WO2020034645A1 - 人脸识别方法、人脸识别系统、及电子设备 - Google Patents

人脸识别方法、人脸识别系统、及电子设备 Download PDF

Info

Publication number
WO2020034645A1
WO2020034645A1 PCT/CN2019/081186 CN2019081186W WO2020034645A1 WO 2020034645 A1 WO2020034645 A1 WO 2020034645A1 CN 2019081186 W CN2019081186 W CN 2019081186W WO 2020034645 A1 WO2020034645 A1 WO 2020034645A1
Authority
WO
WIPO (PCT)
Prior art keywords
comparison
value
face
comparison value
pictures
Prior art date
Application number
PCT/CN2019/081186
Other languages
English (en)
French (fr)
Inventor
汪辉
Original Assignee
浙江宇视科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 浙江宇视科技有限公司 filed Critical 浙江宇视科技有限公司
Priority to EP19849095.5A priority Critical patent/EP3839806B1/en
Priority to US17/268,127 priority patent/US11967176B2/en
Publication of WO2020034645A1 publication Critical patent/WO2020034645A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06V40/168—Feature extraction; Face representation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06V40/168—Feature extraction; Face representation
    • G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships

Definitions

  • the present disclosure relates to the field of face recognition technology, for example, to a face recognition method, a face recognition system, and an electronic device.
  • Face recognition is a technology for identity recognition based on facial feature information of a person.
  • the feature is extracted and compared with the feature information stored in the database to obtain the comparison result, and then the identity is identified.
  • the accuracy of face recognition needs to be improved due to changes in the age of people, changes in makeup and posture.
  • the present disclosure provides a face recognition method, an electronic device, a computer-readable storage medium, and a face recognition system.
  • the present disclosure provides a face recognition method.
  • the method includes:
  • Collect a face picture extract a face feature, and compare the face feature with a plurality of pre-stored original pictures respectively to obtain a first required comparison value, and compare the face feature with a plurality of pre-stored scene-collected pictures respectively Compare to get the second required comparison value.
  • a face feature is re-extracted, and the person to be re-extracted
  • the facial features are compared with a plurality of the original pictures to obtain a third required contrast value, and it is determined whether the third required contrast value is greater than or equal to the preset contrast threshold.
  • the face picture is saved as a new scene collection picture, the number of times that the scene collection picture is successfully compared is refreshed, and the identification is determined to pass .
  • the present disclosure also provides an electronic device including:
  • At least one processor At least one processor
  • Memory configured to store at least one computer program
  • the at least one processor When the at least one computer program is executed by the at least one processor, the at least one processor implements the method as described above.
  • the present disclosure also provides a computer-readable storage medium storing computer-executable instructions for performing the method as described above.
  • the present disclosure also provides a face recognition system.
  • the face recognition system includes an image acquisition device and an electronic device, and the image acquisition device is communicatively connected to the electronic device;
  • the image acquisition device is configured to acquire a face picture and transfer the face picture to the electronic device.
  • the electronic device is an electronic device as described above.
  • FIG. 1 is a schematic diagram of a face recognition system provided by the present disclosure.
  • FIG. 2 is a schematic flowchart of a face recognition method provided by the present disclosure.
  • FIG. 3 is another schematic flowchart of a face recognition method provided by the present disclosure.
  • FIG. 4 is another schematic flowchart of a face recognition method provided by the present disclosure.
  • Icons 10-face recognition system; 11-image acquisition equipment; 12-electronic equipment; 121-memory; 122-image processing module.
  • face recognition is required in many scenarios to verify the identity of a person.
  • face recognition accuracy is not high.
  • the present disclosure provides a face recognition method.
  • the method includes:
  • Collect a face picture extract a face feature, and compare the face feature with a plurality of pre-stored original pictures respectively to obtain a first required comparison value, and compare the face feature with a plurality of pre-stored scene-collected pictures respectively Compare them to get the second required comparison value;
  • a face feature is re-extracted, and the person to be re-extracted Comparing the facial features with a plurality of the original pictures to obtain a third required comparison value, and determining whether the third required comparison value is greater than or equal to the preset contrast threshold;
  • the face picture is saved as a new scene collection picture, the number of times that the scene collection picture is successfully compared is refreshed, and the identification is determined to pass .
  • the method further includes determining that the comparison is unsuccessful based on a determination result that the third required comparison value is less than the preset comparison threshold, and determining that the identification fails.
  • the method further includes:
  • the first required contrast value is greater than or equal to the preset contrast threshold, and that the original picture corresponding to the first required contrast value and the field-collected picture corresponding to the second required contrast value are the same person
  • the judgment result of the picture judges that the comparison is successful, and judges that the recognition is passed.
  • the method further includes:
  • the first required contrast value is greater than or equal to the preset contrast threshold
  • the original picture corresponding to the first required contrast value and the second required contrast value correspond to
  • the method further includes:
  • the comparison is determined to be successful, and the second location is refreshed. It is necessary to compare the number of successful comparisons of on-site collected pictures corresponding to the numerical values, and determine that the identification has passed.
  • the method further includes:
  • the re-extracted face feature corresponds to a comparison result of a field-collected picture corresponding to the second required comparison value that is smaller than the preset comparison threshold value, it is determined that the comparison fails, and the second required value is cleared.
  • the method further includes:
  • the method further includes:
  • the method further includes:
  • the second required contrast value corresponds to based on a determination result that the first required contrast value is less than the preset contrast threshold and the second required contrast value is greater than or equal to the preset contrast threshold Whether the number of successful comparisons of on-site image collection is greater than or equal to two;
  • the method further includes: based on a determination result that the number of successful comparisons of the field-collected pictures corresponding to the second required comparison value is less than two times, re-extracting the facial features, and re-extracting the faces Compare the characteristics with the on-site collected pictures corresponding to the second required comparison value;
  • the comparison value is greater than or equal to the preset comparison threshold, it is determined that the comparison is successful, and the refresh is based on the second required comparison.
  • the method further includes:
  • the deletion is based on the second comparison value.
  • the number of successful comparisons of on-site collected pictures corresponding to the comparison value, the second required comparison value is deleted, and the identification is determined to fail.
  • the face picture is collected, the face feature is extracted, and the face feature is compared with a plurality of pre-stored original pictures to obtain a first required comparison value, and the face feature is compared with The pre-stored multiple scene collection pictures are compared separately to obtain the second required comparison value, including:
  • Detect whether there is a face to be identified based on the judgment result of the existence of the face to be identified, collect a face picture, extract a face feature, and compare the face feature with a plurality of pre-stored original pictures to obtain the first required
  • the comparison value is obtained by comparing the facial features with a plurality of pre-stored scene-collected pictures to obtain a second required comparison value.
  • FIG. 1 is a schematic diagram of a face recognition system 10 provided by the present disclosure.
  • the face recognition system 10 can be applied to a quick pass door.
  • the face recognition system 10 includes an image acquisition device 11 and an electronic device 12.
  • the electronic device 12 includes a memory 121 and at least one processor. Each of the processors may specifically include an image processing module 122.
  • the image acquisition device 11 is communicatively connected to the electronic device 12, and the image acquisition device 11 is used for For obtaining a face picture, and transmitting the obtained face picture to the electronic device 12.
  • the memory 121 stores a computer program that can be run on the image processing module 122.
  • the image processing module 122 runs the computer program
  • the electronic device 12 executes the following face recognition method.
  • FIG. 2 Please refer to FIG. 2 for a flowchart of a face recognition method provided by the present disclosure. The method can be applied to the face recognition system 10 shown in FIG. 1 and executed by the electronic device 12 in the face recognition system 10.
  • This method can be applied to the electronic device 12, which has pre-stored original pictures of multiple persons and pictures collected on the spot, and the number of successful comparisons based on each of the pictures collected on the spot.
  • the method may include:
  • Step S10 Detect whether a face to be identified exists within a valid recognition distance, and if a face to be identified exists, step S11 to step S14 are performed. If there is no face to be identified, the process returns to step S10.
  • Step S11 Collect a face picture and extract facial features.
  • Step S12 comparing the facial features with a plurality of the original pictures, respectively, to obtain a first required comparison value; comparing the facial features with a plurality of the scene collected pictures, respectively, to obtain a second image. Need to compare values.
  • the face features are compared with a plurality of the original pictures to obtain a plurality of contrast values.
  • the highest contrast value is the first required contrast value.
  • the highest comparison value is the second required comparison value. The higher the comparison value, the higher the similarity between the collected face picture and the original picture and the scene collected picture.
  • Step S13 determine whether the first required comparison value and the second required comparison value are greater than or equal to a preset contrast threshold, and if the first required comparison value is less than the preset comparison threshold and the If the second required comparison value is also smaller than the preset comparison threshold, step S14 is performed.
  • Step S14 Re-extracting the facial features, comparing the re-extracted facial features with a plurality of the original pictures, to obtain a third required comparison value, and determining whether the third required comparison value is greater than or equal to the preset For the comparison threshold value, if the third required comparison value is greater than or equal to the preset comparison threshold value, step S15 is performed; if the third required comparison value is less than the preset comparison threshold value, step S16 is performed.
  • Step S15 Save the face picture as a new scene capture picture to the electronic device 12, refresh the number of successful comparisons of the scene capture picture, and determine that the recognition is successful.
  • the first required contrast value is smaller than the preset contrast threshold
  • the second required contrast value is also smaller than the preset contrast threshold
  • re-extract facial features and perform the process again.
  • Face recognition The re-extracted face features are different from the previously extracted face features.
  • the re-extracted face features can be more detailed than the previously extracted face features, and the accuracy of the re-extracted face features can be more accurate. high.
  • the new live capture picture is not completed in the inventory, but only in the next face recognition, After another comparison, the new on-site collected pictures that have been successfully identified will be stored in the library. After multiple verifications and comparisons, the on-site collected pictures are more adaptive in the face recognition process.
  • the collected face picture when the collected face picture is saved to the electronic device 12 as a new on-site captured picture, it can also be determined whether the collected face picture meets a preset condition, wherein the comparison is successful and A face picture that meets the preset condition is saved as a new live capture picture in the electronic device 12, and if a live capture picture is stored in the electronic device 12, it is replaced.
  • multiple on-site captured pictures pre-stored in the electronic device 12 may correspond to different persons, that is, at most one picture of each person is stored in the electronic device 12. If the electronic device 12 stores a live collection picture of the same person as the new live collection picture, the live collection picture may be replaced with the new live collection picture to implement replacement processing.
  • the preset condition includes: judging whether the quality of the collected face picture meets a preset threshold for the scene to collect pictures into the database, and judging whether the collected face distance is at least one of a preset distance range.
  • the collected face distance detection can predict the face distance by collecting the face size or geometric space model.
  • the process of determining whether the preset condition is met may include: if the quality of the collected face picture meets the threshold of the on-site captured picture storage and the collected face distance is within a preset Within a distance range, the comparison is successful, the collected face picture is saved as a new live capture picture to the electronic device 12, the number of successful comparisons based on the new live capture picture is refreshed, and the recognition is determined to pass
  • the process of determining whether the preset condition is met may include: if the quality of the collected face picture meets the threshold of the on-site captured picture storage, then the comparison is successful and the recognition is determined to pass;
  • the process of determining whether the preset conditions are met may include: if the collected face distance is within a preset distance range, the comparison is successful, and the recognition is passed; in one embodiment, whether the The process that meets the preset condition may include: if the quality of the collected face picture does not meet the threshold of the on-site captured picture storage and the collected face Not from within the predetermined distance range, the comparison
  • refreshing the number of successful comparisons of the on-site collected pictures is specifically: refreshing the number of successful comparisons of the new on-site collected pictures.
  • the refreshing the number of successful comparisons of the on-site captured pictures may be specifically: refreshing the number of successful comparisons of the target on-site captured pictures among the pre-stored multiple on-site captured pictures.
  • the target scene collection picture may be a scene collection picture corresponding to the second required comparison value; the target scene collection picture may be a scene collection picture that can be successfully compared with the new scene collection picture;
  • the target scene collection picture may be a picture of the same person as the new scene collection picture.
  • Step S16 The comparison is unsuccessful and the identification fails.
  • the electronic device 12 may store an approximate mismatch database, and the approximate mismatch database stores a list of similar persons. If the first required comparison value is greater than or equal to a preset comparison threshold, step S21 is performed.
  • Step S21 determine whether the original picture corresponding to the first required contrast value and the on-site captured picture corresponding to the second required comparison value are pictures of the same person, and if the first required comparison value is greater than or equal to the A preset comparison threshold, and the original picture corresponding to the first required comparison value and the field-collected picture corresponding to the second required comparison value are pictures of the same person, step S22 is performed; if the first required The comparison value is greater than or equal to the preset comparison threshold, and the original picture corresponding to the first required comparison value and the on-site captured picture corresponding to the second required comparison value are different pictures of the same person, and step S23 is performed.
  • Step S22 The comparison is successful, and it is determined that the identification is passed.
  • the first required comparison value is stored in the electronic device 12 as a record of the current determination.
  • Step S23 Determine whether the second required contrast value is greater than or equal to the preset contrast threshold; if the second required contrast value is less than the preset contrast threshold, perform step S24, and if the second If the required comparison value is greater than or equal to the preset comparison threshold, step S25 is performed.
  • Step S24 The comparison is successful, and it is determined that the identification is passed.
  • the first required comparison value is stored in the electronic device 12 as a record of the current determination.
  • Step S25 determine whether the person corresponding to the original picture corresponding to the first required comparison value is in the approximate mismatch database, and if the person corresponding to the original picture corresponding to the first required comparison value is in the In the approximate mismatch database, step S26 is performed. If the person corresponding to the original picture corresponding to the first required contrast value is not in the approximate mismatch database, step S27 to step S28 are performed.
  • Step S26 It is determined whether the number of successful comparisons of on-site captured pictures corresponding to the second required comparison value is more than two (that is, greater than or equal to two times). If the number of times is more than two times, step S261 is performed; if the number of successful comparisons of on-site pictures corresponding to the second required comparison value is less than one time (that is, less than twice), step S262 is performed.
  • Step S261 The comparison is successful, and it is determined that the identification is passed.
  • the second required comparison value is stored in the electronic device 12 as a record of the judgment.
  • Step S262 Re-extracting the facial features, comparing the re-extracted facial features with the field-collected picture corresponding to the second required contrast value, and determining whether the re-extracted facial features correspond to the second required contrast value Whether the comparison value obtained from the comparison of the field-collected pictures is greater than or equal to the preset contrast threshold, and if the re-extracted face feature corresponds to the comparison value obtained from the field-collected pictures corresponding to the second required comparison value, the comparison value is greater than or equal to the Step S2621 is performed if the preset comparison threshold is exceeded. If the comparison value obtained by comparing the re-extracted face features with the second required comparison value of the on-site captured picture is smaller than the preset comparison threshold, execute step S2622.
  • Step S2621 The comparison is successful, and the number of successful comparisons of the on-site captured pictures corresponding to the second required comparison value is refreshed, and it is determined that the identification is passed.
  • the comparison value is used as the record of the judgment, and the number of successful comparisons of the field-collected pictures corresponding to the second required comparison value after refresh and the record of the judgment are stored in the electronic device. 12 in.
  • Step S2622 When the comparison fails, the number of successful comparisons of the on-site pictures collected based on the second required comparison value is cleared, the second required comparison value is deleted, and the identification is determined to fail.
  • the method further includes step S27: re-extracting facial features, and comparing the re-extracted facial features with a plurality of the original pictures to obtain a fourth required comparison value;
  • Step S28 Determine whether the fourth required contrast value is greater than or equal to the preset contrast threshold; if the fourth required contrast value is less than the preset contrast threshold, perform step S281. If the fourth required comparison value is greater than or equal to the preset comparison threshold, step S282 is performed.
  • Step S281 The comparison is unsuccessful, clearing the number of successful comparisons of the on-site captured pictures corresponding to the second required comparison value, deleting the second required comparison value, and determining that the identification fails.
  • Step S282 The comparison is successful, and it is determined that the identification is passed.
  • the fourth required comparison value is greater than or equal to the preset comparison threshold, the comparison is successful.
  • the first required comparison value is used as the record passed in this judgment and stored in the server, wherein the number of successful comparisons of the on-site picture collection based on the second required comparison value is deleted, and all The second required comparison value is described; if the original picture corresponding to the fourth required comparison value and the field-collected picture corresponding to the second required comparison value are pictures of the same person, the first required comparison value corresponds to the original picture.
  • the list of persons and the second required comparison value are stored in the approximate mismatch database corresponding to the pictures collected on the scene, and the second required comparison value is recorded as the record passed in this judgment and stored in the electronic device 12 Medium; if the original picture corresponding to the fourth required contrast value is the original picture corresponding to the first required contrast value and the second required
  • the on-site picture corresponding to the ratio value is a picture of a person who is not the same, and the fourth required comparison value is used as the record of the current identification and identification, and is stored in the electronic device 12, and the corresponding one based on the second required comparison value is cleared. The number of successful comparisons of pictures collected on site, and the second required comparison value is deleted.
  • step S31 if the first required contrast value is smaller than the preset contrast threshold, and the second required contrast value is greater than or equal to the preset contrast threshold, step S31 is performed.
  • Step S31 judging whether the number of successful comparisons of on-site captured images corresponding to the second required comparison value is more than two; if the number of successful comparisons of on-site captured images corresponding to the second required comparison value is two If the above (that is, two or more times) is performed, step S32 is performed; if the number of successful comparisons of the on-site captured pictures corresponding to the second required comparison value is less than one time (that is, less than two times), step S33 is performed.
  • Step S32 The comparison is successful, and it is determined that the identification is passed.
  • the second required comparison value is stored in the electronic device 12 as a record of the determination.
  • Step S33 re-extracting the facial features, comparing the re-extracted facial features with the on-site collected pictures corresponding to the second required contrast value, and determining that the re-extracted facial features correspond to the second required contrast value Whether the comparison value obtained from the comparison of the field-collected pictures is greater than or equal to the preset contrast threshold, and if the re-extracted face feature corresponds to the comparison value obtained from the field-collected pictures corresponding to the second required comparison value, the comparison value is greater than or equal to the Step S331 is performed if the preset comparison threshold is exceeded. If the comparison value obtained by comparing the re-extracted face features with the scene-acquired picture corresponding to the second required comparison value is smaller than the preset comparison threshold, step S332 is performed.
  • Step S331 The comparison is successful, and the number of successful comparisons of the on-site captured pictures corresponding to the second required comparison value is refreshed, and it is determined that the identification is passed.
  • the comparison value is used as the record of the judgment, and the number of successful comparisons of the field-collected pictures corresponding to the second required comparison value after the refresh and the record of the judgment are stored in the record.
  • Step S332 The comparison fails, the number of successful comparisons of the pictures collected on the scene based on the second required comparison value is cleared, the second required comparison value is deleted, and the identification is determined to fail.
  • the comparison is successful, and the number of successful comparisons of the field-collected pictures corresponding to the second required comparison value is refreshed, and the number of successful comparisons is added to the number of previous successful comparisons. For example, if the original comparison is successful The number of times is 1 time. After refreshing, the number of times of comparison is added to the number of times, and the number of successful comparisons becomes twice.
  • the comparison fails, and the number of times of successful comparison of on-site captured images corresponding to the second required comparison value is cleared.
  • the number of successful comparisons of on-site captured images is 1 time. After the comparison fails, 1 time is changed to zero, that is, Initialize the state, and delete the comparison value (that is, the second required comparison value) obtained by comparing the extracted human face with the scene-collected picture, and determine that the recognition fails.
  • the field-acquired pictures stored in the electronic device 12 will be continuously updated.
  • the updating method may be that during the face recognition process, people who are successfully compared and meet the preset conditions
  • the face collection picture is stored in the electronic device 12 as a new live collection picture instead of the existing live collection picture. It can also be used to periodically collect face pictures that meet the preset conditions during the face recognition process, and use the face picture as a new one.
  • the on-site collection pictures are stored in the electronic device 12 to update the on-site collection pictures in the electronic device 12.
  • the original pictures stored in the electronic device 12 may be periodically entered into new pictures to update the original pictures in the electronic device 12.
  • the present disclosure provides a face recognition system 10 that can be applied to a quick pass door.
  • the face recognition system 10 includes an image acquisition device 11 and an electronic device 12, and the image acquisition device 11 is communicatively connected with the electronic device 12.
  • the image acquisition device 11 is configured to acquire a face picture and transfer the face picture to the electronic device 12.
  • the electronic device 12 includes a memory 121, a processor (the processor may be the image processing module 122), and a computer program stored on the memory 121 and executable on the processor.
  • the processor executes all When the computer program is described, the electronic device 12 in the face recognition system 10 enables the above-mentioned face recognition method.
  • the memory 121 includes a personnel database, and the personnel database stores original pictures and field collected pictures of multiple people.
  • the image processing module 122 is configured to extract facial features in the face picture, and compare the extracted facial features with a plurality of the original pictures and the collected scene pictures, respectively, to the person. Face pictures for recognition.
  • the electronic device 12 may be a terminal device, and the terminal device is provided on the quick-pass door and performs real-time People are detected. For pedestrians who are going through the speed gate, face pictures are collected, facial features are extracted, and face recognition is performed. When face recognition is successful, the terminal device controls the gate to open the door, and pedestrians can pass through the gate. When the pedestrian passes, the gate automatically closes the door after a delay, and continues to detect the next pedestrian.
  • the face recognition system 10 is applied to a fast-pass door, if a pedestrian opens the door when it is too far away from the gate, there will be security risks, such as other pedestrians entering the gate first, etc.
  • the terminal device also needs to The distance detection is performed, and the gate release is performed only when the face is within a proper distance range.
  • the present disclosure sets the distance range to 0.3-1.5m.
  • the terminal device defines two distance modes, including a fast mode and a safe mode, wherein the fast mode distance threshold is greater than the safe mode distance threshold.
  • the fast mode is enabled, when the face distance is detected to be less than or equal to the fast mode distance threshold, the face recognition comparison verification process is started; when the safe mode is enabled, when the detected face distance is less than or equal to the safe mode distance threshold, the start Face recognition comparison verification processing.
  • a face recognition method and a face recognition system store original pictures of multiple persons and pictures collected on the spot by multiple persons in an electronic device, and by collecting the face pictures, extracting Face features. Compare the extracted face features with multiple original pictures and on-site collected pictures to identify the face. If the comparison fails, re-extract a more detailed face. The features are compared with multiple original pictures in the electronic device, and the face is re-recognized to ensure that there is no missing or mismatching in the face recognition process, and the accuracy of the face is effectively guaranteed.
  • the new scene collection picture needs to be checked and verified multiple times to ensure that the new scene collection picture has strong adaptability and is conducive to improving face recognition. Accuracy.
  • An embodiment of the present disclosure also provides a computer-readable storage medium storing computer-executable instructions, where the computer-executable instructions are used to execute the foregoing method.
  • All or part of the processes in the method of the above embodiment can be completed by executing related hardware through a computer program.
  • the program can be stored in a non-transitory computer-readable storage medium.
  • the method can include the method described above.
  • the process of the embodiment, wherein the non-transitory computer-readable storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a RAM.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Analysis (AREA)
  • Collating Specific Patterns (AREA)

Abstract

一种人脸识别方法、电子设备、计算机可读存储介质及人脸识别系统，所述方法包括：根据人脸特征得到第一所需对比数值和第二所需对比数值（S12）；若所述第一所需对比数值小于所述预设的对比阈值且所述第二所需对比数值也小于所述预设的对比阈值，重新提取人脸特征，将重新提取的人脸特征与多个所述原始图片分别进行对比，得到第三所需对比数值，基于所述第三所需对比数值大于等于所述预设的对比阈值的判断结果（S14），判定识别通过（S15）。

Description

人脸识别方法、人脸识别系统、及电子设备
本公开要求在2018年08月13日提交中国专利局、申请号为201810919221.7的中国专利申请的优先权，该申请的全部内容通过引用结合在本公开中。
技术领域
本公开涉及人脸识别技术领域，例如涉及一种人脸识别方法、人脸识别系统、及电子设备。
背景技术
人脸识别，是基于人的脸部特征信息进行身份识别的一种技术，通过提取人脸特征与数据库中已存有特征信息进行比对，获取比对结果，进而进行身份的识别。目前，由于人员年龄变化、妆容姿态变化等原因，导致人脸识别的准确率有待提高。
发明内容
有鉴于此，本公开提供一种人脸识别方法、电子设备、计算机可读存储介质及人脸识别系统。
本公开提供了一种人脸识别方法，所述方法包括：
采集人脸图片，提取人脸特征，将所述人脸特征与预存的多个原始图片分别进行对比，得到第一所需对比数值，将所述人脸特征与预存的多个现场采集图片分别进行对比，得到第二所需对比数值。
判断所述第一所需对比数值与所述第二所需对比数值是否大于等于预设的对比阈值。
基于所述第一所需对比数值小于所述预设的对比阈值且所述第二所需对比数值也小于所述预设的对比阈值的判断结果，重新提取人脸特征，将重新提取 的人脸特征与多个所述原始图片分别进行对比，得到第三所需对比数值，判断第三所需对比数值是否大于等于所述预设的对比阈值。
基于所述第三所需对比数值大于等于所述预设的对比阈值的判断结果，保存所述人脸图片作为新的现场采集图片，刷新所述现场采集图片对比成功的次数，并判定识别通过。
本公开还提供了一种电子设备，所述电子设备包括：
至少一个处理器；
存储器，设置为存储至少一个计算机程序，
当所述至少一个计算机程序被所述至少一个处理器执行，使得所述至少一个处理器实现如上所述的方法。
本公开还提供了一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行如上所述的方法。
本公开还提供了一种人脸识别系统，所述人脸识别系统包括图像获取设备和电子设备，所述图像获取设备与所述电子设备通信连接；
所述图像获取设备，用于获取人脸图片，并将所述人脸图片传递至所述电子设备。
所述电子设备为如上所述的电子设备。
附图说明
图1为本公开所提供的人脸识别系统的示意图。
图2为本公开所提供的人脸识别方法的一种流程示意图。
图3为本公开所提供的人脸识别方法的另一种流程示意图。
图4为本公开所提供的人脸识别方法的又一种流程示意图。
图标：10-人脸识别系统；11-图像获取设备；12-电子设备；121-存储器；122-图像处理模块。
具体实施方式
目前，在很多场景下都需要采用人脸识别进而对人的身份进行核实。通常，随着人的年龄变化及妆容姿态的变化，在通过采集人脸图片，将人脸图片与数据库中的原始图片进行对比，对人脸进行识别会出现漏匹配或误匹配的情况，人脸识别准确性不高。
本公开提供了一种人脸识别方法，所述方法包括：
采集人脸图片，提取人脸特征，将所述人脸特征与预存的多个原始图片分别进行对比，得到第一所需对比数值，将所述人脸特征与预存的多个现场采集图片分别进行对比，得到第二所需对比数值；
判断所述第一所需对比数值与所述第二所需对比数值是否大于等于预设的对比阈值；
基于所述第一所需对比数值小于所述预设的对比阈值且所述第二所需对比数值也小于所述预设的对比阈值的判断结果，重新提取人脸特征，将重新提取的人脸特征与多个所述原始图片分别进行对比，得到第三所需对比数值，判断第三所需对比数值是否大于等于所述预设的对比阈值；
基于所述第三所需对比数值大于等于所述预设的对比阈值的判断结果，保存所述人脸图片作为新的现场采集图片，刷新所述现场采集图片对比成功的次数，并判定识别通过。
在一实施例中，所述方法还包括：基于所述第三所需对比数值小于所述预设的对比阈值的判断结果，判定对比不成功，并判定识别不通过。
在一实施例中，所述方法还包括：
基于所述第一所需对比数值大于等于预设的对比阈值的判断结果，判断所述第一所需对比数值对应的原始图片和所述第二所需对比数值对应的现场采集图片是否为同一人员的图片；
基于所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片为同一人员的图片的判断结果，判定本次对比成功，并判定识别通过。
在一实施例中，所述方法还包括：
基于所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片非同一人员的图片的判断结果，判断所述第二所需对比数值是否大于等于所述预设的对比阈值；
基于所述第二所需对比数值小于所述预设的对比阈值的判断结果，判定对比成功，并判定识别通过。
在一实施例中，所述基于所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片非同一人员的图片的判断结果，判断所述第二所需对比数值是否大于等于所述预设的对比阈值的步骤之后，所述方法还包括：
基于所述第二所需对比数值大于等于所述预设的对比阈值的判断结果，判断所述第一所需对比数值对应的原始图片所对应的人员是否在预存的近似误匹配数据库中；
基于所述第一所需对比数值对应的原始图片所对应的人员处于所述近似误匹配数据库中的判断结果，判断所述第二所需对比数值对应的现场采集图片对比成功的次数是否为大于等于两次，基于所述第二所需对比数值对应的现场采集图片对比成功的次数大于等于两次的判断结果，判定本次比对成功，并判定识别通过。
在一实施例中，所述方法还包括：
基于所述第二所需对比数值对应的现场采集图片对比成功的次数小于两次的判断结果，重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片再次对比；
基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值大于等于所述预设的对比阈值的判断结果，判定比对成功，刷新所述第二所需对比数值对应的现场采集图片对比成功的次数，并判定识别通过。
在一实施例中，所述基于所述第二所需对比数值对应的现场采集图片对比成功的次数小于两次的判断结果，重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片再次对比的步骤之后，所述方法还包括：
基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值小于所述预设的对比阈值的判断结果，判定比对失败，清除所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值，并判定识别不通过。
在一实施例中，所述方法还包括：
基于所述第一所需对比数值对应的原始图片所对应的人员不在所述近似误匹配数据库中的判断结果，重新提取人脸特征，将重新提取的人脸特征与多个所述原始图片进行对比，得到第四所需对比数值；
基于所述第四所需对比数值大于等于所述预设的对比阈值的判断结果，判定对比成功，并判定识别通过。
在一实施例中，所述方法还包括：
基于所述第四所需对比数值小于所述预设的对比阈值的判断结果，判定对比不成功，并判定识别不通过。
在一实施例中，所述方法还包括：
基于所述第一所需对比数值小于所述预设的对比阈值，且所述第二所需对比数值大于等于所述预设的对比阈值的判断结果，判断所述第二所需对比数值对应的现场采集图片对比成功的次数是否为大于等于两次；
基于所述第二所需对比数值对应的现场采集图片对比成功的次数为大于等于两次的判断结果，判定本次比对成功，并判定识别通过。
在一实施例中，所述方法还包括：基于所述第二所需对比数值对应的现场采集图片对比成功的次数为小于两次的判断结果，重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片进行对比；
基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值大于等于所述预设的对比阈值，判定比对成功，刷新基于所述第二所需对比数值对应的现场采集图片对比成功的次数，并判定识别通过。
在一实施例中，所述方法还包括：
基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值小于所述预设的对比阈值的判断结果，判定比对失败，清除基于所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值，并判定识别不通过。
在一实施例中，所述采集人脸图片，提取人脸特征，将所述人脸特征与预存的多个原始图片分别进行对比，得到第一所需对比数值，将所述人脸特征与预存的多个现场采集图片分别进行对比，得到第二所需对比数值的步骤，包括：
检测是否存在待识别人脸，基于存在待识别人脸的判断结果，采集人脸图片，提取人脸特征，将所述人脸特征与预存的多个原始图片分别进行对比，得到第一所需对比数值，将所述人脸特征与预存的多个现场采集图片分别进行对比，得到第二所需对比数值。
请参阅图1，为本公开提供的一种人脸识别系统10的示意图，该人脸识别系统10可以应用于速通门，所述人脸识别系统10包括图像获取设备11和电子设备12，所述电子设备12包括存储器121和至少一个处理器，每个所述处理器具体可以包括图像处理模块122，其中，图像获取设备11与所述电子设备12通信连接，所述图像获取设备11用于获取人脸图片，并将获取的人脸图片传递至所述电子设备12。
所述存储器121存储有可在所述图像处理模块122上运行的计算机程序，所述图像处理模块122运行所述计算机程序时使得所述电子设备12执行下面的人脸识别方法。
请结合参阅图2，为本公开所提供的人脸识别方法的流程图，该方法可以应 用于图1所示的人脸识别系统10，由人脸识别系统10中的电子设备12执行。
该方法可以应用于电子设备12，所述电子设备12预存有多个人员的原始图片及现场采集图片，以及基于每个所述现场采集图片对比成功的次数。该方法可以包括：
步骤S10：检测有效识别距离内是否存在待识别人脸，若存在待识别人脸，则执行步骤S11至步骤S14。若不存在待识别人脸，则返回执行步骤S10。
步骤S11：采集人脸图片，提取人脸特征。
步骤S12：将所述人脸特征与多个所述原始图片分别进行对比，得到第一所需对比数值；将所述人脸特征与多个所述现场采集图片分别进行对比，得到第二所需对比数值。
其中，将所述人脸特征与多个所述原始图片对比，得到多个对比数值，在得到多个对比数值中，对比数值最高的则为第一所需对比数值。同理，将所述人脸特征与多个所述现场采集图片对比，得到多个对比数值，在得到多个对比数值中，对比数值最高的为第二所需对比数值。对比数值越高代表采集的人脸图片与原始图片及现场采集图片的相似度越高。
步骤S13：判断所述第一所需对比数值及所述第二所需对比数值是否大于等于预设的对比阈值，若所述第一所需对比数值小于所述预设的对比阈值且所述第二所需对比数值也小于所述预设的对比阈值，则执行步骤S14。
步骤S14：重新提取人脸特征，将重新提取的人脸特征与多个所述原始图片分别进行对比，得到第三所需对比数值，判断第三所需对比数值是否大于等于所述预设的对比阈值，若所述第三所需对比数值大于等于所述预设的对比阈值，则执行步骤S15，若所述第三所需对比数值小于所述预设的对比阈值，则执行步 骤S16。
步骤S15：将所述人脸图片作为新的现场采集图片保存至所述电子设备12中，刷新所述现场采集图片对比成功的次数，并判定识别通过。
其中，在所述第一所需对比数值小于所述预设的对比阈值，且所述第二所需对比数值也小于所述预设的对比阈值的情况下，重新提取人脸特征，再次进行人脸识别，重新提取的人脸特征与之前提取的人脸特征不同，重新提取的人脸特征相对与之前提取的人脸特征可以更为细化，重新提取的人脸特征的准确性可以更高。
在一实施例中，将所述采集的人脸图片作为新的现场采集图片保存至所述电子设备12时，所述新的现场采集图片没有完成入库存储，只有在下一次人脸识别中，再一次经过对比，并且对比成功识别通过的所述新的现场采集图片才会完成入库存储，这样经过多次的核验对比，使所述现场采集图片在人脸识别过程中更加具有适应性。
为了提高人脸识别的准确性，所述采集的人脸图片作为新的现场采集图片保存至所述电子设备12时，还可以判断采集的人脸图片是否符合预设条件，其中，对比成功且符合所述预设条件的人脸图片则作为新的现场采集图片保存至所述电子设备12中，若所述电子设备12中存储有现场采集图片，则作代替处理。
在具体实现中，预存于电子设备12的多个现场采集图片可以分别对应不同的人员，也即，电子设备12中存储有每个人员的至多一张图片。若所述电子设备12中存储有与所述新的现场采集图片为同一人员的一张现场采集图片，则可以将该张现场采集图片用所述新的现场采集图片替代，以实现代替处理。
所述预设条件包括：判断所述采集的人脸图片质量是否符合预设现场采集图片入库的阈值，和判断采集的人脸距离是否在预设的距离范围内中至少之一。采集的人脸距离检测可通过采集人脸大小或者几何空间模型预测人脸距离。
在一实施例中，判断是否符合所述预设条件的过程可以包括：若所述采集的人脸图片质量符合所述现场采集图片入库的阈值且所述采集的人脸距离在预设的距离范围内，则比对成功，将所述采集的人脸图片作为新的现场采集图片保存至所述电子设备12中，刷新基于所述新的现场采集图片对比成功的次数，并判定识别通过；在一实施例中，判断是否符合所述预设条件的过程可以包括：若所述采集的人脸图片质量符合所述现场采集图片入库的阈值，则比对成功，判定识别通过；在一实施例中，判断是否符合所述预设条件的过程可以包括：若所述采集的人脸距离在预设的距离范围内，比对成功，判断识别通过；在一实施例中，判断是否符合所述预设条件的过程可以包括：若所述采集的人脸图片质量不符合所述现场采集图片入库的阈值以及所述采集的人脸距离不在预设的距离范围内，则比对失败，判定识别不通过。
需要说明的是，在本实施例的步骤S15中，所述刷新所述现场采集图片对比成功的次数，具体为：刷新所述新的现场采集图片对比成功的次数。
在其他实施例中，所述刷新所述现场采集图片对比成功的次数，具体可以为：刷新预存的多个现场采集图片中的目标现场采集图片的对比成功的次数。其中，所述目标现场采集图片可以为所述第二所需对比数值对应的现场采集图片；所述目标现场采集图片可以为能够与所述新的现场采集图片对比成功的一个现场采集图片；所述目标现场采集图片可以与所述新的现场采集图片为同一人员的图片。
步骤S16：对比不成功，识别不通过。
请结合参阅图3，所述电子设备12可以存储有近似误匹配数据库，所述近似误匹配数据库存储有相似人员名单。若所述第一所需对比数值大于等于预设的对比阈值，则执行步骤S21。
步骤S21：判断所述第一所需对比数值对应的原始图片和所述第二所需对比数值对应的现场采集图片是否为同一人员的图片，如果所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片为同一人员的图片，执行步骤S22；如果所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片非同一人员的图片，执行步骤S23。
步骤S22：本次对比成功，判定识别通过。
其中，判定识别通过后，将所述第一所需对比数值作为本次判定通过的记录，存储至所述电子设备12中。
步骤S23：判断所述第二所需对比数值是否大于等于所述预设的对比阈值；如果所述第二所需对比数值小于所述预设的对比阈值，执行步骤S24，如果所述第二所需对比数值大于等于所述预设的对比阈值，执行步骤S25。
步骤S24：对比成功，判定识别通过。
其中，判定识别通过后，将所述第一所需对比数值作为本次判定通过的记录，存储至所述电子设备12中。
步骤S25：判断所述第一所需对比数值对应的原始图片所对应的人员是否在所述近似误匹配数据库中，如果所述第一所需对比数值对应的原始图片所对应 的人员处于所述近似误匹配数据库中，执行步骤S26，如果所述第一所需对比数值对应的原始图片所对应的人员不处于所述近似误匹配数据库，执行步骤S27-步骤S28。
步骤S26：判断所述第二所需对比数值对应的现场采集图片对比成功的次数是否为两次以上(即大于等于两次)，若所述第二所需对比数值对应的现场采集图片对比成功的次数为两次以上，执行步骤S261；若所述第二所需对比数值对应的现场采集图片对比成功的次数为一次以下(即小于两次)，执行步骤S262。
步骤S261：本次比对成功，判定识别通过。
其中，判定识别通过后，将第二所需对比数值作为本次判定通过的记录，存储至电子设备12中。
步骤S262：重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片再次对比，判断重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值是否大于等于所述预设的对比阈值，若重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值大于等于所述预设的对比阈值，执行步骤S2621，若重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值小于所述预设的对比阈值，执行步骤S2622。
步骤S2621：比对成功，刷新所述第二所需对比数值对应的现场采集图片对比成功的次数，并判定识别通过。
其中，判断识别通过后，将本次对比数值作为本次判定通过的记录，将刷新后基于所述第二所需对比数值对应的现场采集图片对比成功的次数以及判定通过的记录存储至电子设备12中。
步骤S2622：比对失败，清除基于所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值，并判定识别不通过。
如图3所示的实施例中，所述方法还包括，步骤S27：重新提取人脸特征，将重新提取的人脸特征与多个所述原始图片进行对比，得到第四所需对比数值；
步骤S28：判断第四所需对比数值是否大于等于所述预设的对比阈值；若所述第四所需对比数值小于所述预设的对比阈值，执行步骤S281。若所述第四所需对比数值大于等于所述预设的对比阈值，执行步骤S282。
步骤S281：对比不成功，清除基于所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值，并判定识别不通过。
步骤S282：对比成功，判定识别通过。
其中，若第四所需对比数值大于等于所述预设的对比阈值，对比成功，判定识别通过后，若第四所需对比数值对应的原始图片与第一所需对比数值对应的原始图片为同一人的图片，则将第一所需对比数值作为本次判断识别通过的记录，存储至服务器，其中，清除基于所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值；若第四所需对比数值对应的原始图片与第二所需对比数值对应的现场采集图片为同一人的图片，将第一所需对比数值对应原始图片所对应的人员名单及第二所需对比数值对应现场采集图片所对应的人员名单存储至所述近似误匹配数据库，将第二所需对比数值作为本次判断识别通过的记录，存储至所述电子设备12中；若第四所需对比数值对应的原始图片与第一所需对比数值对应的原始图片及第二所需对比数值对应的现场采集图片非同一人的图片，将第四所需对比数值作为本次判定识别通过的记录，存储至所述电子设备12中，清除基于所述第二所需对比数值对应的 现场采集图片对比成功的次数，删除所述第二所需对比数值。
请结合参阅图4，如果所述第一所需对比数值小于所述预设的对比阈值，且所述第二所需对比数值大于等于所述预设的对比阈值，则执行步骤S31。
步骤S31：判断基于所述第二所需对比数值对应的现场采集图片对比成功的次数是否为两次以上；若基于所述第二所需对比数值对应的现场采集图片对比成功的次数为两次以上(即大于等于两次)，执行步骤S32；若基于所述第二所需对比数值对应的现场采集图片对比成功的次数为一次以下(即小于两次)，执行步骤S33。
步骤S32：本次比对成功，判定识别通过。
判定识别通过后，将第二所需对比数值作为本次判定通过的记录，存储至所述电子设备12中。
步骤S33：重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片进行对比，判断重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值是否大于等于所述预设的对比阈值，若重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值大于等于所述预设的对比阈值，执行步骤S331，若重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值小于所述预设的对比阈值，执行步骤S332。
步骤S331：比对成功，刷新基于所述第二所需对比数值对应的现场采集图片对比成功的次数，并判定识别通过。
其中，判断识别通过后，将本次对比数值作为本次判定通过的记录，将刷新后基于所述第二所需对比数值对应的现场采集图片对比成功的次数以及判定 通过的记录存储至所述电子设备12中。
步骤S332：比对失败，清除基于所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值，并判定识别不通过。
上述过程中，比对成功，刷新基于所述第二所需对比数值对应的现场采集图片对比成功的次数，在原先对比成功的次数上加上此次对比成功的次数，例如，若原先对比成功的次数为1次，刷新过后，则在1次的基础上加上此次的对比次数，进而，对比成功的次数变为两次。
比对失败，清除基于所述第二所需对比数值对应的现场采集图片对比成功的次数，例如现场采集图片对比成功的次数为1次，对比失败后，则将1次变为零次，即初始化状态，并将提取的人脸与所述现场采集图片对比得到的对比数值(也即第二所需对比数值)删除，判定识别不通过。
为了保证人脸识别的准确性，所述电子设备12中所存储的现场采集图片会不断更新，其中，更新方式可以为，在人脸识别过程中，将比对成功，且符合预设条件的人脸采集图片作为新的现场采集图片以代替已有现场采集图片存储至电子设备12中，也可以为定期在人脸识别过程中，采集符合预设条件的人脸图片，将人脸图片作为新的现场采集图片存储至电子设备12中，以更新电子设备12中的现场采集图片。而所述电子设备12中所存储的原始图片则可通过定期录入新的图片，以更新所述电子设备12中的原始图片。
本公开提供一种人脸识别系统10，可以应用于速通门，所述人脸识别系统10包括图像获取设备11和电子设备12，所述图像获取设备11与所述电子设备12通信连接。
所述图像获取设备11，配置为获取人脸图片，并将所述人脸图片传递至所 述电子设备12。
所述电子设备12包括存储器121、处理器(所述处理器可以为图像处理模块122)以及存储在所述存储器121上并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时使得所述人脸识别系统10中的所述电子设备12实现上述的人脸识别方法。
在一实施例中，所述存储器121包括人员数据库，所述人员数据库存储有多个人员的原始图片及现场采集图片。
所述图像处理模块122配置为提取所述人脸图片中的人脸特征，并将提取的所述人脸特征与多个所述原始图片及所述现场采集图片分别进行对比，对所述人脸图片进行识别。
例如，当本公开提供的所述人脸识别系统10应用于所述速通门时，所述电子设备12可以为终端设备，所述终端设备设置于所述速通门上，实时对经过的人进行检测，对要经过速通门的行人，采集人脸图片，提取人脸特征，进行人脸识别。当人脸识别成功时，终端设备控制闸机开门，行人则可通过闸机，当行人通过后，闸机延时一段时间自动关门，继续对下一个行人进行检测。当所述人脸识别系统10应用于速通门时，行人若距离闸机过远时开门，会存在安全风险，比如其他行人抢先进入闸机等，因此，所述终端设备还需要对人脸的距离进行检测，仅在人脸在合适距离范围内时，才执行开闸放行，可选的，本公开将距离范围设置为0.3-1.5m。在一实施例中，当所述人脸识别系统10应用于速通门时，所述终端设备会定义两个距离模式，包括快速模式和安全模式，其中，快速模式距离阈值大于安全模式距离阈值，当启用快速模式时，检测到人脸距离小于等于快速模式距离阀值，启动人脸识别比对核验处理；当启用安全 模式时，当检测到人脸距离小于等于安全模式距离阀值，启动人脸识别比对核验处理。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的人脸识别系统10的具体工作过程，可以参考前述方法中的对应过程，在此不再过多赘述。
综上所述，本公开所提供的一种人脸识别方法及人脸识别系统，在电子设备中存储有多个人员的原始图片及多个人员的现场采集图片，通过采集人脸图片，提取人脸特征，将提取的人脸特征分别与多个原始图片及现场采集图片进行对比，进而对人脸进行识别，在两者对比都失败的情况下，则重新提取更为细化的人脸特征，与电子设备中的多个原始图片再次进行对比，进而对人脸进行再次识别，确保了人脸识别过程无漏匹配或错匹配的情况，有效保证人脸的准确性。同时，本公开在将人脸图片作为新的现场采集图片重新入库存储时，需要对新的现场采集图片进行多次对比核验，保证新的现场采集图片具有强适应性，利于提高人脸识别的准确性。
本公开实施例还提供一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行上述方法。
上述实施例方法中的全部或部分流程可以通过计算机程序来执行相关的硬件来完成的，该程序可存储于一个非暂态计算机可读存储介质中，该程序在执行时，可包括如上述方法的实施例的流程，其中，该非暂态计算机可读存储介质可以为磁碟、光盘、只读存储记忆体(Read‐Only Memory，ROM)或RAM等。

Claims (19)

  1. 一种人脸识别方法，所述方法包括：
    采集人脸图片，提取人脸特征，将所述人脸特征与预存的多个原始图片分别进行对比，得到第一所需对比数值，将所述人脸特征与预存的多个现场采集图片分别进行对比，得到第二所需对比数值；
    判断所述第一所需对比数值与所述第二所需对比数值是否大于等于预设的对比阈值；
    基于所述第一所需对比数值小于所述预设的对比阈值且所述第二所需对比数值也小于所述预设的对比阈值的判断结果，重新提取人脸特征，将重新提取的人脸特征与多个所述原始图片分别进行对比，得到第三所需对比数值，判断第三所需对比数值是否大于等于所述预设的对比阈值；
    基于所述第三所需对比数值大于等于所述预设的对比阈值的判断结果，保存所述人脸图片作为新的现场采集图片，刷新所述现场采集图片对比成功的次数，并判定识别通过。
  2. 根据权利要求1所述的人脸识别方法，其中，所述方法还包括：基于所述第三所需对比数值小于所述预设的对比阈值的判断结果，判定对比不成功，并判定识别不通过。
  3. 根据权利要求1所述的人脸识别方法，其中，所述方法还包括：
    基于所述第一所需对比数值大于等于预设的对比阈值的判断结果，判断所述第一所需对比数值对应的原始图片和所述第二所需对比数值对应的现场采集图片是否为同一人员的图片；
    基于所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片为同一人员的图片的判断结果，判定本次对比成功，并判定识别通过。
  4. 根据权利要求1所述的人脸识别方法，其中，所述方法还包括：
    基于所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片非同一人员的图片的判断结果，判断所述第二所需对比数值是否大于等于所述预设的对比阈值；
    基于所述第二所需对比数值小于所述预设的对比阈值的判断结果，判定对比成功，并判定识别通过。
  5. 根据权利要求4所述的人脸识别方法，其中，所述基于所述第一所需对比数值大于等于所述预设的对比阈值，且所述第一所需对比数值对应的原始图片及所述第二所需对比数值对应的现场采集图片非同一人员的图片的判断结果，判断所述第二所需对比数值是否大于等于所述预设的对比阈值的步骤之后，所述方法还包括：
    基于所述第二所需对比数值大于等于所述预设的对比阈值的判断结果，判断所述第一所需对比数值对应的原始图片所对应的人员是否在预存的近似误匹配数据库中；
    基于所述第一所需对比数值对应的原始图片所对应的人员处于所述近似误匹配数据库中的判断结果，判断所述第二所需对比数值对应的现场采集图片对比成功的次数是否为大于等于两次，基于所述第二所需对比数值对应的现场采集图片对比成功的次数大于等于两次的判断结果，判定本次比对成功，并判定 识别通过。
  6. 根据权利要求5所述人脸识别方法，其中，所述方法还包括：
    基于所述第二所需对比数值对应的现场采集图片对比成功的次数小于两次的判断结果，重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片再次对比；
    基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值大于等于所述预设的对比阈值的判断结果，判定比对成功，刷新所述第二所需对比数值对应的现场采集图片对比成功的次数，并判定识别通过。
  7. 根据权利要求6所述人脸识别方法，其中，所述基于所述第二所需对比数值对应的现场采集图片对比成功的次数小于两次的判断结果，重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片再次对比的步骤之后，所述方法还包括：
    基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值小于所述预设的对比阈值的判断结果，判定比对失败，清除所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值，并判定识别不通过。
  8. 根据权利要求5所述的人脸识别方法，其中，所述方法还包括：
    基于所述第一所需对比数值对应的原始图片所对应的人员不在所述近似误匹配数据库中的判断结果，重新提取人脸特征，将重新提取的人脸特征与多个所述原始图片进行对比，得到第四所需对比数值；
    基于所述第四所需对比数值大于等于所述预设的对比阈值的判断结果，判 定对比成功，并判定识别通过。
  9. 根据权利要求8所述的人脸识别方法，其中，所述方法还包括：
    基于所述第四所需对比数值小于所述预设的对比阈值的判断结果，判定对比不成功，并判定识别不通过。
  10. 根据权利要求1所述的人脸识别方法，其中，所述方法还包括：
    基于所述第一所需对比数值小于所述预设的对比阈值，且所述第二所需对比数值大于等于所述预设的对比阈值的判断结果，判断所述第二所需对比数值对应的现场采集图片对比成功的次数是否为大于等于两次；
    基于所述第二所需对比数值对应的现场采集图片对比成功的次数为大于等于两次的判断结果，判定本次比对成功，并判定识别通过。
  11. 根据权利要求10所述的人脸识别方法，其中，所述方法还包括：基于所述第二所需对比数值对应的现场采集图片对比成功的次数为小于两次的判断结果，重新提取人脸特征，将重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片进行对比；
    基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值大于等于所述预设的对比阈值，判定比对成功，刷新基于所述第二所需对比数值对应的现场采集图片对比成功的次数，并判定识别通过。
  12. 根据权利要求11所述的人脸识别方法，其中，所述方法还包括：
    基于重新提取的人脸特征与所述第二所需对比数值对应的现场采集图片对比得到的对比数值小于所述预设的对比阈值的判断结果，判定比对失败，清除基于所述第二所需对比数值对应的现场采集图片对比成功的次数，删除所述第二所需对比数值，并判定识别不通过。
  13. 根据权利要求1所述的人脸识别方法，其中，所述采集人脸图片，提取人脸特征，将所述人脸特征与预存的多个原始图片分别进行对比，得到第一所需对比数值，将所述人脸特征与预存的多个现场采集图片分别进行对比，得到第二所需对比数值的步骤，包括：
    检测是否存在待识别人脸，基于存在待识别人脸的判断结果，采集人脸图片，提取人脸特征，将所述人脸特征与预存的多个原始图片分别进行对比，得到第一所需对比数值，将所述人脸特征与预存的多个现场采集图片分别进行对比，得到第二所需对比数值。
  14. 一种电子设备，所述电子设备包括：
    至少一个处理器；
    存储器，设置为存储至少一个计算机程序，
    当所述至少一个计算机程序被所述至少一个处理器执行时，使得所述至少一个处理器实现如权利要求1‐13中任一所述的方法。
  15. 如权利要求14所述的电子设备，其中，所述存储器包括人员数据库，所述人员数据库存储有多个人员的原始图片及多个人员的现场采集图片。
  16. 一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行如权利要求1‐13任一所述的方法。
  17. 一种人脸识别系统，其中，所述人脸识别系统包括图像获取设备和电子设备，所述图像获取设备与所述电子设备通信连接；
    所述图像获取设备，所述图像获取设备配置为获取人脸图片，并将所述人脸图片传递至所述电子设备；
    所述电子设备为如权利要求14所述的电子设备。
  18. 根据权利要求17所述的人脸识别系统，其中，
    所述存储器包括人员数据库，所述人员数据库存储有多个人员的原始图片 及多个人员的现场采集图片；
    所述图像处理器配置为提取所述人脸图片中的人脸特征，并将提取的所述人脸特征与多个所述原始图片及所述现场采集图片分别进行对比，对其进行识别。
  19. 根据权利要求17或18所述的人脸识别系统，其中，所述人脸识别系统应用于速通门。
PCT/CN2019/081186 2018-08-13 2019-04-03 人脸识别方法、人脸识别系统、及电子设备 WO2020034645A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP19849095.5A EP3839806B1 (en) 2018-08-13 2019-04-03 Facial recognition method, facial recognition system, and electronic device
US17/268,127 US11967176B2 (en) 2018-08-13 2019-04-03 Facial recognition method, facial recognition system, and electronic device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810919221.7A CN109145801B (zh) 2018-08-13 2018-08-13 一种人脸识别方法及人脸识别系统
CN201810919221.7 2018-08-13

Publications (1)

Publication Number Publication Date
WO2020034645A1 true WO2020034645A1 (zh) 2020-02-20

Family

ID=64793063

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/081186 WO2020034645A1 (zh) 2018-08-13 2019-04-03 人脸识别方法、人脸识别系统、及电子设备

Country Status (4)

Country Link
US (1) US11967176B2 (zh)
EP (1) EP3839806B1 (zh)
CN (1) CN109145801B (zh)
WO (1) WO2020034645A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3852076A1 (en) * 2020-01-15 2021-07-21 Climax Technology Co., Ltd. Smart home security system and method of disarming a security setting

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109145801B (zh) * 2018-08-13 2021-02-02 浙江宇视科技有限公司 一种人脸识别方法及人脸识别系统
CN110633677B (zh) * 2019-09-18 2023-05-26 威盛电子股份有限公司 人脸识别的方法及装置
CN111260838A (zh) * 2020-02-27 2020-06-09 广州羊城通有限公司 一种基于本地识别的刷脸通行方法及闸机设备
CN112818784B (zh) * 2021-01-22 2024-02-06 浙江大华技术股份有限公司 一种门禁设备的控制方法、装置及存储介质
CN118285946B (zh) * 2024-06-03 2024-10-01 先临三维科技股份有限公司 扫描杆管理方法、装置、存储介质及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080317298A1 (en) * 2005-09-28 2008-12-25 Facedouble Incorporated Digital Image Search System And Method
CN103400108A (zh) * 2013-07-10 2013-11-20 北京小米科技有限责任公司 人脸识别方法、装置和移动终端
CN105574500A (zh) * 2015-12-15 2016-05-11 北京天诚盛业科技有限公司 提高人脸识别通过率的方法和装置
CN106446816A (zh) * 2016-09-14 2017-02-22 北京旷视科技有限公司 人脸识别方法及装置
CN109145801A (zh) * 2018-08-13 2019-01-04 浙江宇视科技有限公司 一种人脸识别方法及人脸识别系统

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2825635A1 (en) * 2012-08-28 2014-02-28 Solink Corporation Transaction verification system
KR101498670B1 (ko) * 2013-10-07 2015-03-05 이화여자대학교 산학협력단 생체정보 기반 신뢰구간집합을 이용한 패스워드 생성 방법 및 장치
CN204155293U (zh) * 2014-09-01 2015-02-11 上海智达商投资管理合伙企业(有限合伙) 一种基于人脸识别的验证装置及验证系统
KR102010378B1 (ko) * 2014-09-24 2019-08-13 삼성전자주식회사 객체를 포함하는 영상의 특징을 추출하는 방법 및 장치
CN105550671A (zh) 2016-01-28 2016-05-04 北京麦芯科技有限公司 一种人脸识别的方法及装置
CN107437048A (zh) * 2016-05-27 2017-12-05 鸿富锦精密工业（深圳）有限公司 人脸识别系统及人脸识别方法
WO2019064743A1 (ja) * 2017-09-28 2019-04-04 パナソニックＩｐマネジメント株式会社 認証装置、認証システム、認証方法、及びプログラム
CN107818308A (zh) * 2017-10-31 2018-03-20 平安科技（深圳）有限公司 一种人脸识别智能比对方法、电子装置及计算机可读存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080317298A1 (en) * 2005-09-28 2008-12-25 Facedouble Incorporated Digital Image Search System And Method
CN103400108A (zh) * 2013-07-10 2013-11-20 北京小米科技有限责任公司 人脸识别方法、装置和移动终端
CN105574500A (zh) * 2015-12-15 2016-05-11 北京天诚盛业科技有限公司 提高人脸识别通过率的方法和装置
CN106446816A (zh) * 2016-09-14 2017-02-22 北京旷视科技有限公司 人脸识别方法及装置
CN109145801A (zh) * 2018-08-13 2019-01-04 浙江宇视科技有限公司 一种人脸识别方法及人脸识别系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3839806A4

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3852076A1 (en) * 2020-01-15 2021-07-21 Climax Technology Co., Ltd. Smart home security system and method of disarming a security setting

Also Published As

Publication number Publication date
CN109145801A (zh) 2019-01-04
EP3839806A4 (en) 2022-06-08
EP3839806B1 (en) 2024-09-04
CN109145801B (zh) 2021-02-02
EP3839806A1 (en) 2021-06-23
US11967176B2 (en) 2024-04-23
US20210326622A1 (en) 2021-10-21

Similar Documents

Publication Publication Date Title
WO2020034645A1 (zh) 人脸识别方法、人脸识别系统、及电子设备
CN106991395B (zh) 信息处理方法、装置及电子设备
CN108038176B (zh) 一种路人库的建立方法、装置、电子设备及介质
KR101781358B1 (ko) 디지털 영상 내의 얼굴 인식을 통한 개인 식별 시스템 및 방법
KR20070119463A (ko) 얼굴 인증 장치, 얼굴 인증 방법, 및 출입 관리 장치
US20230368559A1 (en) Information processing apparatus, information processing method, and non-transitory computer-readable storage medium
US10970953B2 (en) Face authentication based smart access control system
KR20170123950A (ko) 얼굴 인식을 이용한 출입 관리 방법 및 시스템
US11978282B2 (en) Authentication system, authentication method, and storage medium
AU2024201525B2 (en) Gate system, gate apparatus, image processing method therefor, program, and arrangement method for gate apparatus
JP6558178B2 (ja) 迷惑行為者推定システム、迷惑行為者推定システムの制御方法及び制御プログラム
CN110287841B (zh) 图像传输方法及装置、图像传输系统、存储介质
CN111444757A (zh) 无人超市的行人重识别方法、装置、设备及存储介质
JP7302329B2 (ja) 画像処理装置、人物判別方法、およびプログラム
KR101718244B1 (ko) 얼굴 인식을 위한 광각 영상 처리 장치 및 방법
CN114926717A (zh) 对象特征融合方法、装置以及存储介质
Lee et al. Ocular and iris recognition baseline algorithm
CN116012909A (zh) 一种快速人脸识别方法、系统
CN118470777A (zh) 一种无感通行人脸比对方法及装置
CN111967314A (zh) 行人重识别方法、装置、电子设备及存储介质
CN112995582A (zh) 监控设备的控制方法、装置、及服务器

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: 19849095

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2019849095

Country of ref document: EP

Effective date: 20210315