CN108509885A - A kind of efficient identity card picture screening technique - Google Patents
A kind of efficient identity card picture screening technique Download PDFInfo
- Publication number
- CN108509885A CN108509885A CN201810253794.0A CN201810253794A CN108509885A CN 108509885 A CN108509885 A CN 108509885A CN 201810253794 A CN201810253794 A CN 201810253794A CN 108509885 A CN108509885 A CN 108509885A
- Authority
- CN
- China
- Prior art keywords
- text
- line
- identity card
- face
- region
- 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
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of efficient identity card picture screening techniques, the approximate region of identity card is determined by finding the face in image, line of text detection is carried out in this region, the text line discipline of ID card information is closed by line of text testing result set symbol, to filter out non-identity card picture.The method of the present invention copes with the photo under different light environments, edge blurry, noise, the interference such as endless full circle, strong antijamming capability, keeps extremely low misclassification rate, screening accuracy is high and takes short.It is screened relative to artificial screening and traditional image processing method, this method performs better than in comprehensive performance, and disclosure satisfy that actual demand, is embodied in robustness, high efficiency.
Description
Technical field
The invention belongs to computer vision, technical field of image processing, are related to a kind of optical sieving technology, more specifically
It says, is to be related to a kind of efficient identity card picture screening technique.
Background technology
In recent years, with the improvement of people's living standards, the heavy traffic of bank etc., and credit card is handled etc.
Vocational window handle more busy, and due to the interest of internet, direct web handling becomes key.Web handling is such
Business needs diversified personal information, wherein mostly important is exactly identity card, that is, needs to upload the identity card shot
Photo.Such picture data amount is prodigious, if can carry out carrying out identity before last information verification in staff
The intelligent screening of card will greatly facilitate staff, while can also improve working efficiency.
Existing screening technique includes artificial screening and traditional images processing filtering algorithm.Artificial screening is simplest
Screening technique reaches screening purpose by the eye of people, brain cooperation.This method defect is it is also obvious that since human eye judges simultaneously
Nonstandard, specification, error rate is also relatively high, and artificial efficiency is extremely inefficient, in huge data now it is more aobvious its
Efficiency it is low.Traditional image procossing filtering algorithm, is operated by simple image binaryzation first, and then is carried out image and gone
It makes an uproar processing, later by edge extracting, such as Canny operators, so that it is determined that the approximate region of identity card, finally passes through this region again
Binaryzation information determine whether for ID Card Image.The influences such as this method is illuminated by the light, shooting angle are serious, and are carrying out
When identity card extracted region, difference between foreground picture and Background must be sufficiently large, in this way could could in edge extracting
Obtain more complete marginal information.
It will appear various interference, such as illumination light and shade in the environment of ID Card Image screening, obscure, white Gaussian
The influences such as noise, shooting angle.Above each factor is removed, it is prior to be the part identities such as ethnic group card above also
Contain more information, such as the spoken and written languages of national minorities, in addition, how to exclude captured photo include former identity card and
Non- copy, these are all very crucial problems.Requirement of these problems for filtering algorithm is very high, and algorithm of today is difficult
Meet demand.Many Face datections are all the detection interfaces carried using OpenCV, that is, use haar to detect face, although convenient
It calls, but there are poor performance, slow-footed problems.Moreover, the line of text detection in natural scene is also to learn all the time
One of research emphasis of art circle such as carries out text detection using support vector machines (SVM), there is no apply in ID Card Image at present
Good Method for text detection in screening.
Invention content
To solve screening accuracy present in existing identity card screening technique not high, robustness takes and stability
The problems such as comprehensive performance is poor, the invention discloses a kind of efficient identity card picture screening techniques, by finding the people in image
Face determines the approximate region of identity card, carries out line of text detection in this region, passes through the fit part of line of text testing result set symbol
The text line discipline for demonstrate,proving information, to filter out non-identity card picture.
In order to achieve the above object, the present invention provides the following technical solutions：
A kind of efficient identity card picture screening technique, includes the following steps：
Step 1, Face datection
It is right in one direction when not detecting face using the human face region and facial angle in the detection image of the libraries dlib
Row detection again after 90 ° of image rotation, until detecting face；
Step 2, adjustment are horizontal
According to the key point of the face detected in step 1, facial angle is calculated, is adjusted again according to this angle
To image level；
Step 3, line of text detection
Using the deep learning MODEL C TPN algorithms based on caffe to obtaining the region of identity card in Face datection into style of writing
One's own profession detects, and obtains the position of line of text；
Step 4 does not conform to table images according to screening rule deletion
Screening rule is set, is not inconsistent if any a rule, filters, meet strictly all rules and then pass through；The screening rule packet
It includes：
(1) line of text line number is more than or equal to 5；
(2) face location is on the right side of all line of text in addition to most having a line；
(3) the length longest of last column text；
(4) face location is in the upside of last column line of text.
Further, further include copy filtration step between the step 2 and step 3：
Color Channel analysis is carried out to the human face region obtained after step 1 Face datection, if in this region major part pixel
The value of the RGB triple channels of point is essentially identical, then it is determined that being copy, disease is filtered.
Further, direction of rotation is clockwise in the step 1.
Further, Shuo≤3 time Ci are rotated in the step 1.
Further, facial angle is calculated by following procedure in the step 2:
Facial angle is calculated separately out according to two test points of two vertical test points of face and level, the two takes
Averagely it is worth to final facial angle.
Compared with prior art, the invention has the advantages that and advantageous effect：
The method of the present invention copes with the photo under different light environments, edge blurry, noise, the interference such as endless full circle,
Strong antijamming capability keeps extremely low misclassification rate, and screening accuracy is high and takes short.Relative to artificial screening and traditional
Image processing method screens, and this method performs better than in comprehensive performance, and disclosure satisfy that actual demand, in robustness, height
It is embodied in effect property.
Description of the drawings
Fig. 1 is a kind of efficient identity card picture screening technique steps flow chart schematic diagram provided by the invention.
Fig. 2 is that example photo carries out screening process figure according to the method for the present invention.
Fig. 3 is the extraction figure in face detection process and identity card region.
Fig. 4 is the line of text detection figure of non-ID Card Image.
Specific implementation mode
Technical solution provided by the invention is described in detail below with reference to specific embodiment, it should be understood that following specific
Embodiment is only illustrative of the invention and is not intended to limit the scope of the invention.
Firstly, it is necessary to be handled by the image that server shoots the hardware devices such as mobile phone, the present invention is then used
Whether comprising clearly ID card information in method intelligent distinguishing image, reach screening purpose.The step flow of present invention entirety
As shown in Figure 1.Face datection is carried out by rotating 90 ° of artwork successively first, if detecting face, according to the people detected
Face angle is rotated to level, then is intercepted the rough identity card region of acquisition and passed through trained text detection depth later
Learning model detects line of text, finally according to the position relationship of line of text, it is determined whether is ID Card Image.
The present invention is the screening technique detected based on Face datection and line of text, and can be by carrying out figure in human face region
The RGB threeway trace analysis of picture carries out screening out for copy.Specifically, the method for the present invention mainly includes the following steps：
Step 1, Face datection
This step determines the approximate region of identity card by finding the face in image.In view of high efficiency and accuracy,
The present invention has selected the libraries dlib to detect human face region and facial angle, and summation considers detection result, computing resource, detection
These factors of speed.In today that deep learning is prevailing, although detection speed of the deep learning algorithm compared with dlib will with precision
It is higher, but great loss will be had in terms of computing resource, the text detection and high concurrent operation behind opposite have
It is certain to influence, therefore dlib is used to carry out Face datection.
The photo of general natural scene shooting, the angle of face is indefinite, such as Fig. 3 (a), -180 °~180 ° it
Between be likely to, and Face datection can just detected between -60 °~60 °, and angle is bigger, and detection result is poorer.It presses successively
The same direction (such as clockwise) be rotated by 90 °, as soon as often rotation is primary to carry out time Face datection, no longer when detecting face
It is rotated, such figure rotates through up three times, at most carries out 4 Face datections.Mode three times, energy are rotated through up in this way
It is enough that the facial angle of detection has been narrowed down to -45 °~45 °, to improve the effect of detection.Fig. 3 (b) is that artwork 3 (a) carries out
After primary rotation as a result, face can be detected at this time, such as Fig. 3 (c).
Step 2, adjustment are horizontal
The face key point detected in (c) according to fig. 3, if two vertical test points of face are (among Face datection point
Uppermost point and nethermost point on line) and horizontal two test points (the leftmost point of Face datection point and rightmost
Point) calculate separately out facial angle, the two is averaged to obtain final facial angle.Others can certainly be used
Test point carries out facial angle calculating, such as two points of forehead or so, but considers calculating speed and precision, has only used face
Two test points of vertical two test points and level.Subtle rotation is carried out again according to this angle to adjust to level,
Purpose is in order to make the word in image all in horizontal direction, to which the effect of line of text detection is more preferable.Final adjustment knot
Fruit such as Fig. 3 (d).
As an improvement, the present invention also adds copy filter method after step 2.
Copy of ID Card now is substantially black and white, different and colored identity card artwork.It filters out such
Image, most simple and effective way are exactly to be analyzed by the RGB triple channels of image.Black and white copying part is shot in face
What region showed is also black and white, so the human face region directly obtained after previous step Face datection carries out Color Channel
Analysis.If essentially identical in the value of the RGB triple channels of this region major part pixel, it is determined that being copy, at this time
It can filter this out.
Step 3, line of text detection
Line of text detection above identity card is very simple, but due to natural scene picture shooting angle, illumination etc.
The limitation of aspect, the effect to obtain very robust is highly difficult, and the series of algorithms performance in terms of deep learning is very
Excellent, then abandon traditional text line detection method based on image procossing, the detection side for the use of deep learning
Method, i.e. the deep learning MODEL C TPN algorithms based on caffe.After Face datection obtains identity card approximate region, this region
Image carry out line of text detection, obtain the position of line of text.
Step 4 does not conform to table images according to screening rule deletion
Those non-identity cards, unintelligible, high light image are filtered out, it can be by text line position that CTPN algorithms obtain
It sets and is screened.If Fig. 3 is non-ID Card Image, the position of its line of text and the text line position of identity card can be clearly found out
It sets and is very different, according to these differences, setting series of rules can filter this out.
Rule is as follows：
(1) line of text line number is more than or equal to 5, passes through；Otherwise, it filters.
(2) face location passes through on the right side of all line of text (in addition to last column)；Otherwise, it filters.
(3) the length longest of last column text, passes through；Otherwise, it filters.
(4) face location passes through in the upside of last column line of text；Otherwise, it filters.
The above screening rule seems very simple, but can screen out the overwhelming majority non-identity card, it is unintelligible,
High light image, and breakneck acceleration quickly can be reached.
As an improvement, we can also realize above step high concurrent operation.According to screening technique above into stroke
Sequence writes test and test, it is found that (most of include identity card to the image set for all including to all kinds of interference, and total amount of images is
100) it is that (hardware condition is i7 processors to 500ms or so, and GTX 1080 is aobvious to carry out the average breakneck acceleration that single thread is tested
Card).As can be seen that the speed of single thread or inadequate.Program is optimized at this time, program is no longer that single thread carries out, and is changed
For multithreading high concurrent, obtained average breakneck acceleration is 150ms or so.Such speed has had reached breakneck acceleration and has wanted
It asks.
The technical means disclosed in the embodiments of the present invention is not limited only to the technological means disclosed in the above embodiment, further includes
By the above technical characteristic arbitrarily the formed technical solution of combination.It should be pointed out that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (5)
1. a kind of efficient identity card picture screening technique, which is characterized in that include the following steps：
Step 1, Face datection
Using the human face region and facial angle in the detection image of the libraries dlib, when not detecting face in one direction to image
Row detection again after being rotated by 90 °, until detecting face；
Step 2, adjustment are horizontal
According to the key point of the face detected in step 1, facial angle is calculated, is adjusted again to figure according to this angle
As horizontal；
Step 3, line of text detection
Line of text is carried out to the region for obtaining identity card in Face datection using the deep learning MODEL C TPN algorithms based on caffe
Detection, obtains the position of line of text；
Step 4 does not conform to table images according to screening rule deletion
Screening rule is set, is not inconsistent if any a rule, filters, meet strictly all rules and then pass through；The screening rule includes：
(1) line of text line number is more than or equal to 5；
(2) face location is on the right side of all line of text in addition to most having a line；
(3) the length longest of last column text；
(4) face location is in the upside of last column line of text.
2. efficient identity card picture screening technique according to claim 1, it is characterised in that：The step 2 and step
It further include copy filtration step between three：
Color Channel analysis is carried out to the human face region obtained after step 1 Face datection, if in this region major part pixel
The value of RGB triple channels is essentially identical, then it is determined that being copy, disease is filtered.
3. efficient identity card picture screening technique according to claim 1 or 2, it is characterised in that：In the step 1
Direction of rotation is clockwise.
4. efficient identity card picture screening technique according to claim 1 or 2, it is characterised in that：In the step 1
Rotate Shuo≤3 time Ci.
5. efficient identity card picture screening technique according to claim 1 or 2, it is characterised in that：In the step 2
Facial angle is calculated by following procedure:
Facial angle is calculated separately out according to two test points of two vertical test points of face and level, the two is averaged
It is worth to final facial angle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810253794.0A CN108509885A (en) | 2018-03-26 | 2018-03-26 | A kind of efficient identity card picture screening technique |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810253794.0A CN108509885A (en) | 2018-03-26 | 2018-03-26 | A kind of efficient identity card picture screening technique |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108509885A true CN108509885A (en) | 2018-09-07 |
Family
ID=63378528
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810253794.0A Pending CN108509885A (en) | 2018-03-26 | 2018-03-26 | A kind of efficient identity card picture screening technique |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108509885A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109886002A (en) * | 2019-02-27 | 2019-06-14 | 苏州迈荣祥信息科技有限公司 | A kind of the use control method and mobile terminal of application software |
| CN110889341A (en) * | 2019-11-12 | 2020-03-17 | 广州供电局有限公司 | Form image recognition method and device based on AI (Artificial Intelligence), computer equipment and storage medium |
| CN111950554A (en) * | 2020-08-17 | 2020-11-17 | 深圳市丰巢网络技术有限公司 | Identification card identification method, device, equipment and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521569A (en) * | 2011-11-30 | 2012-06-27 | 康佳集团股份有限公司 | Method and system for identifying identity card by using smart phone and mobile phone |
| CN103345622A (en) * | 2013-07-09 | 2013-10-09 | 浙江省公安厅居民身份证制作中心 | Method for controlling quality of character pictures on second-generation identification cards |
| CN103488998A (en) * | 2013-09-11 | 2014-01-01 | 东华大学 | Identity card recognition method based on neural network and image processing technology |
| CN105528602A (en) * | 2015-10-30 | 2016-04-27 | 小米科技有限责任公司 | Region identification method and device |
| CN105701488A (en) * | 2016-01-01 | 2016-06-22 | 广州恒巨信息科技有限公司 | Identity card identification method |
| CN106547912A (en) * | 2016-11-25 | 2017-03-29 | 西安理工大学 | The identification of non-China second-generation identity card photo and elimination method in identity card database |
| CN106909882A (en) * | 2017-01-16 | 2017-06-30 | 广东工业大学 | A kind of face identification system and method for being applied to security robot |
| CN106991421A (en) * | 2017-03-22 | 2017-07-28 | 湖南联信科技有限公司 | A kind of ID card information extraction system |
| CN107346420A (en) * | 2017-06-19 | 2017-11-14 | 中国科学院信息工程研究所 | Text detection localization method under a kind of natural scene based on deep learning |
| CN107358201A (en) * | 2017-07-13 | 2017-11-17 | 杭州有盾网络科技有限公司 | A kind of photo array method, apparatus and system |
| CN107369086A (en) * | 2017-07-06 | 2017-11-21 | 上海你我贷互联网金融信息服务有限公司 | A kind of identity card stamp system and method |
-
2018
- 2018-03-26 CN CN201810253794.0A patent/CN108509885A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521569A (en) * | 2011-11-30 | 2012-06-27 | 康佳集团股份有限公司 | Method and system for identifying identity card by using smart phone and mobile phone |
| CN103345622A (en) * | 2013-07-09 | 2013-10-09 | 浙江省公安厅居民身份证制作中心 | Method for controlling quality of character pictures on second-generation identification cards |
| CN103488998A (en) * | 2013-09-11 | 2014-01-01 | 东华大学 | Identity card recognition method based on neural network and image processing technology |
| CN105528602A (en) * | 2015-10-30 | 2016-04-27 | 小米科技有限责任公司 | Region identification method and device |
| CN105701488A (en) * | 2016-01-01 | 2016-06-22 | 广州恒巨信息科技有限公司 | Identity card identification method |
| CN106547912A (en) * | 2016-11-25 | 2017-03-29 | 西安理工大学 | The identification of non-China second-generation identity card photo and elimination method in identity card database |
| CN106909882A (en) * | 2017-01-16 | 2017-06-30 | 广东工业大学 | A kind of face identification system and method for being applied to security robot |
| CN106991421A (en) * | 2017-03-22 | 2017-07-28 | 湖南联信科技有限公司 | A kind of ID card information extraction system |
| CN107346420A (en) * | 2017-06-19 | 2017-11-14 | 中国科学院信息工程研究所 | Text detection localization method under a kind of natural scene based on deep learning |
| CN107369086A (en) * | 2017-07-06 | 2017-11-21 | 上海你我贷互联网金融信息服务有限公司 | A kind of identity card stamp system and method |
| CN107358201A (en) * | 2017-07-13 | 2017-11-17 | 杭州有盾网络科技有限公司 | A kind of photo array method, apparatus and system |
Non-Patent Citations (1)
| Title |
|---|
| ZHI TIAN ET AL: "Detecting Text in Natural Image with Connectionist Text Proposal Network", 《EUROPEAN CONFERENCE ON COMPUTER VISION 2016》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109886002A (en) * | 2019-02-27 | 2019-06-14 | 苏州迈荣祥信息科技有限公司 | A kind of the use control method and mobile terminal of application software |
| CN110889341A (en) * | 2019-11-12 | 2020-03-17 | 广州供电局有限公司 | Form image recognition method and device based on AI (Artificial Intelligence), computer equipment and storage medium |
| CN111950554A (en) * | 2020-08-17 | 2020-11-17 | 深圳市丰巢网络技术有限公司 | Identification card identification method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Patel et al. | Live face video vs. spoof face video: Use of moiré patterns to detect replay video attacks | |
| WO2019153739A1 (en) | Identity authentication method, device, and apparatus based on face recognition, and storage medium | |
| CN105631417B (en) | Video enhancement system and method applied to internet video live streaming | |
| CN106529414A (en) | Method for realizing result authentication through image comparison | |
| WO2019134536A1 (en) | Neural network model-based human face living body detection | |
| WO2019137178A1 (en) | Face liveness detection | |
| WO2019061658A1 (en) | Method and device for positioning eyeglass, and storage medium | |
| CN106056064A (en) | Face recognition method and face recognition device | |
| CN105894655B (en) | Paper currency detection and recognition methods under complex environment based on RGB-D cameras | |
| CN104794693B (en) | A kind of portrait optimization method of face key area automatic detection masking-out | |
| Lu et al. | Robust blur kernel estimation for license plate images from fast moving vehicles | |
| CN105787427B (en) | Lip region localization method | |
| CN111259891B (en) | Method, device, equipment and medium for identifying identity card in natural scene | |
| CN108509885A (en) | A kind of efficient identity card picture screening technique | |
| CN107172354A (en) | Method for processing video frequency, device, electronic equipment and storage medium | |
| CN109598210A (en) | A kind of image processing method and device | |
| CN104361357B (en) | Photo album categorizing system and sorting technique based on image content analysis | |
| CN111079688A (en) | Living body detection method based on infrared image in face recognition | |
| CN104598881B (en) | Feature based compresses the crooked scene character recognition method with feature selecting | |
| CN105913389B (en) | Image processing method and device during skin abnormality | |
| Fujio et al. | Face/Fingerphoto Spoof Detection under Noisy Conditions by using Deep Convolutional Neural Network. | |
| CN114332983A (en) | Face image definition detection method, face image definition detection device, electronic equipment and medium | |
| CN112907206A (en) | Service auditing method, device and equipment based on video object identification | |
| CN111126283A (en) | Rapid in-vivo detection method and system for automatically filtering fuzzy human face | |
| Liu et al. | Effectively localize text in natural scene images |
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: 20180907 |