CN110503031B - Method for improving face recognition accuracy and passing speed - Google Patents
Method for improving face recognition accuracy and passing speed Download PDFInfo
- Publication number
- CN110503031B CN110503031B CN201910774446.2A CN201910774446A CN110503031B CN 110503031 B CN110503031 B CN 110503031B CN 201910774446 A CN201910774446 A CN 201910774446A CN 110503031 B CN110503031 B CN 110503031B
- Authority
- CN
- China
- Prior art keywords
- user
- replacement
- photos
- score
- photo
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 8
- 101100153581 Bacillus anthracis topX gene Proteins 0.000 claims abstract description 13
- 101150041570 TOP1 gene Proteins 0.000 claims abstract description 13
- 238000012795 verification Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 8
- 238000005286 illumination Methods 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
Images
Classifications
-
- 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
- G06V40/161—Detection; Localisation; Normalisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (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)
- Collating Specific Patterns (AREA)
Abstract
The invention discloses a method for improving face recognition accuracy and passing speed, which comprises the following steps: the maximum comparison score obtained by the user in the registration of all the people in the bottom library in the recognition model A, namely the top1 score meets the following conditions: score _ top1 is more than or equal to Th, and Th is a preset identification threshold value and represents that the user identification passes; if the Score _ top1 is less than or equal to Th1, Th1 is greater than Th, Th1 is set by the user, which indicates that the user recognizes that the similarity of the user's base picture and the current real person is greatly different, and the live picture in the time period is added or updated to the face library of the user as a replacement picture.
Description
Technical Field
The invention relates to the field of face recognition, in particular to a method for improving face recognition accuracy and passing speed.
Background
The face recognition is influenced by different factors including light, registration photo, camera installation height and the like in an actual use scene. These uncontrolled factors can affect the speed and accuracy of face recognition traffic. In practical use, the interference factors generally reduce the face recognition accuracy, and cause the problem of difficulty in recognition of the detected person.
The face recognition verification score is not full every time, which is caused by the difference between the actual use environment and the style of the registered photo (such as the registered photo is p-chart, beautified, and the like), the difference between the age and the light of the registered photo of the tested person and the appearance of the equipment, the difference between the brightness and the darkness, even the influence of other color light sources on the facial tone), the variation of the micro expression, the deviation of the face angle, and the like. In actual use, due to the interference factors, a person may need to recognize for passing through for multiple times, so that the recognition accuracy and the passing speed are greatly reduced. In addition, in order to obtain a recognition result quickly on an embedded device, the size and the feature dimension of a recognition model trained by the neural network are limited.
Disclosure of Invention
In order to solve the technical problems in the prior art, the invention provides a method for improving the accuracy rate of face recognition and the passing speed, which comprises the following steps:
the maximum comparison score obtained by the user in the registration of all the people in the bottom library in the recognition model A, namely the top1 score meets the following conditions: score _ top1 is more than or equal to Th, and Th is a preset identification threshold value and represents that the user identification passes;
if the Score _ top1 is less than or equal to Th1, Th1 is greater than Th, Th1 is set by the user, which indicates that the user recognizes that the similarity of the user's base picture and the current real person is greatly different, and the live picture in the time period is added or updated to the face library of the user as a replacement picture.
Further, the field photo can meet the following conditions as a replacement photo:
1. the field illumination meets the human face quality check, including human face size, angle, ambiguity and illumination;
2. and (3) performing secondary verification by using an identification model B which has higher accuracy and larger scale than the identification model A, comparing all the registration photos and the replacement photos in the non-period with the field photos respectively, wherein the maximum comparison score is not less than Th2, the minimum comparison score is not less than Th3, and Th2 and Th3 are set by the user and are used for restricting the confidence of whether the replacement photos take effect or not.
Further, whether a larger identification model B is started or not is determined by monitoring the CPU usage, namely when the service is busy, only the candidate replacement is stored, and when the service is idle, the identification model B is started for secondary verification.
Further, the constraint of whether to replace is as follows:
1. if the time interval replacement photo does not exist, the field photo is additionally registered in the user face library as the replacement photo;
2. if the replacement photographs of the time period exist before, the scores of the similarity scores of the new and old replacement photographs of the time period, all the registration photographs and the registration photographs not in the time period are mapped through the mapping function are compared, and the higher score is obtained to serve as the replacement photograph of the time period.
Further, the mapping function is as follows:
F(Scores)= (1-(1-s1)*(1-s2)*...*(1-sn))
wherein Scors is s1, s2, the combination of Scores of sn, n represents the number of registered photos and the number of registered photos not in the period, and s1 to sn represent the similarity of the new and old replacement photos and the n compared photos respectively.
The invention can reduce the interference of external interference factors, improve the traffic rate and reduce the face recognition verification time under the condition of ensuring and even improving the recognition accuracy rate.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The invention will be further explained with reference to the drawings.
As shown in fig. 1, when the user uses the face recognition technology to verify the identity, ideally, the maximum comparison score obtained in the recognition model a, i.e. the top1 score, of the registered photos of the user and all the people in the base repository satisfies: score _ top1 is more than or equal to Th, Th is a preset identification threshold value, which represents that the user passes the identification and simultaneously indicates that the identity of the user is the identity of the person corresponding to the registration photo with the top1 Score. If at this time, Score _ top1 is less than or equal to Th1, Th1 is greater than Th, and Th1 is set by the user, which indicates that the user passes the recognition, and the similarity between the photos of the user in the base library and the current real person is not particularly high, at this time, the live photos added or updated in the time period are added as the replacement photos to the user face library, so that at the next comparison, the user replacement photos exist in the base library, and the probability of successful recognition of the user is increased.
The field photo can meet the following conditions as a replacement photo:
1. the field illumination needs to meet the requirements of human face quality verification, including conditions of human face size, angle, ambiguity, illumination and the like.
2. And performing secondary verification by using a higher-accuracy and larger identification model B, and comparing all the registration photos and the replacement photos not in the time period of the user with the identification scene photos respectively, wherein the replacement photos not in the time period refer to that the field identification photos are additionally registered in the user face library by using the method, and each user to be detected can generate the replacement photos in different time periods, and can be distinguished according to morning, noon and evening. If the maximum comparison score is larger than or equal to Th2 and the minimum comparison score is larger than or equal to Th3, the identification scene photo meets the replacement requirement, and Th2 and Th3 are set by the user to restrict the confidence of whether the replacement photo is effective.
Because the identification model B consumes more computing power, the equipment determines whether to start the larger identification model B by monitoring the CPU usage, namely when the service is busy (the CPU consumption is higher), only the candidate replacement is stored, and when the service is idle (the CPU consumption is lower), the identification model B is started to perform secondary verification.
Constraint of whether to replace: 1, if the time interval replacement photo does not exist, additionally registering the field photo as a replacement photo in the user face library; 2, if the replacement photograph of the period exists before, comparing the similarity scores of the new and old replacement photographs of the period with all the registration photographs and the registration photographs not in the period after mapping by the mapping function, and taking the higher score as the replacement photograph of the period.
The mapping function is as follows:
f (Scores) = (1- (1-s1) × (1-s2) · × 1-sn)), Scores are fractional combinations of s1, s 2.., sn, n represents the number of registered photos and the number of registered photos not in the period, and s1 to sn represent the similarity of the new and old replacement photos to the n compared photos, respectively.
According to the invention, through a registration photo replacement mechanism, the equipment has the capability of 'semi-self learning', the influence of interference factors such as non-uniform registration photo style, age difference, light, micro-expression change, deviation of face angle and the like on the face identification accuracy and the passing speed is well solved, and the identification accuracy and the passing speed can be increased.
Claims (1)
1. A method for improving face recognition accuracy and passing speed comprises the following steps:
the maximum comparison score obtained by the user in the registration of all the people in the bottom library in the recognition model A, namely the top1 score meets the following conditions: score _ top1 is more than or equal to Th, and Th is a preset identification threshold value and represents that the user identification passes;
if the Score _ top1 is less than or equal to Th1, Th1 is greater than Th, Th1 is set by the user, which indicates that the user recognizes that the similarity difference between the user's image in the base library and the current real person is large, and the live photo in the time period is added or updated to be added as a replacement photo to the face library of the user;
the field photo can meet the following conditions as a replacement photo:
(1) the field illumination meets the human face quality check, including human face size, angle, ambiguity and illumination;
(2) performing secondary verification by using an identification model B which has higher accuracy and larger scale than the identification model A, respectively comparing all the registration photos and the replacement photos in the non-time period of the user with the site photos, wherein the maximum comparison score is not less than Th2, the minimum comparison score is not less than Th3, and Th2 and Th3 are set by the user and are used for restricting the confidence of whether the replacement photos take effect;
whether a larger identification model B is started or not is determined by monitoring the CPU usage, namely when the service is busy, only the candidate replacement image is stored, and when the service is idle, the identification model B is started for secondary verification;
the constraints of whether to replace are as follows:
(1) if the time interval replacement photo does not exist, the field photo is additionally registered in the user face library as the replacement photo;
(2) if the replacement photographs of the time period exist before, the scores of the similarity scores of the new and old replacement photographs of the time period and all the registration photographs and the registration photographs not in the time period after mapping through a mapping function are compared, and the higher score is obtained to be used as the replacement photographs of the time period;
the mapping function is as follows:
F(Scores)= (1-(1-s1)*(1-s2)*...*(1-sn))
wherein Scors is s1, s2, the combination of Scores of sn, n represents the number of registered photos and the number of registered photos not in the period, and s1 to sn represent the similarity of the new and old replacement photos and the n compared photos respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910774446.2A CN110503031B (en) | 2019-08-21 | 2019-08-21 | Method for improving face recognition accuracy and passing speed |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910774446.2A CN110503031B (en) | 2019-08-21 | 2019-08-21 | Method for improving face recognition accuracy and passing speed |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110503031A CN110503031A (en) | 2019-11-26 |
| CN110503031B true CN110503031B (en) | 2021-11-26 |
Family
ID=68589086
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910774446.2A Active CN110503031B (en) | 2019-08-21 | 2019-08-21 | Method for improving face recognition accuracy and passing speed |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110503031B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822245B (en) * | 2021-11-22 | 2022-03-04 | 杭州魔点科技有限公司 | Face recognition method, electronic device, and medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056075A (en) * | 2016-05-27 | 2016-10-26 | 广东亿迅科技有限公司 | Important person identification and tracking system in community meshing based on unmanned aerial vehicle |
| CN105938552B (en) * | 2016-06-29 | 2020-04-24 | 北京旷视科技有限公司 | Face recognition method and device for automatically updating base map |
| CN107657202A (en) * | 2016-07-25 | 2018-02-02 | 鸿富锦精密工业（深圳）有限公司 | Face identification method |
| CN108229260B (en) * | 2016-12-21 | 2020-12-29 | 杭州海康威视系统技术有限公司 | Identity information verification method and system |
| CN107818301B (en) * | 2017-10-16 | 2021-04-02 | 创新先进技术有限公司 | Method and device for updating biological characteristic template and electronic equipment |
| CN107992855A (en) * | 2017-12-22 | 2018-05-04 | 中国科学院重庆绿色智能技术研究院 | A kind of triple verification methods of airport security based on recognition of face |
-
2019
- 2019-08-21 CN CN201910774446.2A patent/CN110503031B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110503031A (en) | 2019-11-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020038136A1 (en) | Facial recognition method and apparatus, electronic device and computer-readable medium | |
| JP5899472B2 (en) | Person attribute estimation system and learning data generation apparatus | |
| KR102174595B1 (en) | System and method for identifying faces in unconstrained media | |
| WO2020098074A1 (en) | Face sample picture marking method and apparatus, computer device, and storage medium | |
| WO2019119505A1 (en) | Face recognition method and device, computer device and storage medium | |
| CN109766759A (en) | Emotion recognition method and related products | |
| CN104091176A (en) | Technology for applying figure and head portrait comparison to videos | |
| TW202115622A (en) | Face attribute recognition method, electronic device and computer-readable storage medium | |
| KR20170006355A (en) | Method of motion vector and feature vector based fake face detection and apparatus for the same | |
| CN104298983A (en) | Tongue fur image acquisition and analysis system with distributed user terminals | |
| CN110414376B (en) | Method for updating face recognition model, face recognition camera and server | |
| CN103020602B (en) | Based on the face identification method of neural network | |
| WO2019056503A1 (en) | Store monitoring evaluation method, device and storage medium | |
| TWI525574B (en) | Collaborative face annotation method and collaborative face annotation system | |
| WO2023273297A1 (en) | Multi-modality-based living body detection method and apparatus, electronic device, and storage medium | |
| CN110765134A (en) | File establishing method, equipment and storage medium | |
| TW201917636A (en) | A method of face recognition based on online learning | |
| CN113705469A (en) | Face recognition method and device, electronic equipment and computer readable storage medium | |
| JP5004181B2 (en) | Region identification device and content identification device | |
| CN110503031B (en) | Method for improving face recognition accuracy and passing speed | |
| CN111382596A (en) | Face recognition method and device and computer storage medium | |
| CN111080827A (en) | Attendance system and method | |
| CN112037235B (en) | Injury picture automatic auditing method and device, electronic equipment and storage medium | |
| CN104217503A (en) | Self-service terminal identity identification method and corresponding house property certificate printing method | |
| CN109345427B (en) | Classroom video frequency point arrival method combining face recognition technology and pedestrian recognition technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 13 / F, building a, it park, Wenyi West Road, Yuhang District, Hangzhou City, Zhejiang Province Patentee after: Hangzhou Yufan Intelligent Technology Co.,Ltd. Country or region after: China Address before: 13 / F, building a, it park, Wenyi West Road, Yuhang District, Hangzhou City, Zhejiang Province Patentee before: UNIVERSAL UBIQUITOUS TECHNOLOGY Co.,Ltd. Country or region before: China |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A method to improve the accuracy and speed of facial recognition Granted publication date: 20211126 Pledgee: Guotou Taikang Trust Co.,Ltd. Pledgor: Hangzhou Yufan Intelligent Technology Co.,Ltd. Registration number: Y2024980020891 |