CN102567737A - Method for locating eyeball cornea - Google Patents
Method for locating eyeball cornea Download PDFInfo
- Publication number
- CN102567737A CN102567737A CN2011104497518A CN201110449751A CN102567737A CN 102567737 A CN102567737 A CN 102567737A CN 2011104497518 A CN2011104497518 A CN 2011104497518A CN 201110449751 A CN201110449751 A CN 201110449751A CN 102567737 A CN102567737 A CN 102567737A
- Authority
- CN
- China
- Prior art keywords
- cornea
- circle
- center
- image
- eye
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method for locating an eyeball cornea. The method comprises the steps of utilizing a CV mode identification solution in opencv (open source computer vision library) for training to obtain an eye classifier and generating a relevant XML (extensible makeup language) file; setting a rectangular eye region captured by a camera as a region of interest (ROI); carrying out smoothing to an ROI image, and carrying out edge detection; and carrying out Hough circular transformation in an obtained edge image by utilizing the characteristic that an eye cornea has remarkable circular shape, and identifying a circular sequence in the region, namely, determining that the sequence is the region where the cornea is located and the position where the center of circle is located is the pupil. An accurate and reliable position of the eye cornea can be obtained by adopting the process of the invention. The detection method has the advantages of wide application scope, high identification precision, simple detection equipment and the like.
Description
Technical field
The present invention relates to computer vision and handle, specifically be meant the method for a kind of cornea location.
Background technology
Because the development of the technology of Flame Image Process and video identification, image recognition also are applied to comprise security protection and get rid of the danger daily life that monitoring anti-theft etc. are various and industrial field.Human eye recognition and tracking technology then is an important drop applications of image recognition, in its main application sets with biomechanics and field of human-computer interaction.Detection method in the eye detection technology mainly contains three kinds: 1, light reflection detection 2, contact lense detection 3, skin potential detect.4, sclera iris edge reflections photosensitive tube detection method etc.But but this detection method applied range is narrow, and exists accuracy of identification relatively poor, and checkout equipment is required high shortcoming.
Summary of the invention
The objective of the invention is to overcome the shortcoming and defect of prior art, the method for a kind of cornea location is provided.Advantages such as the present invention can obtain the position of eyes cornea accurately and reliably, and it is wide to have the application scope, and checkout equipment is simple.
The present invention realizes through following technical proposals:
The method of a kind of cornea location is used the CV pattern-recognition solution among the opencv, and Flame Image Process is carried out in identification eyeball zone.
Under Visual C++6.0 platform, at first utilize the CV pattern-recognition solution among the opencv, training obtains the eyes sorter, generates relevant XML file; Eye rectangular area that camera captures is set as area-of-interest; Then the region of interest area image is carried out smoothing processing, rim detection; In the edge image that obtains, utilize cornea to have tangible rounded form characteristic, carry out Hough circle transformation, identify circle sequence in this zone, judge that promptly this sequence is the cornea region, its position, center of circle is pupil; Obtain the accurate position of eyes cornea through above-mentioned processing.
The method of above-mentioned cornea location comprises the steps:
1) affirmation of ocular
The boosted sorter of training cascade: use 5000 positive routine photo and 10000 counter-example photo, get positive and negative sample; Utilize the Haartraining training aids, be used to generate the xml autoexec of eyes classification; Utilize the xml file that obtains, in the video flowing that camera is taken, can accurately find and the corresponding to eyes sequence of sample characteristics, and this rectangular area is set to area-of-interest;
2) the stable seizure of area-of-interest
Eye areas sequence first address in that first frame picture is obtained is scheduled to the round center of circle as one, and the eye areas sequence first address of next catching and the distance in this center of circle do not exceed predetermined radius of a circle, i.e. ((X0-X1)
2+ (Y0-Y1)
2)
1/
2<=R0, we do not change home position, and use this center of circle as eye areas sequence first address always, up to the address that occurs exceeding predetermined circle being arranged just to the central coordinate of circle renewal, can effectively eliminate jitter phenomenon;
3) area-of-interest Filtering Processing
Utilize gaussian kernel function (k (|| x-xc||)=exp{-||x-xc||^2/ (2* σ) ^2) }) that (param1*param2) carried out convolution to reduce image information content; When the sigma parameter was set to 9*9, it is actual that filter effect meets the cornea location, so just realized the The disposal of gentle filter to image;
4) rim detection
Through after the filtering, the area-of-interest of each frame of this video flowing is carried out rim detection, the x and the y direction of image are asked first order derivative; Be combined as the derivative of 4 directions then; The point that the derivative of these directions reaches local maximum is exactly a candidate point of forming the edge, in order to make image edge information not lose, further searches and draw complete eye contour; From bianry image, seek profile through cvFindContours () function, edge at this moment is the border between the positive and negative.So just realized match to imperfect profile;
5) cornea location
Utilize Hough circle transformation cvHoughCircle () function, find round sequence in the binary map after CANNY detects; To each non-zero points of gained bianry image after the rim detection, consider its partial gradient, i.e. the gradient that obtains of Sobel first order derivative through cvSobel () function calculation x and y direction; Utilize this gradient, have each point on the straight line of slope appointment all in totalizer, to add up, slope is the peaked distance from the minimum value of an appointment to appointment here; Simultaneously, each non-0 positions of elements in the marker edge image; From these points that add up thinking highly of, select the candidate center then, these centers are all greater than given threshold value and greater than its all neighbours; Next all non-0 elements are considered at each center; These elements are according to the distance ordering at itself and center; The minor increment of maximum radius is counted from the road, and a radius selecting non-0 element to support most is if center receives the support of the fullest of non-0 pixel of edge image; And to the early stage selecteed center enough distances are arranged, it will be retained; At last, utilize the coordinate at this center and the radius of confirming not being drawn and the corresponding to circle of cornea and the center of circle, thereby realize accurate location eye cornea by each two field picture of area-of-interest (ROI) video flowing of filtering and CANNY rim detection.
Advantage of the present invention and effect are: the developing instrument OpenCV that the present invention adopted is a cross-platform computer vision storehouse, on Linux, Windows and Mac OS operating system, can move.Constitute by a series of C functions and a small amount of C++ class, the interface of language such as Python, Ruby, MATLAB is provided simultaneously, realized a lot of general-purpose algorithms of Flame Image Process and computer vision aspect.The visual processes algorithm that OpenCV provides is very abundant; And its part is with the C language; Add the characteristic that it is increased income, handle proper, need not add new outside support also can be complete compiling link generate executive routine; So much human is done the transplanting of algorithm with it, the code of OpenCV can operate in dsp system and the SCM system through suitable rewriting normally.
Therefore, the present invention proposes with regard to the method improving eye detection and follow the tracks of, makes every effort to lower cost simpler equipment and human eye is followed the tracks of more accurately and located.
Description of drawings
Fig. 1 is common eyeball picture;
Fig. 2 is after smoothing processing and the rim detection and the design sketch of inverse;
Fig. 3 is a circle Sequence Detection design sketch behind the inverse; Among the figure: detect the also round sequence 1 of match; The center of circle 2 of circle sequence.
Embodiment
Following specific embodiments of the invention is done further detailed explanation, but embodiment of the present invention is not limited thereto.
Like Fig. 1, Fig. 2, shown in Figure 3.The method of cornea of the present invention location is used the CV pattern-recognition solution among the opencv, and Flame Image Process is carried out in identification eyeball zone.
Under Visual C++6.0 platform, at first utilize the CV pattern-recognition solution among the opencv, training obtains the eyes sorter, generates relevant XML file; Eye rectangular area that camera captures is set as area-of-interest (ROI); Then the ROI image is carried out smoothing processing, rim detection; In the edge image that obtains, utilize cornea to have tangible rounded form characteristic, carry out Hough circle transformation, identify circle sequence in this zone, judge that promptly this sequence is the cornea region, its position, center of circle is pupil.Can obtain the position of eyes cornea accurately and reliably through above-mentioned processing.
Below in conjunction with Fig. 1, Fig. 2, Fig. 3, specify the method step of cornea location:
1) affirmation of ocular
The boosted sorter of training cascade:
Use about 5000 positive routine photo and about 10000 counter-example photo, get positive and negative sample; Utilize the Haartraining training aids, be used to generate the xml autoexec of eyes classification; Utilize the xml file that obtains, in the video flowing that camera is taken, can accurately find and the corresponding to eyes sequence of sample characteristics, and this rectangular area is set to area-of-interest (ROI).
2) the stable seizure of area-of-interest (ROI)
Find in the research; The randomized jitter phenomenon appears in the area-of-interest that captures; This is because when each two field picture of video flowing carried out human eye detection, the eye areas sequence first address that returns each open all different, therefore when calling these first addresss and carrying out the video flowing reorganization as the picture space owner pointer; Will cause the video flowing of area-of-interest (ROI) unstable, thereby influence the accurate location of eye cornea.
In order to eliminate shake.We are scheduled to the round center of circle at the eye areas sequence first address that first frame picture is obtained as one, so as long as the eye areas sequence first address of next catching and the distance in this center of circle do not exceed predetermined radius of a circle, i.e. ((X0-X1)
2+ (Y0-Y1)
2)
1/
2<=R0, we do not change home position, and use this center of circle as eye areas sequence first address always, up to the address that occurs exceeding predetermined circle being arranged just to the central coordinate of circle renewal, can effectively eliminate jitter phenomenon.
3) area-of-interest (ROI) Filtering Processing
Usually in the image that camera captures, particularly people's face has various spin offs; Some skins are uneven, the number of people is different, can occur approaching with the round speckle regions in many places after detecting on the edge of, and these noises all can have influence on the detection to the eye cornea circle; And find under study for action; Utilize gaussian kernel function (k (|| x-xc||)=exp{-||x-xc||^2/ (2* σ) ^2) }) that (param1*param2) carried out convolution to reduce image information content, research shows, when the sigma parameter is set to 9*9; It is actual that filter effect meets the cornea location most, so just realized the The disposal of gentle filter to image.
4) rim detection
Through after the filtering; Area-of-interest (ROI) to each frame of this video flowing carries out rim detection; X and y direction to image are asked first order derivative, are combined as the derivative of 4 directions then, and the point that the derivative of these directions reaches local maximum is exactly a candidate point of forming the edge; In order to make image edge information not lose; We are necessary further to search and draw complete eye contour, from bianry image, seek profile through cvFindContours () function, and edge at this moment is the border between the positive and negative.So just realized match to imperfect profile.
5) cornea location
Utilize Hough circle transformation cvHoughCircle () function, find round sequence in the binary map after CANNY detects; To each non-zero points of gained bianry image after the rim detection, consider its partial gradient, i.e. the gradient that obtains of Sobel first order derivative through cvSobel () function calculation x and y direction; Utilize this gradient, have each point on the straight line of slope appointment all in totalizer, to add up, slope is the peaked distance from the minimum value of an appointment to appointment here; Simultaneously, each non-0 positions of elements in the marker edge image; From these points that add up thinking highly of, select the candidate center then, these centers are all greater than given threshold value and greater than its all neighbours; Next all non-0 elements are considered at each center; These elements are according to the distance ordering at itself and center; The minor increment of maximum radius is counted from the road, and a radius selecting non-0 element to support most is if center receives the support of the fullest of non-0 pixel of edge image; And to the early stage selecteed center enough distances are arranged, it will be retained; At last, utilize the coordinate at this center and the radius of confirming not being drawn and the corresponding to circle of cornea and the center of circle, thereby successfully realize accurate location eye cornea by each two field picture of area-of-interest (ROI) video flowing of filtering and CANNY rim detection.
As stated, just can realize the present invention preferably.The foregoing description is merely preferred embodiment of the present invention, is not to be used for limiting practical range of the present invention; Be that all equalizations of doing according to content of the present invention change and modification, all contained by claim of the present invention scope required for protection.
Claims (3)
1. the method for a cornea location is characterized in that: use the CV pattern-recognition solution among the opencv, discern the eyeball zone, carry out Flame Image Process.
2. the method for cornea according to claim 1 location is characterized in that: under visual c++ 6.0 platforms, at first utilize the CV pattern-recognition solution among the opencv, training obtains the eyes sorter, generates relevant XML file; Eye rectangular area that camera captures is set as area-of-interest; Then the region of interest area image is carried out smoothing processing, rim detection; In the edge image that obtains, utilize cornea to have tangible rounded form characteristic, carry out Hough circle transformation, identify circle sequence in this zone, judge that promptly this sequence is the cornea region, its position, center of circle is pupil; Obtain the accurate position of eyes cornea through above-mentioned processing.
3. the method for cornea according to claim 2 location is characterized in that comprising the steps:
1) affirmation of ocular
The boosted sorter of training cascade: use 5000 positive routine photo and 10000 counter-example photo, get positive and negative sample; Utilize the Haartraining training aids, be used to generate the xml autoexec of eyes classification; Utilize the xml file that obtains, in the video flowing that camera is taken, can accurately find and the corresponding to eyes sequence of sample characteristics, and this rectangular area is set to area-of-interest;
2) the stable seizure of area-of-interest
Eye areas sequence first address in that first frame picture is obtained is scheduled to the round center of circle as one, and the eye areas sequence first address of next catching and the distance in this center of circle do not exceed predetermined radius of a circle, i.e. ((X0-X1)
2+ (Y0-Y1)
2)
1/
2<=R0, we do not change home position, and use this center of circle as eye areas sequence first address always, up to the address that occurs exceeding predetermined circle being arranged just to the central coordinate of circle renewal, can effectively eliminate jitter phenomenon;
3) area-of-interest Filtering Processing
Utilize gaussian kernel function (k (|| x-xc||)=exp{-||x-xc||^2/ (2* σ) ^2) }) that (param1*param2) carried out convolution to reduce image information content; When the sigma parameter was set to 9*9, it is actual that filter effect meets the cornea location, so just realized the The disposal of gentle filter to image;
4) rim detection
Through after the filtering, the area-of-interest of each frame of this video flowing is carried out rim detection, the x and the y direction of image are asked first order derivative; Be combined as the derivative of 4 directions then; The point that the derivative of these directions reaches local maximum is exactly a candidate point of forming the edge, in order to make image edge information not lose, further searches and draw complete eye contour; From bianry image, seek profile through cvFindContours () function, edge at this moment is the border between the positive and negative.So just realized match to imperfect profile;
5) cornea location
Utilize Hough circle transformation cvHoughCircle () function, find round sequence in the binary map after CANNY detects; To each non-zero points of gained bianry image after the rim detection, consider its partial gradient, i.e. the gradient that obtains of Sobel first order derivative through cvSobel () function calculation x and y direction; Utilize this gradient, have each point on the straight line of slope appointment all in totalizer, to add up, slope is the peaked distance from the minimum value of an appointment to appointment here; Simultaneously, each non-0 positions of elements in the marker edge image; From these points that add up thinking highly of, select the candidate center then, these centers are all greater than given threshold value and greater than its all neighbours; Next all non-0 elements are considered at each center; These elements are according to the distance ordering at itself and center; The minor increment of maximum radius is counted from the road, and a radius selecting non-0 element to support most is if center receives the support of the fullest of non-0 pixel of edge image; And to the early stage selecteed center enough distances are arranged, it will be retained; At last, utilize the coordinate at this center and the radius of confirming not being drawn and the corresponding to circle of cornea and the center of circle, thereby realize accurate location eye cornea by each two field picture of the region of interest domain video stream of filtering and CANNY rim detection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011104497518A CN102567737A (en) | 2011-12-28 | 2011-12-28 | Method for locating eyeball cornea |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011104497518A CN102567737A (en) | 2011-12-28 | 2011-12-28 | Method for locating eyeball cornea |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102567737A true CN102567737A (en) | 2012-07-11 |
Family
ID=46413110
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2011104497518A Pending CN102567737A (en) | 2011-12-28 | 2011-12-28 | Method for locating eyeball cornea |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102567737A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104463921A (en) * | 2015-01-09 | 2015-03-25 | 厦门美图之家科技有限公司 | Image processing method for positioning center of eyeball |
CN107122597A (en) * | 2017-04-12 | 2017-09-01 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of corneal damage intelligent diagnosis system |
WO2018108124A1 (en) * | 2016-12-15 | 2018-06-21 | 腾讯科技(深圳)有限公司 | Method and system for positioning pupil |
CN108567404A (en) * | 2017-09-15 | 2018-09-25 | 分界线(天津)网络技术有限公司 | A kind of measuring system and method for teenager's eyes pseudo-myopia part |
CN109063634A (en) * | 2018-07-29 | 2018-12-21 | 国网上海市电力公司 | Using the method for hough-circle transform identification target object region electrical symbol in power monitoring |
CN104125418B (en) * | 2013-04-29 | 2019-05-10 | 托比公司 | Energy-efficient image sensing apparatus and its operation method and eye/gaze tracking system |
CN110188767A (en) * | 2019-05-08 | 2019-08-30 | 浙江大学 | Keratonosus image sequence feature extraction and classifying method and device based on deep neural network |
CN112414552A (en) * | 2020-11-24 | 2021-02-26 | 西南交通大学 | Body temperature detection device for elevator and body temperature calculation method thereof |
CN112634221A (en) * | 2020-12-17 | 2021-04-09 | 广西壮族自治区人民医院 | Image and depth-based cornea level identification and lesion positioning method and system |
CN116777794A (en) * | 2023-08-17 | 2023-09-19 | 简阳市人民医院 | Cornea foreign body image processing method and system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080273763A1 (en) * | 2007-04-26 | 2008-11-06 | Stmicroelectronics Rousset Sas | Method and device for locating a human iris in an eye image |
CN101499129A (en) * | 2008-01-30 | 2009-08-05 | 中国科学院自动化研究所 | Distant range iris recognition system and method |
CN101916362A (en) * | 2010-05-28 | 2010-12-15 | 深圳大学 | Iris positioning method and iris identification system |
CN102129553A (en) * | 2011-03-16 | 2011-07-20 | 上海交通大学 | Method for eye detection based on single infrared light supply |
-
2011
- 2011-12-28 CN CN2011104497518A patent/CN102567737A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080273763A1 (en) * | 2007-04-26 | 2008-11-06 | Stmicroelectronics Rousset Sas | Method and device for locating a human iris in an eye image |
CN101499129A (en) * | 2008-01-30 | 2009-08-05 | 中国科学院自动化研究所 | Distant range iris recognition system and method |
CN101916362A (en) * | 2010-05-28 | 2010-12-15 | 深圳大学 | Iris positioning method and iris identification system |
CN102129553A (en) * | 2011-03-16 | 2011-07-20 | 上海交通大学 | Method for eye detection based on single infrared light supply |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104125418B (en) * | 2013-04-29 | 2019-05-10 | 托比公司 | Energy-efficient image sensing apparatus and its operation method and eye/gaze tracking system |
US10372208B2 (en) | 2013-04-29 | 2019-08-06 | Tobii Ab | Power efficient image sensing apparatus, method of operating the same and eye/gaze tracking system |
CN104463921A (en) * | 2015-01-09 | 2015-03-25 | 厦门美图之家科技有限公司 | Image processing method for positioning center of eyeball |
WO2018108124A1 (en) * | 2016-12-15 | 2018-06-21 | 腾讯科技(深圳)有限公司 | Method and system for positioning pupil |
US10878593B2 (en) | 2016-12-15 | 2020-12-29 | Tencent Technology (Shenzhen) Company Limited | Pupil localizing method and system |
CN107122597B (en) * | 2017-04-12 | 2020-04-24 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Intelligent diagnosis system for corneal damage |
CN107122597A (en) * | 2017-04-12 | 2017-09-01 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of corneal damage intelligent diagnosis system |
CN108567404A (en) * | 2017-09-15 | 2018-09-25 | 分界线(天津)网络技术有限公司 | A kind of measuring system and method for teenager's eyes pseudo-myopia part |
CN109063634A (en) * | 2018-07-29 | 2018-12-21 | 国网上海市电力公司 | Using the method for hough-circle transform identification target object region electrical symbol in power monitoring |
CN110188767A (en) * | 2019-05-08 | 2019-08-30 | 浙江大学 | Keratonosus image sequence feature extraction and classifying method and device based on deep neural network |
CN110188767B (en) * | 2019-05-08 | 2021-04-27 | 浙江大学 | Corneal disease image serialization feature extraction and classification method and device based on deep neural network |
CN112414552A (en) * | 2020-11-24 | 2021-02-26 | 西南交通大学 | Body temperature detection device for elevator and body temperature calculation method thereof |
CN112634221A (en) * | 2020-12-17 | 2021-04-09 | 广西壮族自治区人民医院 | Image and depth-based cornea level identification and lesion positioning method and system |
CN112634221B (en) * | 2020-12-17 | 2024-07-16 | 广西壮族自治区人民医院 | Cornea hierarchy identification and lesion positioning method and system based on images and depth |
CN116777794A (en) * | 2023-08-17 | 2023-09-19 | 简阳市人民医院 | Cornea foreign body image processing method and system |
CN116777794B (en) * | 2023-08-17 | 2023-11-03 | 简阳市人民医院 | Cornea foreign body image processing method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102567737A (en) | Method for locating eyeball cornea | |
Fuhl et al. | Pupilnet v2. 0: Convolutional neural networks for cpu based real time robust pupil detection | |
Santini et al. | PuReST: Robust pupil tracking for real-time pervasive eye tracking | |
Tan et al. | Towards online iris and periocular recognition under relaxed imaging constraints | |
JP4464686B2 (en) | Real-time eye detection and tracking under various light conditions | |
US11026571B2 (en) | Method for processing pupil tracking image | |
Eivazi et al. | Improving real-time CNN-based pupil detection through domain-specific data augmentation | |
CN106133752A (en) | Eye gaze is followed the tracks of | |
JP6058089B2 (en) | Blink detection method and apparatus | |
US9082000B2 (en) | Image processing device and image processing method | |
Elloumi et al. | Ocular diseases diagnosis in fundus images using a deep learning: approaches, tools and performance evaluation | |
Choi et al. | Accurate eye pupil localization using heterogeneous CNN models | |
CN102129553A (en) | Method for eye detection based on single infrared light supply | |
Noman et al. | Mobile-based eye-blink detection performance analysis on android platform | |
Hajari et al. | A review of issues and challenges in designing Iris recognition Systems for noisy imaging environment | |
Cheng et al. | A fatigue detection system with eyeglasses removal | |
CN109446935B (en) | Iris positioning method for iris recognition in long-distance traveling | |
Fathee et al. | Iris segmentation in uncooperative and unconstrained environments: state-of-the-art, datasets and future research directions | |
KR101712673B1 (en) | Apparatus and Method for Detecting Center of Pupil | |
Wiśniewska et al. | Robust eye gaze estimation | |
George et al. | Escaf: Pupil centre localization algorithm with candidate filtering | |
Mashudi et al. | A review on iris recognition in non-cooperative environment | |
Karakaya et al. | An iris segmentation algorithm based on edge orientation for off-angle iris recognition | |
Fusek et al. | Pupil localization using self-organizing migrating algorithm | |
Ponz et al. | Topography-based detection of the iris centre using multiple-resolution images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120711 |