别错的可能性很大。人脸识别比对得出TOP N时(超过最佳相似度阈值),当N>1时,对TOP N分别通过人脸特征向量查出对应用户ID,如果存在多个特征向量为同一个人,则对同一人的多张人脸特征向量的相似度进行权 重累加。
在一个实施例中,某个人当前采集的照片和三张人脸照片的特征向量的相似度分别为a,b,c(最佳相似度为55%,高相似度阈值为65%,55%<a,b,c<65%),则该人当前照片与人脸库中的照片的相似度可例如为:a+n*b+m*c,其中n,m是累加系数(0<n,m<1)。
在属于同一用户ID下的第一图片的第一相似度进行合并累积之后,根据结果重新对输出结果按照用户ID(每人唯一ID,可对应多张照片)进行一个TOP P的排序。
在S108中,将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度。其中,第二相似度比较包括年龄比较与性别比较。
在一个实施例中,可例如,将待识别的面部图像与多个第二图片进行年龄比较,生成年龄相似度;将待识别的面部图像与多个第二图片进行性别比较,生成性别相似度;通过年龄相似度与性别相似度确定所述第二相似度。
在S110中,根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。可例如,基于对当前人脸照片进行性别和年龄的识别,基于TOP P中每个人真实性别和年龄再重新进行TOP M排名。
在一个实施例中,人脸库中每个人脸的真实性别和年龄属性,可来自身份证号码信息。
根据本公开的面部识别方法,在通过面部特征对当前图片和数据库中的图片进行面部识别之后,再利用年龄属性特征对识别结果进行二次识别,并给出最终识别结果的方式,能够在海量数据的情况下快速准确的对人脸进行面部特征识别,输出识别结果。
应清楚地理解,本公开描述了如何形成和使用特定示例,但本公开的原理不限于这些示例的任何细节。相反,基于本公开公开的内容的教导,这些原理能够应用于许多其它实施例。
图2是根据本申请第二实施方式的一种面部识别方法的流程图。图2所示的面部特征识别方法20是“在多个第一相似度中的最大第一相似度大于第一相似度门限内时”的详细描述。
如图2所示,在S202中,在多个第一相似度中的最大第一相似度大于第一相似度门限时,将最大的第一相似度对应的第一图片作为所述待识别面部图像 的识别结果。
在S204中,根据所述识别结果确定待识别面部图像在数据库中对应的用户标识。
在S206中,根据所述用户标识在所述数据库中对应的第一图片数量将所述待识别的面部图像添加至所述数据库。
在S208中,根据所述用户标识在所述数据库中对应的第一图片数量用所述待识别的面部图像更新所述数据库内的第一图像。
第一相似度的门限的最高值可例如65%,在待识别图像与数据库中的第一图像进行比对搜索过程中,如果出现相似度超过了高阈值(65%)的第一图像,则认为当前照片就是数据库中该第一图像对应的用户。
此时可进行后续判断,在一个实施例中,当该人在人脸库中人脸照片数不超过M个(比如10个)时,可将待识别的面部图像添加至数据库该用户的ID下,并使用相同的神经网络模型提取人脸特征库,作为新增的该用户的第一图片存入数据库中,以支持一人多脸。
在一个实施例中,如果该用户的第一照片数已达到10个,则可例如将数据库中,该用户对应的第一图片中的照片中时间超过X年(如三年)以上的进行第一图片删除,然后将待识别的面部图像添加至数据库该用户的ID下,并使用相同的神经网络模型提取人脸特征库,作为新增的该用户的第一图片存入数据库中。
在一个实施例中,如果该用户的第一照片数已达到10个,但是数据库中该用户的照片均未超过时间限制,则不将待识别的面部图像作为第一图片添加到数据库中。
根据本公开的面部识别方法,在数据库中,每个用户对应的多张照片,通过一个用户对应多张照片的方式,能够覆盖人脸采集时候的各种现场场景,包括光线,人脸大小,人脸角度等,能够为后续的人脸识别提供多样性的第一照片。
根据本公开的面部识别方法,在数据库中,每个用户对应的多张照片,并根据当前照片对数据库中时间较长的第一照片进行替换的方式,能够避免用户年龄变化对面部识别造成不良影响。
图3是根据本申请第三实施方式的一种面部识别方法的流程图。图3所示的流程30是对图2所示的流程中S102“将待识别的面部图像与数据库中多个第 一图片进行第一相似度比较,获取多个第一相似度”的详细描述。
如图3所示,在S302中,提取待识别的面部图像的特征点。
在S304中,将所述特征点与数据库中多个第一图片的特征点进行特征比对。
在S306中,根据比对结果获取所述多个第一相似度。
人脸识别比对包括人脸图像预处理、特征提取、特征比对三个过程。人脸预处理又包括人脸检测和人脸配准两个过程。人脸比对是衡量两个人脸之间相似度的算法。人脸比对算法的输入是两个人脸特征输出是两个特征之间的相似度。将提取的人脸特征数据与数据库中储存的特征模板进行搜索匹配,设定一个最佳的相似度阈值,当相似度超过该阈值,则输出匹配后的结果。
图4是根据本申请第四实施方式示出的一种面部识别方法的流程图。图4所示的流程40是对图2所示的流程中S208“将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度”的详细描述。
如图4所示,在S402中,将待识别的面部图像与多个第二图片进行年龄比较,生成年龄相似度。人脸年龄识别的难度在于单人的不同年龄段识别和多人的不同年龄段识别,人脸年龄识别常和人脸识别进行组合识别,能更正确的判断在一定年限内“是否是一个人”的问题。年龄识别的特征包括但不限于:成年人眼睛位置一般在头部上下1/2处。外眼角到嘴角与它到耳屏的距离相等。老人五官的位置由于牙的脱落,眼以下面部稍短于1/2。小孩五官的位置由于下颏尚未长全,眼以上面部稍长于1/2。
对通过人脸识别分类性别和年龄(划分为多个年龄段,如5岁一段)的过程,基于大量人脸训练的模型提取人脸特征,将人脸特征输入预先训练好聚类分类器,输入性别和年龄的分类结果。其中,人脸数据训练的过程为:1.提取人脸特征点2.构造比例特征,长度特征等特征3.根据样本进行聚类分类训练。
在一个实施例中,可例如确定所述待识别的面部图像的第一年龄;确定所述多个第二图片的多个第二年龄;以及通过向量距离计算确定第一年龄与多个第二年龄之间的年龄相似度。
在一个实施例中,向量距离计算包括:欧式距离计算、方差计算、以及余弦距离计算。
在S404中,将待识别的面部图像与多个第二图片进行性别比较,生成性别相似度。人脸性别分类是一个二类问题,人脸性别分类问题需要解决的两个关 键问题是人脸特征提取和分类器的选择。性别分类的人脸特征包括但不限于,男性头骨:棱角分明、线条刚直、眉弓较女性突出额头坡度较大。眼眶较女性小、鼻骨和下颌骨较发达。女性头骨:棱角柔和、浑圆,额结节较男性突出。眼眶较男性大、鼻骨和下颌萎缩,整个头形显得略小。
在一个实施例中,可例如通过特征识别确定所述待识别的面部图像的性别,同时通过第二图片对应的用户的身份信息确定性别,最终进行性别比较。
在S406中,通过年龄相似度与性别相似度确定所述第二相似度。
人脸识别中的人脸特性向量提取的方法,是通过大量人脸样本数据训练出人脸深度神经网络模型,该模型用于人脸底库中人脸图像特征向量提取,同样也用于当前采集和检测出来的人脸图像的特征提取;人脸性别和年龄属性识别,也是通过大量标记过不同性别,不同年龄人脸图像数据进行训练,并通过性别二分类法和年龄段分类的方法,来对当前采集和检测到的人脸进行性别和年龄的识别,采用的神经网络与人脸识别特征提取的不同。
根据本公开的面部识别方法,通过人脸性别和年龄识别从另外两个维度对人脸特征进行监督学习,能够辅助面部识别结果的筛选和重排,在人脸识别领域具有高置信度。
图5是根据本申请第五实施方式示出的一种面部识别方法的流程图。图5所示的面部识别方法50中面部特征识别全过程的详细描述如图5所示,在S502中,将待识别图像与第一图像进行比对,得到第一相似度。
在S504中,第一相似度的最高值是否大于55%。
在S506中,未识别出结果。
在S508中,第一相似度的最高值是否大于65%。
在S510中,将最高的第一相似度对应的第一图像作为识别结果。
在S512中,将具有相同用户标识的第一图片进行合并。
在S514中,根据合并后的结果将第一相似度结果进行重排。
在S516中,进行性别和年龄比较得到第二相似度,并进行排序。
在S518中,将最高的第二相似度对应的第二图像作为识别结果。
在S520中,输出结果。
通过预设高精确率相似度阈值,如相似度阈值为65%,在比对搜索过程中,出现相似度超过了高阈值(65%),则认为当前照片就是人脸库中该人,当该人 在人脸库中人脸照片数不超过M个(比如10个)时,同时系统自动增加当前采集人脸到人脸底库中,并使用相同的神经网络模型提取人脸特征库,作为新增人脸存入库中,以支持一人多脸,如果该人的人脸照片数已达到10个,则最已有人脸照片中时间超过X年(如三年)以上的进行更新,否则不进行添加人脸或更新人脸照片;每人10张照片,尽可能的覆盖当前人脸采集的场景,包括光线,人脸大小,人脸角度等,对人脸照片进行更新是为了避免年龄变化对识别造成影响。
针对当前采集人脸图像在和人脸库中人脸特征比对时,当出现最高相似度(TOP1)大于高相似度阈值(65%)时,则直接认为该识别TOP1就是目标人脸,当最高相似度(TOP1)出现在最佳相似度阈值(如55%)和高相似度阈值(65%)之间时(相似度出现在不太确定的阈值边缘时,即识别错的可能性很大),人脸识别比对得出TOP N时(超过最佳相似度阈值),当N>1时,对TOP N分别通过人脸特征向量查出对应用户ID,如果存在多个特征向量为同一个人,则对同一人的多张人脸特征向量的相似度进行权重累加,如,某个人当前采集的照片和三张人脸照片的特征向量的相似度分别为a,b,c(最佳相似度为55%,高相似度阈值为65%,55%<a,b,c<65%),则该人当前照片与人脸库中的照片的相似度则为(包括但不限于以下计算方法):a+n*b+m*c,其中n,m是累加系数(0<n,m<1),同一人相似度进行合并,重新对输出结果按照用户ID(每人唯一ID,可对应多张照片)进行一个TOP P的排序,并输出,此方法可能在很大程度上提升识别率;
当前采集照片提取特征向量与人脸库中所有人脸特征进行比对,最高相似度并未超出高相似度阈值,出现在最佳相似度和高相似度之间,且对同一人多张照片相似度进行了合并,基于输出基于用户ID的TOP P,则对当前采集人脸图像进行性别和年龄的识别,与高于所有最佳相似度阈值的TOP P的人脸对应人的实际真实性别和年龄进行新的向量距离计算(性别和年龄这两个属性带有一个置信度的系数,向量距离计算,包括但不限于采用欧式距离,方差,余弦距离等),得出一个新的TOP M,根据业务需要来使用TOP M,可例如将TOP M中的最大值对应的图片作为识别结果,还可例如,将TOP M中所有的图片均作为识别结果,将这些图片展示在面部识别终端中,通过人工的方式辅助进行面部识别,本申请不以此为限。
本公开的面部识别方法,针对人脸采集终端采集当前人脸图像在一个人脸特征数量较大的底库(百万以上)进行(1:N)搜索比对的场景,给出了一种通过增加和更新人脸底库的方法,增加人脸识别比对结果TOP N中出现与该人最为相似的多张人脸为同一人,并基于人(唯一UserID)进行重新TOP N的重新排名TOP P,以此确保最高相似度识别的准确率,同时基于当前人脸照片进行性别和年龄的识别(识别的特征提取方法与人脸比对的特征方法和模型不同),因此,基于每个用户(唯一UserID)可获得的真实性别和年龄属性,对TOP P再进行筛选排序形成TOP M,作为最终识别结果。通过此方案,可大幅提升在大量人脸库(百万级或千万级)的识别正确率。上述人脸底库,当前人脸的特征提取,比对获得TOP N,以及基于同一人脸的合并累加重排TOP P和基于识别人脸性别和年龄的重排TOP M均可在云端服务器侧完成。
本公开的面部识别方法,是一种在人脸识别大库搜索场景下的显著提升人脸识别正确率的方法。
本领域技术人员可以理解实现上述实施例的全部或部分步骤被实现为由CPU执行的计算机程序。在该计算机程序被CPU执行时,执行本公开提供的上述方法所限定的上述功能。所述的程序可以存储于一种计算机可读存储介质中,该存储介质可以是只读存储器,磁盘或光盘等。
此外,需要注意的是,上述附图仅是根据本公开示例性实施例的方法所包括的处理的示意性说明,而不是限制目的。易于理解,上述附图所示的处理并不表明或限制这些处理的时间顺序。另外,也易于理解,这些处理可以是例如在多个模块中同步或异步执行的。
下述为本公开装置实施例,可以用于执行本公开方法实施例。对于本公开装置实施例中未披露的细节,请参照本公开方法实施例。
图6是根据本申请第六实施方式示出的一种面部识别装置的功能框图。面部识别装置60包括:第一比较模块602,门限模块604,第二图片生成模块606,第二比较模块608,以及第一结果模块610。
第一比较模块602用于将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度;其中,第一相似度比较为面部特征比较。
门限模块604用于在多个第一相似度中的最大第一相似度位于第一相似度 门限内时,提取多个第一相似度中的部分第一相似度;第一相似度门限内可例如为55%-65%。
第二图片生成模块606用于通过部分第一相似度对应的多个第一图片确定多个第二图片;可例如,将具有相同用户标识的第一图片进行合并;以及根据合并结果生成多个第二图片以及对应的多个第一相似度。
第二比较模块608用于将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度;其中,第二相似度比较包括年龄比较与性别比较。
第一结果模块610用于根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。
根据本公开的面部识别装置,在通过面部特征对当前图片和数据库中的图片进行面部识别之后,再利用年龄属性特征对识别结果进行二次识别,并给出最终识别结果的方式,能够在海量数据的情况下快速准确的对人脸进行面部特征识别,输出识别结果。
图7是根据本申请第七实施方式示出的一种面部识别装置的框图。面部识别装置70在面部识别装置60的基础上还包括:第二结果模块702,用户标识模块704,图片更新模块706。
第二结果模块702用于在多个第一相似度中的最大第一相似度大于第一相似度门限时,将最大的第一相似度对应的第一图片作为所述待识别面部图像的识别结果。
用户标识模块704用于根据所述识别结果确定待识别面部图像在数据库中对应的用户标识;
图片更新模块706用于根据所述用户标识在所述数据库中对应的第一图片数量确定后续处理方式。
图8是根据本申请第八实施方式示出的一种电子设备的结构示意图。
下面参照图8来描述根据本公开的这种实施方式的电子设备200。图8显示的电子设备200仅仅是一个示例,不应对本公开实施例的功能和使用范围带来任何限制。
如图8所示,电子设备200以通用计算设备的形式表现。电子设备200的组件可以包括但不限于:至少一个处理单元210、至少一个存储单元220、连接 不同系统组件(包括存储单元220和处理单元210)的总线230、显示单元240等。
其中,所述存储单元存储有程序代码,所述程序代码可以被所述处理单元210执行,使得所述处理单元210执行本说明书上述电子处方流转处理方法部分中描述的根据本公开各种示例性实施方式的步骤。例如,所述处理单元210可以执行如图1,图2,图3,图4,以及图5中所示的步骤。
所述存储单元220可以包括易失性存储单元形式的可读介质,例如随机存取存储单元(RAM)2201和/或高速缓存存储单元2202,还可以进一步包括只读存储单元(ROM)2203。
所述存储单元220还可以包括具有一组(至少一个)程序模块2205的程序/实用工具2204,这样的程序模块2205包括但不限于:操作系统、一个或者多个应用程序、其它程序模块以及程序数据,这些示例中的每一个或某种组合中可能包括网络环境的实现。
总线230可以为表示几类总线结构中的一种或多种,包括存储单元总线或者存储单元控制器、外围总线、图形加速端口、处理单元或者使用多种总线结构中的任意总线结构的局域总线。
电子设备200也可以与一个或多个外部设备300(例如键盘、指向设备、蓝牙设备等)通信,还可与一个或者多个使得用户能与该电子设备200交互的设备通信,和/或与使得该电子设备200能与一个或多个其它计算设备进行通信的任何设备(例如路由器、调制解调器等等)通信。这种通信可以通过输入/输出(I/O)接口250进行。并且,电子设备200还可以通过网络适配器260与一个或者多个网络(例如局域网(LAN),广域网(WAN)和/或公共网络,例如因特网)通信。网络适配器260可以通过总线230与电子设备200的其它模块通信。应当明白,尽管图中未示出,可以结合电子设备200使用其它硬件和/或软件模块,包括但不限于:微代码、设备驱动器、冗余处理单元、外部磁盘驱动阵列、RAID系统、磁带驱动器以及数据备份存储系统等。
通过以上的实施方式的描述,本领域的技术人员易于理解,这里描述的示例实施方式可以通过软件实现,也可以通过软件结合必要的硬件的方式来实现。因此,根据本公开实施方式的技术方案可以以软件产品的形式体现出来,该软件产品可以存储在一个非易失性存储介质(可以是CD-ROM,U盘,移动硬盘 等)中或网络上,包括若干指令以使得一台计算设备(可以是个人计算机、服务器、或者网络设备等)执行根据本公开实施方式的上述方法。
本申请的实施方式还提供了一种计算机程序产品400,计算机程序产品400包括存储在非易失性计算机可读存储介质上的计算程序,计算机程序包括程序指令,当程序指令被计算机执行时,使计算机执行上述的基于光纤通信网络的机器人控制方法。
图9是根据本申请第九实施方式示出的一种计算机可读存储介质的结构示意图。
参考图9所示,描述了根据本公开的实施方式的用于实现上述方法的程序产品400,其可以采用便携式紧凑盘只读存储器(CD-ROM)并包括程序代码,并可以在终端设备,例如个人电脑上运行。然而,本公开的程序产品不限于此,在本文件中,可读存储介质可以是任何包含或存储程序的有形介质,该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。
可以以一种或多种程序设计语言的任意组合来编写用于执行本公开操作的程序代码,所述程序设计语言包括面向对象的程序设计语言—诸如Java、C++等,还包括常规的过程式程序设计语言—诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户计算设备上部分在远程计算设备上执行、或者完全在远程计算设备或服务器上执行。在涉及远程计算设备的情形中,远程计算设备可以通过任意种类的网络,包括局域网(LAN)或广域网(WAN),连接到用户计算设备,或者,可以连接到外部计算设备(例如利用因特网服务提供商来通过因特网连接)。
上述计算机可读介质承载有一个或者多个程序,当上述一个或者多个程序被一个该设备执行时,使得该计算机可读介质实现如下功能:将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度;在多个第一相似度中的最大第一相似度位于第一相似度门限内时,提取多个第一相似度中的部分第一相似度;通过部分第一相似度对应的多个第一图片确定多个第二图片;将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度;以及根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。
本领域技术人员可以理解上述各模块可以按照实施例的描述分布于装置中,也可以进行相应变化唯一不同于本实施例的一个或多个装置中。上述实施例的模块可以合并为一个模块,也可以进一步拆分成多个子模块。
通过以上的实施例的描述,本领域的技术人员易于理解,这里描述的示例实施例可以通过软件实现,也可以通过软件结合必要的硬件的方式来实现。因此,根据本公开实施例的技术方案可以以软件产品的形式体现出来,该软件产品可以存储在一个非易失性存储介质(可以是CD-ROM,U盘,移动硬盘等)中或网络上,包括若干指令以使得一台计算设备(可以是个人计算机、服务器、移动终端、或者网络设备等)执行根据本公开实施例的方法。
以上具体地示出和描述了本公开的示例性实施例。应可理解的是,本公开不限于这里描述的详细结构、设置方式或实现方法;相反,本公开意图涵盖包含在所附权利要求的精神和范围内的各种修改和等效设置。
Claims (18)
- 一种面部识别方法,其特征在于,包括:将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度;在多个第一相似度中的最大第一相似度位于第一相似度门限内时,提取多个第一相似度中的部分第一相似度;通过部分第一相似度对应的多个第一图片确定多个第二图片;将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度;以及根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。
- 如权利要求1所述的方法,其特征在于,还包括:在多个第一相似度中的最大第一相似度大于第一相似度门限时,将最大的第一相似度对应的第一图片作为所述待识别面部图像的识别结果。
- 如权利要求2所述的方法,其特征在于,还包括:根据所述识别结果确定待识别面部图像在数据库中对应的用户标识;根据所述用户标识在所述数据库中对应的第一图片数量确定后续处理方式。
- 如权利要求3所述的方法,其特征在于,根据所述用户标识在所述数据库中对应的第一图片数量确定后续处理方式包括:根据所述用户标识在所述数据库中对应的第一图片数量将所述待识别的面部图像添加至所述数据库;或根据所述用户标识在所述数据库中对应的第一图片数量用所述待识别的面部图像更新所述数据库内的第一图像。
- 如权利要求1所述的方法,其特征在于,第一相似度比较为面部特征比较,所述将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度包括:提取待识别的面部图像的特征点;将所述特征点与数据库中多个第一图片的特征点进行特征比对;以及根据比对结果获取所述多个第一相似度。
- 如权利要求1所述的方法,其特征在于,提取多个第一相似度中的部分第一相似度包括:提取多个第一相似度中的所有在所述第一相似度门限内的第一相似度。
- 如权利要求1所述的方法,其特征在于,通过部分第一相似度对应的多个第一图片确定多个第二图片包括:确定部分第一相似度对应的多个第一图片在数据库中对应的用户标识;将具有相同用户标识的第一图片进行合并;以及根据合并结果生成多个第二图片以及对应的多个第一相似度。
- 如权利要求7所述的方法,其特征在于,所述根据合并结果生成多个第二图片以及对应的多个第一相似度包括:将具有相同用户标识的多个第一图片对应的多个第一相似度进行权重累加合并。
- 如权利要求1所述的方法,其特征在于,第二相似度比较包括年龄比较与性别比较,所述将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度包括:将待识别的面部图像与多个第二图片进行年龄比较,生成年龄相似度;将待识别的面部图像与多个第二图片进行性别比较,生成性别相似度;通过年龄相似度与性别相似度确定所述第二相似度。
- 如权利要求9所述的方法,其特征在于,将待识别的面部图像与多个第二图片进行年龄比较,生成年龄相似度包括:确定所述待识别的面部图像的第一年龄;确定所述多个第二图片的多个第二年龄;以及通过向量距离计算确定第一年龄与多个第二年龄之间的年龄相似度。
- 如权利要求10所述的方法,其特征在于,向量距离计算包括以下至少一者:欧式距离计算、方差计算、以及余弦距离计算。
- 如权利要求1所述的方法,其特征在于,所述根据所述多个第二相似 度确定所述待识别的面部图像的面部特征识别结果包括:将多个第二相似度中的最大第二相似度对应的第二图片作为所述待识别面部图像的识别结果。
- 一种面部识别装置,其特征在于,包括:第一比较模块,用于将待识别的面部图像与数据库中多个第一图片进行第一相似度比较,获取多个第一相似度;门限模块,用于在多个第一相似度中的最大第一相似度位于第一相似度门限内时,提取多个第一相似度中的部分第一相似度;第二图片生成模块,用于通过部分第一相似度对应的多个第一图片确定多个第二图片;第二比较模块,用于将待识别的面部图像与多个第二图片进行第二相似度比较,获取多个第二相似度;以及第一结果模块,用于根据所述多个第二相似度确定所述待识别的面部图像的面部特征识别结果。
- 如权利要求13所述的装置,其特征在于,还包括:第二结果模块,用于在多个第一相似度中的最大第一相似度大于第一相似度门限时,将最大的第一相似度对应的第一图片作为所述待识别面部图像的识别结果。
- 如权利要求14所述的装置,其特征在于,还包括:用户标识模块,用于根据所述识别结果确定待识别面部图像在数据库中对应的用户标识;以及图片更新模块,用于根据所述用户标识在所述数据库中对应的第一图片数量确定后续处理方式。
- 一种电子设备,其特征在于,包括:一个或多个处理器;存储装置,用于存储一个或多个程序;当所述一个或多个程序被所述一个或多个处理器执行,使得所述一个或多个处理器实现如权利要求1-12中任一所述的方法。
- 一种计算机可读介质,其上存储有计算机程序,其特征在于,所述程序被处理器执行时实现如权利要求1-12中任一所述的方法。
- 一种计算机程序产品,所述计算机程序产品包括存储在非易失性计算机可读存储介质上的计算程序,所述计算机程序包括程序指令,当所述程序指令被计算机执行时,使所述计算机执行如权利要求1-12中任一项所述的方法。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810971687.1A CN109117808B (zh) | 2018-08-24 | 2018-08-24 | 面部识别方法、装置、电子设备及计算机可读介质 |
| CN201810971687.1 | 2018-08-24 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020038136A1 true WO2020038136A1 (zh) | 2020-02-27 |
Family
ID=64860653
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2019/095286 Ceased WO2020038136A1 (zh) | 2018-08-24 | 2019-07-09 | 面部识别方法、装置、电子设备及计算机可读介质 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109117808B (zh) |
| WO (1) | WO2020038136A1 (zh) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111414949A (zh) * | 2020-03-13 | 2020-07-14 | 杭州海康威视系统技术有限公司 | 图片聚类方法、装置、电子设备及存储介质 |
| CN111488919A (zh) * | 2020-03-24 | 2020-08-04 | 北京迈格威科技有限公司 | 目标识别方法及装置,电子设备以及计算机可读存储介质 |
| CN111814570A (zh) * | 2020-06-12 | 2020-10-23 | 深圳禾思众成科技有限公司 | 一种基于动态阈值的人脸识别方法、系统及存储介质 |
| CN111914769A (zh) * | 2020-08-06 | 2020-11-10 | 腾讯科技(深圳)有限公司 | 用户合法性判定方法、装置、计算机可读存储介质及设备 |
| CN112287753A (zh) * | 2020-09-23 | 2021-01-29 | 武汉天宝莱信息技术有限公司 | 一种基于机器学习提升人脸识别精度的系统及其算法 |
| CN112579803A (zh) * | 2020-11-16 | 2021-03-30 | 北京迈格威科技有限公司 | 一种图像数据清洗方法、装置、电子设备及存储介质 |
| CN112861742A (zh) * | 2021-02-18 | 2021-05-28 | 咪咕文化科技有限公司 | 人脸识别方法、装置、电子设备及存储介质 |
| CN113269154A (zh) * | 2021-06-29 | 2021-08-17 | 北京市商汤科技开发有限公司 | 一种图像识别方法、装置、设备及存储介质 |
| CN113408348A (zh) * | 2021-05-14 | 2021-09-17 | 桂林电子科技大学 | 一种基于视频的人脸识别方法、装置及存储介质 |
| CN114078271A (zh) * | 2020-08-21 | 2022-02-22 | 浙江宇视科技有限公司 | 阈值确定方法、目标人员识别方法、装置、设备及介质 |
| CN114863575A (zh) * | 2022-05-25 | 2022-08-05 | 深圳市巨龙创视科技有限公司 | 生物特征识别方法、装置、计算机设备及存储介质 |
| CN115019346A (zh) * | 2022-06-22 | 2022-09-06 | 浙江大华技术股份有限公司 | 一种信息检测方法及装置 |
| CN115880761A (zh) * | 2023-02-09 | 2023-03-31 | 数据空间研究院 | 基于策略优化的人脸识别方法、系统、存储介质及应用 |
| CN115966007A (zh) * | 2022-12-28 | 2023-04-14 | 支付宝(杭州)信息技术有限公司 | 一种对抗攻击检测方法、装置以及设备 |
| CN117037249A (zh) * | 2023-08-09 | 2023-11-10 | 世优(北京)科技有限公司 | 人脸图像的分堆方法、装置及电子设备 |
| CN117292425A (zh) * | 2023-10-09 | 2023-12-26 | 杭州网易云音乐科技有限公司 | 人像识别方法、装置、存储介质及电子设备 |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109117808B (zh) * | 2018-08-24 | 2020-11-03 | 深圳前海达闼云端智能科技有限公司 | 面部识别方法、装置、电子设备及计算机可读介质 |
| CN109961031A (zh) * | 2019-01-25 | 2019-07-02 | 深圳市星火电子工程公司 | 人脸融合识别认定方法、目标人员信息显示方法、预警布控方法及系统 |
| CN109977765A (zh) * | 2019-02-13 | 2019-07-05 | 平安科技(深圳)有限公司 | 人脸图像识别方法、装置及计算机设备 |
| CN109902644A (zh) * | 2019-03-07 | 2019-06-18 | 北京海益同展信息科技有限公司 | 人脸识别方法、装置、设备和计算机可读介质 |
| CN109800744B (zh) * | 2019-03-18 | 2021-08-20 | 深圳市商汤科技有限公司 | 图像聚类方法及装置、电子设备和存储介质 |
| CN110110115B (zh) * | 2019-03-28 | 2023-05-23 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | 人机交互筛选识别机载目标图像的方法 |
| CN110222662B (zh) * | 2019-06-12 | 2020-11-03 | 北京博雅英杰科技股份有限公司 | 对象的识别方法、装置及存储介质 |
| CN112149479B (zh) * | 2019-06-28 | 2024-07-26 | 西安光启智能技术有限公司 | 人脸识别的方法、存储介质和电子装置 |
| CN110825765B (zh) * | 2019-10-23 | 2022-10-04 | 中国建设银行股份有限公司 | 一种人脸识别的方法和装置 |
| CN111353419B (zh) * | 2020-02-26 | 2023-08-11 | 北京百度网讯科技有限公司 | 图像比对方法、装置、电子设备及存储介质 |
| CN112199998B (zh) * | 2020-09-09 | 2023-06-20 | 浙江大华技术股份有限公司 | 一种人脸识别方法、装置、设备及介质 |
| EP4226275A1 (en) * | 2020-10-12 | 2023-08-16 | Assa Abloy Ab | Access control with face recognition and heterogeneous information |
| CN112364801A (zh) * | 2020-11-19 | 2021-02-12 | 天津中科虹星科技有限公司 | 一种动态阈值人脸识别方法 |
| CN114743229A (zh) * | 2021-01-07 | 2022-07-12 | 腾讯科技(深圳)有限公司 | 一种基于人工智能的人脸识别方法、设备和存储介质 |
| CN115331278A (zh) * | 2021-05-11 | 2022-11-11 | 阿里巴巴新加坡控股有限公司 | 数据处理方法、装置、电子设备和存储介质 |
| CN114155590B (zh) * | 2021-12-02 | 2025-04-18 | 杭州国芯微电子股份有限公司 | 一种人脸识别方法 |
| CN114238223A (zh) * | 2021-12-17 | 2022-03-25 | 深圳市雷鸟网络传媒有限公司 | 图片移除方法、装置、计算机设备及计算机可读存储介质 |
| CN114429644A (zh) * | 2021-12-31 | 2022-05-03 | 杭州萤石软件有限公司 | 图像底库的更新方法、装置、设备及存储介质 |
| CN115565017B (zh) * | 2022-10-31 | 2026-01-20 | 北京京东拓先科技有限公司 | 食材识别方法及装置、存储介质及电子设备 |
| CN116758607B (zh) * | 2023-05-16 | 2025-09-23 | 杭州臻稀生物科技有限公司 | 基于眼部与面部特征的人员识别方法 |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016073105A1 (en) * | 2014-11-06 | 2016-05-12 | Intel Corporation | Face recognition using gradient based feature analysis |
| CN105814582A (zh) * | 2013-11-30 | 2016-07-27 | 北京市商汤科技开发有限公司 | 用于识别人脸的方法和系统 |
| CN106056083A (zh) * | 2016-05-31 | 2016-10-26 | 腾讯科技(深圳)有限公司 | 一种信息处理方法及终端 |
| CN108009521A (zh) * | 2017-12-21 | 2018-05-08 | 广东欧珀移动通信有限公司 | 人脸图像匹配方法、装置、终端及存储介质 |
| CN109117808A (zh) * | 2018-08-24 | 2019-01-01 | 深圳前海达闼云端智能科技有限公司 | 面部识别方法、装置、电子设备及计算机可读介质 |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184238A (zh) * | 2015-08-26 | 2015-12-23 | 广西小草信息产业有限责任公司 | 一种人脸识别方法及系统 |
| CN108009465B (zh) * | 2016-10-31 | 2021-08-27 | 杭州海康威视数字技术股份有限公司 | 一种人脸识别方法及装置 |
| CN106815566B (zh) * | 2016-12-29 | 2021-04-16 | 天津中科智能识别产业技术研究院有限公司 | 一种基于多任务卷积神经网络的人脸检索方法 |
| CN107886079A (zh) * | 2017-11-22 | 2018-04-06 | 北京旷视科技有限公司 | 对象识别方法、装置及系统 |
| CN107944020B (zh) * | 2017-12-11 | 2019-12-17 | 深圳云天励飞技术有限公司 | 人脸图像查找方法及装置、计算机装置和存储介质 |
| CN108021908B (zh) * | 2017-12-27 | 2020-06-16 | 深圳云天励飞技术有限公司 | 人脸年龄段识别方法及装置、计算机装置及可读存储介质 |
-
2018
- 2018-08-24 CN CN201810971687.1A patent/CN109117808B/zh active Active
-
2019
- 2019-07-09 WO PCT/CN2019/095286 patent/WO2020038136A1/zh not_active Ceased
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105814582A (zh) * | 2013-11-30 | 2016-07-27 | 北京市商汤科技开发有限公司 | 用于识别人脸的方法和系统 |
| WO2016073105A1 (en) * | 2014-11-06 | 2016-05-12 | Intel Corporation | Face recognition using gradient based feature analysis |
| CN106056083A (zh) * | 2016-05-31 | 2016-10-26 | 腾讯科技(深圳)有限公司 | 一种信息处理方法及终端 |
| CN108009521A (zh) * | 2017-12-21 | 2018-05-08 | 广东欧珀移动通信有限公司 | 人脸图像匹配方法、装置、终端及存储介质 |
| CN109117808A (zh) * | 2018-08-24 | 2019-01-01 | 深圳前海达闼云端智能科技有限公司 | 面部识别方法、装置、电子设备及计算机可读介质 |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111414949B (zh) * | 2020-03-13 | 2023-06-27 | 杭州海康威视系统技术有限公司 | 图片聚类方法、装置、电子设备及存储介质 |
| CN111414949A (zh) * | 2020-03-13 | 2020-07-14 | 杭州海康威视系统技术有限公司 | 图片聚类方法、装置、电子设备及存储介质 |
| CN111488919A (zh) * | 2020-03-24 | 2020-08-04 | 北京迈格威科技有限公司 | 目标识别方法及装置,电子设备以及计算机可读存储介质 |
| CN111488919B (zh) * | 2020-03-24 | 2023-12-22 | 北京迈格威科技有限公司 | 目标识别方法及装置,电子设备以及计算机可读存储介质 |
| CN111814570B (zh) * | 2020-06-12 | 2024-04-30 | 深圳禾思众成科技有限公司 | 一种基于动态阈值的人脸识别方法、系统及存储介质 |
| CN111814570A (zh) * | 2020-06-12 | 2020-10-23 | 深圳禾思众成科技有限公司 | 一种基于动态阈值的人脸识别方法、系统及存储介质 |
| CN111914769A (zh) * | 2020-08-06 | 2020-11-10 | 腾讯科技(深圳)有限公司 | 用户合法性判定方法、装置、计算机可读存储介质及设备 |
| CN111914769B (zh) * | 2020-08-06 | 2024-01-26 | 腾讯科技(深圳)有限公司 | 用户合法性判定方法、装置、计算机可读存储介质及设备 |
| CN114078271A (zh) * | 2020-08-21 | 2022-02-22 | 浙江宇视科技有限公司 | 阈值确定方法、目标人员识别方法、装置、设备及介质 |
| CN112287753B (zh) * | 2020-09-23 | 2024-09-06 | 武汉天宝莱信息技术有限公司 | 一种基于机器学习提升人脸识别精度的系统及其算法 |
| CN112287753A (zh) * | 2020-09-23 | 2021-01-29 | 武汉天宝莱信息技术有限公司 | 一种基于机器学习提升人脸识别精度的系统及其算法 |
| CN112579803B (zh) * | 2020-11-16 | 2024-04-02 | 北京迈格威科技有限公司 | 一种图像数据清洗方法、装置、电子设备及存储介质 |
| CN112579803A (zh) * | 2020-11-16 | 2021-03-30 | 北京迈格威科技有限公司 | 一种图像数据清洗方法、装置、电子设备及存储介质 |
| CN112861742A (zh) * | 2021-02-18 | 2021-05-28 | 咪咕文化科技有限公司 | 人脸识别方法、装置、电子设备及存储介质 |
| CN113408348A (zh) * | 2021-05-14 | 2021-09-17 | 桂林电子科技大学 | 一种基于视频的人脸识别方法、装置及存储介质 |
| CN113269154B (zh) * | 2021-06-29 | 2023-10-24 | 北京市商汤科技开发有限公司 | 一种图像识别方法、装置、设备及存储介质 |
| CN113269154A (zh) * | 2021-06-29 | 2021-08-17 | 北京市商汤科技开发有限公司 | 一种图像识别方法、装置、设备及存储介质 |
| CN114863575A (zh) * | 2022-05-25 | 2022-08-05 | 深圳市巨龙创视科技有限公司 | 生物特征识别方法、装置、计算机设备及存储介质 |
| CN115019346B (zh) * | 2022-06-22 | 2025-07-25 | 浙江大华技术股份有限公司 | 一种信息检测方法及装置 |
| CN115019346A (zh) * | 2022-06-22 | 2022-09-06 | 浙江大华技术股份有限公司 | 一种信息检测方法及装置 |
| CN115966007A (zh) * | 2022-12-28 | 2023-04-14 | 支付宝(杭州)信息技术有限公司 | 一种对抗攻击检测方法、装置以及设备 |
| CN115880761A (zh) * | 2023-02-09 | 2023-03-31 | 数据空间研究院 | 基于策略优化的人脸识别方法、系统、存储介质及应用 |
| CN117037249A (zh) * | 2023-08-09 | 2023-11-10 | 世优(北京)科技有限公司 | 人脸图像的分堆方法、装置及电子设备 |
| CN117292425A (zh) * | 2023-10-09 | 2023-12-26 | 杭州网易云音乐科技有限公司 | 人像识别方法、装置、存储介质及电子设备 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109117808B (zh) | 2020-11-03 |
| CN109117808A (zh) | 2019-01-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109117808B (zh) | 面部识别方法、装置、电子设备及计算机可读介质 | |
| US12175684B2 (en) | Pedestrian tracking method, computing device, pedestrian tracking system and storage medium | |
| US10509985B2 (en) | Method and apparatus for security inspection | |
| CN114998934B (zh) | 基于多模态智能感知和融合的换衣行人重识别和检索方法 | |
| CN109472198B (zh) | 一种姿态鲁棒的视频笑脸识别方法 | |
| CN112016464A (zh) | 检测人脸遮挡的方法、装置、电子设备及存储介质 | |
| WO2022100337A1 (zh) | 人脸图像质量评估方法、装置、计算机设备及存储介质 | |
| WO2017107957A9 (zh) | 人脸图像的检索方法及装置 | |
| CN107944020A (zh) | 人脸图像查找方法及装置、计算机装置和存储介质 | |
| CN112487886A (zh) | 一种有遮挡的人脸识别方法、装置、存储介质及终端 | |
| WO2021208601A1 (zh) | 基于人工智能的图像处理方法、装置、设备及存储介质 | |
| US20220147735A1 (en) | Face-aware person re-identification system | |
| WO2022105118A1 (zh) | 基于图像的健康状态识别方法、装置、设备及存储介质 | |
| CN111814620A (zh) | 人脸图像质量评价模型建立方法、优选方法、介质及装置 | |
| CN101493887A (zh) | 基于半监督学习和哈希索引的眉毛图像分割方法 | |
| CN105469117B (zh) | 一种基于鲁棒特征提取的图像识别方法与装置 | |
| Haji et al. | Real time face recognition system (RTFRS) | |
| CN110188630A (zh) | 一种人脸识别方法和相机 | |
| CN110232331A (zh) | 一种在线人脸聚类的方法及系统 | |
| CN107563319A (zh) | 一种基于图像的亲子间人脸相似性度量计算方法 | |
| WO2020232697A1 (zh) | 一种在线人脸聚类的方法及系统 | |
| CN111666976A (zh) | 基于属性信息的特征融合方法、装置和存储介质 | |
| Shashikala et al. | Attendance monitoring system using face recognition | |
| CN115331280A (zh) | 人脸身份识别方法、装置及电子设备 | |
| Ketab et al. | LipPrint: Using Lip Movements as a Silent Password |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19851409 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 10/06/2021) |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 19851409 Country of ref document: EP Kind code of ref document: A1 |