CN108446682A - A kind of recognition of face calibration detection method of full-automation - Google Patents
A kind of recognition of face calibration detection method of full-automation Download PDFInfo
- Publication number
- CN108446682A CN108446682A CN201810448835.1A CN201810448835A CN108446682A CN 108446682 A CN108446682 A CN 108446682A CN 201810448835 A CN201810448835 A CN 201810448835A CN 108446682 A CN108446682 A CN 108446682A
- Authority
- CN
- China
- Prior art keywords
- face
- recognition
- lbp
- dbn
- rbm
- 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.)
- Pending
Links
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
- G06V40/168—Feature extraction; Face representation
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/30—Scenes; Scene-specific elements in albums, collections or shared content, e.g. social network photos or video
-
- 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
Abstract
The invention discloses a kind of recognitions of face of full-automation to calibrate detection method, includes extracting description facial image Local textural feature using LBP operators to a large amount of facial images in database；Using LBP textural characteristics as the input of DBN, DBN training is carried out；The facial image for the need identification that camera is got after carrying out feature extraction using LBP operators, identifies facial image in DBN top layers, is compared with the feature in database, when result is true, realizes recognition of face detection.Detection method is calibrated in the recognition of face of full-automation of the present invention, and database constantly, which is added, in new facial image and feature is identified calibration, to which the accuracy rate and precision of identification be continuously improved.
Description
Technical field
The present invention relates to technical field of face recognition, and in particular, to a kind of recognition of face calibration detection of full-automation
Method.
Background technology
With the continuous development of technology, the recognition methods such as traditional magnetic card, password are easily lost, are easy due to existing
The problems such as replicating, being easy to forget is subjected to more and more challenging, and reliability constantly declines.Compared with other physiological characteristics, face
Feature has the characteristics that active, direct, simple, close friend, progresses into research and uses scope.
Face recognition process is divided into three parts.One, facial image is acquired, face database is established.Secondly, propose
Feature, then extracts characteristic in feature database.Finally, personally identifiable information is identified, with specific matching algorithm
The face characteristic of identified person is matched, to realize the identification to the characteristic in database.Existing face recognition technology
It is influenced by following factor, causes the decline of recognition correct rate.1. expression, when the expression shape change of face will cause algorithm discrimination
Drastically decline.Second is that appearance changes, different age brackets, face has nuance, and the change of face characteristic increases identification hardly possible
Degree.Third, ornament, makeup, glasses or cap impact recognition result.Fourth, light, the unstability shadow of shadow condition
Ring recognition correct rate.
Invention content
The purpose of the present invention is to provide a kind of recognitions of face of full-automation to calibrate detection method, to solve above-mentioned background
The problem of being proposed in technology.
To achieve the above object, the present invention provides the following technical solutions：
A kind of recognition of face calibration detection method of full-automation, includes the following steps：
S1. to a large amount of facial images in database, description face figure is extracted using LBP (local binary) pattern operators
As Local textural feature, original LBP operator definitions be in the window of 3*3, will be adjacent using window center pixel as threshold value
The gray value of 8 pixels is compared with it, if surrounding pixel values are more than center pixel value, then the position of the pixel is labeled
It is 1, is otherwise 0；In this way, 8 points in the fields 3*3 can generate the unsigned number of 8bit, that is, obtain the LBP values of the window, and
Reflect the texture information in the region with this value.Here, LBP operators can be retouched as a kind of effective texture description operator
The distribution at bright spot, dim spot, edge in facial image etc. is stated, it is calculated, and simple, arithmetic speed is fast, and has illumination and rotation
Constant characteristic.
S2. using LBP textural characteristics as the input of DBN (depth belief network), carry out DBN training, i.e., it is special with LBP textures
Sign is input, carries out unsupervised training to first layer RBM, obtains the optimized parameter of this layer；Secondly by the output number of low one layer of RBM
Unsupervised training is carried out according to the input data as high one layer of RBM, and to this layer of RBM, obtains optimized parameter；Finally utilize the overall situation
Trained method is finely adjusted trained each layer parameter, and DBN is made to converge to global optimum.
S3. the facial image of need identification camera got after carrying out feature extraction using LBP operators, is pushed up in DBN
Layer identification facial image, is compared with the feature in database, when result is true, realizes recognition of face detection.
Further, in the S2 steps, first individually each convolutional network is trained then to pass through all convolution of fixation
Training of the real-time performance to RBM makes model in initialization closer to a good local minimum, finally, whole network
It is finely adjusted by the reverse propagated error of the RBM from top to the convolutional neural networks of all low layers.
Further, in the S3 steps, when comparison result be fictitious time, by the facial image newly obtained and the line of extraction
It manages feature and is stored in database, carry out DBN training.
Due to higher to hardware requirement, further, S1, S2 step carries out on specialized server.
Compared with prior art, the beneficial effects of the invention are as follows：The recognition of face of full-automation of the present invention is calibrated
Detection method describes facial image Local textural feature using LBP operator extractions, then is trained by DBP, by feature
It is compared under the conditions of DBP, database constantly, which is added, in new facial image and feature is identified calibration, to constantly carry
The accuracy rate and precision of height identification.
Description of the drawings
Fig. 1 is the block diagram of the embodiment of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Shown in referring to Fig.1, a kind of recognition of face calibration detection method of full-automation includes the following steps：
S1. to a large amount of facial images in database, description face figure is extracted using LBP (local binary) pattern operators
As Local textural feature, original LBP operator definitions be in the window of 3*3, will be adjacent using window center pixel as threshold value
The gray value of 8 pixels is compared with it, if surrounding pixel values are more than center pixel value, then the position of the pixel is labeled
It is 1, is otherwise 0；In this way, 8 points in the fields 3*3 can generate the unsigned number of 8bit, that is, obtain the LBP values of the window, and
Reflect the texture information in the region with this value.Here, LBP operators can be retouched as a kind of effective texture description operator
The distribution at bright spot, dim spot, edge in facial image etc. is stated, it is calculated, and simple, arithmetic speed is fast, and has illumination and rotation
Constant characteristic.
S2. using LBP textural characteristics as the input of DBN (depth belief network), carry out DBN training, i.e., it is special with LBP textures
Sign is input, carries out unsupervised training to first layer RBM, obtains the optimized parameter of this layer；Secondly by the output number of low one layer of RBM
Unsupervised training is carried out according to the input data as high one layer of RBM, and to this layer of RBM, obtains optimized parameter；Finally utilize the overall situation
Trained method is finely adjusted trained each layer parameter, and DBN is made to converge to global optimum.
S3. the facial image of need identification camera got after carrying out feature extraction using LBP operators, is pushed up in DBN
Layer identification facial image, is compared with the feature in database, when result is true, realizes recognition of face detection.
Further, in the S2 steps, first individually each convolutional network is trained then to pass through all convolution of fixation
Training of the real-time performance to RBM makes model in initialization closer to a good local minimum, finally, whole network
It is finely adjusted by the reverse propagated error of the RBM from top to the convolutional neural networks of all low layers.
Further, in the S3 steps, when comparison result be fictitious time, by the facial image newly obtained and the line of extraction
It manages feature and is stored in database, carry out DBN training.
Due to higher to hardware requirement, further, S1, S2 step carries out on specialized server.
It although an embodiment of the present invention has been shown and described, for the ordinary skill in the art, can be with
Understanding without departing from the principles and spirit of the present invention can carry out these embodiments a variety of variations, modification, replace
And modification, the scope of the present invention is defined by the appended.
Claims (4)
1. detection method is calibrated in a kind of recognition of face of full-automation, which is characterized in that include the following steps：
S1. to a large amount of facial images in database, description facial image office is extracted using LBP (local binary) pattern operators
Portion's textural characteristics, original LBP operator definitions are in the window of 3*3, using window center pixel as threshold value, by adjacent 8
The gray value of pixel is compared with it, if surrounding pixel values are more than center pixel value, then the position of the pixel is marked as
1, it is otherwise 0；In this way, 8 points in the fields 3*3 can generate the unsigned number of 8bit, that is, obtain the LBP values of the window, be used in combination
This value reflects the texture information in the region；
S2. using LBP textural characteristics as the input of DBN (depth belief network), DBN training is carried out, i.e., is with LBP textural characteristics
Input carries out unsupervised training to first layer RBM, obtains the optimized parameter of this layer；Secondly the output data of low one layer of RBM is made
For the input data of high one layer of RBM, and unsupervised training is carried out to this layer of RBM, obtains optimized parameter；Finally utilize global training
Method trained each layer parameter is finely adjusted, so that DBN is converged to global optimum；
S3. the facial image of need identification camera got after carrying out feature extraction using LBP operators, is known in DBN top layers
Others' face image, is compared with the feature in database, when result is true, realizes recognition of face detection.
2. detection method is calibrated in the recognition of face of full-automation according to claim 1, which is characterized in that walked in the S2
In rapid, each convolutional network is first individually trained, then, the training to RBM is realized by all convolutional networks of fixation, model is made to exist
Closer to a good local minimum when initialization, finally, whole network passes through the RBM from top to all low layers
The reverse propagated error of convolutional neural networks is finely adjusted.
3. detection method is calibrated in the recognition of face of full-automation according to claim 1 or 2, which is characterized in that described
In S3 steps, when comparison result is fictitious time, the facial image newly obtained and the textural characteristics of extraction is stored in database, are carried out
DBN is trained.
4. detection method is calibrated in the recognition of face of full-automation according to claim 3, which is characterized in that described S1, S2
Step carries out on specialized server.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810448835.1A CN108446682A (en) | 2018-05-11 | 2018-05-11 | A kind of recognition of face calibration detection method of full-automation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810448835.1A CN108446682A (en) | 2018-05-11 | 2018-05-11 | A kind of recognition of face calibration detection method of full-automation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108446682A true CN108446682A (en) | 2018-08-24 |
Family
ID=63203585
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810448835.1A Pending CN108446682A (en) | 2018-05-11 | 2018-05-11 | A kind of recognition of face calibration detection method of full-automation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108446682A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114626860A (en) * | 2022-05-12 | 2022-06-14 | 武汉和悦数字科技有限公司 | Dynamic identity identification method and device for online commodity payment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102982321A (en) * | 2012-12-05 | 2013-03-20 | 深圳Tcl新技术有限公司 | Acquisition method and device for face database |
| US20160110590A1 (en) * | 2014-10-15 | 2016-04-21 | University Of Seoul Industry Cooperation Foundation | Facial identification method, facial identification apparatus and computer program for executing the method |
| CN106778512A (en) * | 2016-11-25 | 2017-05-31 | 南京蓝泰交通设施有限责任公司 | Face identification method under the conditions of a kind of unrestricted based on LBP and depth school |
| CN107578007A (en) * | 2017-09-01 | 2018-01-12 | 杭州电子科技大学 | A kind of deep learning face identification method based on multi-feature fusion |
| CN107909034A (en) * | 2017-11-15 | 2018-04-13 | 清华大学深圳研究生院 | A kind of method for detecting human face, device and computer-readable recording medium |
-
2018
- 2018-05-11 CN CN201810448835.1A patent/CN108446682A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102982321A (en) * | 2012-12-05 | 2013-03-20 | 深圳Tcl新技术有限公司 | Acquisition method and device for face database |
| US20160110590A1 (en) * | 2014-10-15 | 2016-04-21 | University Of Seoul Industry Cooperation Foundation | Facial identification method, facial identification apparatus and computer program for executing the method |
| CN106778512A (en) * | 2016-11-25 | 2017-05-31 | 南京蓝泰交通设施有限责任公司 | Face identification method under the conditions of a kind of unrestricted based on LBP and depth school |
| CN107578007A (en) * | 2017-09-01 | 2018-01-12 | 杭州电子科技大学 | A kind of deep learning face identification method based on multi-feature fusion |
| CN107909034A (en) * | 2017-11-15 | 2018-04-13 | 清华大学深圳研究生院 | A kind of method for detecting human face, device and computer-readable recording medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114626860A (en) * | 2022-05-12 | 2022-06-14 | 武汉和悦数字科技有限公司 | Dynamic identity identification method and device for online commodity payment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105917353B (en) | Feature extraction and matching for biological identification and template renewal | |
| CN110543846B (en) | Multi-pose face image obverse method based on generation countermeasure network | |
| CN104915972B (en) | Image processing apparatus, image processing method and program | |
| CN110009057B (en) | Graphic verification code identification method based on deep learning | |
| CN109002755B (en) | Age estimation model construction method and estimation method based on face image | |
| CN111967592B (en) | Method for generating countermeasure image machine identification based on separation of positive and negative disturbance | |
| CN109657612B (en) | Quality sorting system based on facial image features and application method thereof | |
| CN106875007A (en) | End-to-end deep neural network is remembered based on convolution shot and long term for voice fraud detection | |
| CN106991380A (en) | A kind of preprocess method based on vena metacarpea image | |
| CN110232326A (en) | A kind of D object recognition method, device and storage medium | |
| CN106529395A (en) | Signature image recognition method based on deep brief network and k-means clustering | |
| Barua et al. | Fingerprint identification | |
| CN113793357A (en) | Bronchopulmonary segment image segmentation method and system based on deep learning | |
| CN110008912B (en) | Social platform matching method and system based on plant identification | |
| Liu et al. | Iris recognition in visible spectrum based on multi-layer analogous convolution and collaborative representation | |
| CN113128364B (en) | Fingerprint biological key generation method based on deep neural network coding | |
| CN107516083A (en) | A kind of remote facial image Enhancement Method towards identification | |
| CN112116012B (en) | Finger vein instant registration and identification method and system based on deep learning | |
| CN116822548B (en) | Method for generating high recognition rate AI two-dimensional code and computer readable storage medium | |
| CN108446682A (en) | A kind of recognition of face calibration detection method of full-automation | |
| CN111027616B (en) | Line characteristic description system based on end-to-end learning | |
| CN114882582A (en) | Gait recognition model training method and system based on federal learning mode | |
| CN114550266A (en) | Face recognition method and device, intelligent door lock and computer readable storage medium | |
| Lei et al. | Iris recognition based on few‐shot learning | |
| Chirchi et al. | Modified circular fuzzy segmentor and local circular encoder to iris segmentation and recognition |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180824 |