KR20160026565A - method for 3-D eye-gage tracking - Google Patents
method for 3-D eye-gage tracking Download PDFInfo
- Publication number
- KR20160026565A KR20160026565A KR1020140115680A KR20140115680A KR20160026565A KR 20160026565 A KR20160026565 A KR 20160026565A KR 1020140115680 A KR1020140115680 A KR 1020140115680A KR 20140115680 A KR20140115680 A KR 20140115680A KR 20160026565 A KR20160026565 A KR 20160026565A
- Authority
- KR
- South Korea
- Prior art keywords
- user
- gaze
- camera
- information
- eyes
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
Abstract
Description
The present invention relates to a three-dimensional (3D) line-of-sight tracking method, and more particularly, to a three-dimensional line-of-sight tracking method using binocular eye information.
A pointing method for tracking a user's gaze in a real space and selecting an object existing within a user's viewing range is mainly applied to an interface device such as a user-electronic device for a severely disabled person.
The prior art of
Such a gaze tracking device is a method of tracking a user's gaze on a monitor, and its application field is extremely limited. It is easy to use as an interface device for a special disabled person as well as a serious disabled person, but it is more realistic and it is necessary to pursue development of application of the gaze to the actual space and application of this gaze tracking result in order to expand the application field.
The present invention provides a method capable of three-dimensional line-of-sight tracking and a system for applying the method.
Accordingly, the present invention provides a gaze tracking method with improved accuracy and a system for applying the same.
A three-dimensional line-of-sight tracking method according to the present invention:
Obtaining conversion information capable of acquiring a user's gazing position from the user's gaze information for a plurality of objects having different distances from the user;
Obtaining gaze information of both eyes of a user gazing at an object;
Obtaining binocular disparity differences for each of the individuals using the binocular visual information;
And comparing the gaze difference of each of the individuals to determine an entity representing the minimum value as an entity to which the user is looking.
A three-dimensional line-of-sight tracking system according to the present invention:
10. A line-of-sight tracking system for performing the method according to
A first camera for photographing both eyes of the user;
A second camera for photographing an object to be looked at by the user;
And an analysis system for determining the direction of the user's gaze from the image information from the first camera and the second camera.
According to an embodiment of the present invention, the entity may be an electronic product located in a space in which the user resides.
According to another embodiment of the present invention, the conversion information may include information on one entity arranged at a different distance from the user.
According to another embodiment of the present invention, the transformation information may be a calibration matrix or a transformation matrix obtained by a geometric transform.
According to another embodiment of the present invention, the gaze information of the user is acquired from a camera that photographs the user's eyes, and eye images obtained from the camera are subjected to histogram analysis-based binarization and component labeling to determine pupil center coordinates Can be obtained.
According to the present invention, it is possible to recognize objects arranged at different distances on a three-dimensional plane even if only image data in two-dimensional images are used. According to the present invention, a plurality of transformation matrices for gaze distance measurement can be obtained through calibration at a plurality of distances, and the gaze in the three-dimensional space can be accurately determined.
1 is a schematic block diagram of a three-dimensional line-of-sight tracking system according to the present invention.
FIG. 2 is a flowchart of a gaze tracking algorithm applied to a three-dimensional line-of-sight tracking method according to the present invention.
Fig. 3 is a conceptual diagram of a geometric transformation showing the mapping relationship between the pupil position and anterior image. Fig.
FIG. 4 illustrates a three-dimensional line-of-sight tracking method according to the present invention. FIG. 4 illustrates a user calibration process and a forward line of sight of the user at this time.
FIG. 5 illustrates a three-dimensional line-of-sight tracking method according to the present invention. FIG. 5 shows a method of predicting a target object (object) using the difference between the left and right eye lines.
6 (a), 6 (b) and 6 (c) show the difference of the binocular vision for each individual when a transformation matrix is applied to each of the individuals located at different distances from the user.
FIG. 7 schematically shows the entire flow of a three-dimensional line-of-sight tracking method according to the present invention.
FIG. 8 is a photograph of an experimental procedure of a three-dimensional line-of-sight tracking method according to the present invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of a gaze tracking method and a system to which the present invention is applied will be described with reference to the accompanying drawings.
1 shows an arrangement of a
The
According to another embodiment of the present invention, the illumination lamp of the
In addition, according to the present invention, the
The eye tracking method according to the present invention and a system to which the present invention is applied can be applied to a conventional method of detecting circular edge detection, local binarization, component labeling and region filling, And performs a geometric transform to detect the position and eye line of the pupil.
First, non-patent document 5 can be referred to for understanding the basic gaze tracking method, which will be briefly described below.
FIG. 2 shows a flowchart of a gaze tracking algorithm applied to the present invention. After acquiring the eye image (21), the center of the pupil is extracted using the algorithm described later (22). After the extraction of the pupil center, if the initial calibration is not performed and it is determined that calibration is necessary (23), calibration (25) of the following process is performed. If it is determined that calibration has been performed (23), the position in the front image region corresponding to the detected position of the center of the pupil is calculated and the
Specifically, the whole process is as follows. A local binarization, a component labeling and a region filling method are performed to find the center of the pupil in the eye image acquired by the
The circular detection algorithm determines the initial pupil region through circular template matching. (Refer to non-patent document [1]). Since the detected pupil can be displayed in an elliptical shape instead of a circle depending on the gaze position and the camera photographing angle, the position determined by the circular template matching method is not accurate. Therefore, a local binarization process is performed based on the determined position. Since the rectangular region is classified into two types, ie, a pupil region (foreground) and a non-pupil region (background), the binarization threshold value is determined by Gonzalez's method (refer to Non-Patent Document 2) The proposed histogram-based binarization method (refer to non-patent document [3]) is used.
After local binarization, there may be noises due to eyebrows or shadows, and if the reflected light is present inside the pupil region, it may appear as an opening. In order to solve this problem, a method of labeling a binarized region by applying a method of labeling to regions adjacent to each other, removing an area having the largest area and having another identity, Remove the noise area. Finally, after performing a morphological closing operation to fill the perforated region, the center of gravity of the filled region is determined to be the final pupil center.
In the calibration, as shown in FIG. 2, four vertexes defined in the front image, that is, the upper left (Px 1 , Py 1 ), the upper right (Px 2 , Py 2 ), lower right (Px 3 , Py 3 ), and lower left (Px 4 , Py 4 ). Since the
In the above equation, M = T × C, which can be expressed as T = M × C -1 . According to this, the transformation matrix T by the constants a to h as an unknown matrix can be obtained from the inverse of the matrix C for the matrix M. [ That is, the elements of the matrix C are located at the four positions of the pupil when viewing the four vertices defined in the forward image obtained from the image obtained from the first camera, that is, the upper left (Px 1 , Py 1 ) 2 , Py 2 ), the lower right (Px 3 , Py 3 ) and the lower left (Px 4 , Py 4 ), and the elements of the matrix M are composed of four vertices Mx 1 , My 1 ), (Mx 2 , My 2 ), (Mx 3 , My 3 ), and (Mx 4 , My 4 ).
The coordinates (Mxc, Myc) for the object can be obtained from the user's coincidence coordinates (Pxc, Pyc) by the following equation (4) using the transformation matrix T obtained by the above equation.
An example of obtaining the transformation matrix according to the above equation is as follows.
For example, if the coordinates of the four corners of the object obtained from the first camera are {11, 28}, {-15, 28}, {-16, 19}, {14, Px2 = -15, Px3 = -16, Px4 = 14, and Py1 = 28, Py2 = 28, Py3 = -16 and Py4 = 18 when applied to the above matrix of mathematical expressions. The same matrix C is constructed.
Then, the inverse matrix (C -1 ) of the above matrix C is calculated as follows.
Assuming that the object to which the user is interested in the above inverse matrix is a 1920x1080 resolution monitor, the matrix M by the four corners of the monitor is constructed as follows.
Thus, the transformation matrix T (Calibration Matrix, CM) is obtained by multiplying the inverse matrix of C by the matrix M, and the solution is as follows.
The transform matrix may be obtained for all entities located at a specific distance from the user and may be obtained by multiplying the matrix of eye coordinates at the time the user looks at the entity, do.
For example, if the coordinates of the eye of the user gazing at the object are {5, 27}, the matrix E consisting of eye coordinates is as follows.
Therefore, from the product of the matrix E and T, the following entity can obtain the gaze coordinate matrix P:
According to the above results, the user can judge that the coordinates (pixel) of {456, 111} are stared at the monitor having the resolution of 1920X1080.
The computation of these gaze coordinates is obtained for each eye, and both coordinates may be coincident or spaced apart by some distance depending on the distance between the user and the object and the conditions of both eyes of the user. The distance of the distance is referred to as the difference of the line of sight in the present invention, and it is possible to obtain the gaze coordinates of the user's gaze object and the corresponding object with respect to various objects having different distances by comparing the sizes of the gaze lines.
The gaze tracking by the above method can accurately track the direction of the user's gaze when the distance between the user and the subject remains the same as the distance at the time of calibration. However, when the distance between the user and the object changes, the accuracy of the eye tracking is degraded, and the tracking may fail. In the present invention, a transformation matrix for a plurality of distances is obtained in consideration of a change in distance between a user and an object, and the distance (Z) of the line of sight as well as the line X-Y coordinate is obtained.
In order to prevent the failure of eye tracking due to the movement of the user, two or more calibration results (expressions) are obtained while varying the distance between the user and the object, and using the calibration results, Though it will track or calculate accurate gaze.
Unlike the prior art, the present invention obtains a plurality of transformation matrices having different distances, and this transformation matrix is calculated for both eyes.
The present invention tracks the user's gaze direction and distance using the difference between the left eye and the right eye absolute values of the user's eyes, that is, the specific position. The present invention performs a calibration as described above for an object located at a certain distance to obtain a transformation matrix T, wherein the calibration is performed for both eyes. The calibration is performed for a plurality of distances different in distance from the user and the object, and a plurality of calibration matrices (matrices) obtained for each distance, that is, a plurality of transformation matrices, are used to track the gaze direction and the distance. By multiplying the plurality of transformation matrices obtained by the calibration by the coordinates of the pupil of both eyes, pupil coordinates for a different object, for example, a specific portion of the monitor, are obtained. The differences of the pupil coordinates thus obtained are obtained. The difference in pupil coordinates includes the difference in the X-X '(left and right) direction and the difference in the Y-Y' (up and down) direction, respectively.
Table 1 above illustrates the eye-gaze (X-Y) coordinates of the user staring at an arbitrary object from the arbitrary transformation matrix obtained at a Z distance and the gaze difference of both eyes converted from the coordinates. Here, the arbitrary plurality of transformation matrices obtained at a plurality of Z distances are applied to each of nine eye lines gazed at one Z distance, and when the average of the difference (absolute value) is the smallest, Z distance ".
Figure 3 illustrates a calibration method for multiple entities.
The user takes an image of both
Based on the acquired eye images, the pupil coordinates for the four reference points (edges) of the
These entities use the four corner coordinates extracted from the forward image and the pupil (reference point) coordinates extracted from the eye image to obtain a transformation matrix by the geometric transformation by the above-described equations. Here, the transformation matrix corresponds to the number of entities, and thus three transformation matrices CM1, CM2, CM3 are obtained as shown in Fig.
(CM1, CM2, CM3), and obtains a binocular image when an actual user looks at an object, and obtains pupil coordinates (Pxc, (Pxc ', Pyc') and multiplying the matrix of these coordinates by the three transformation matrices T (CM1, CM2, CM3) to obtain three coordinates Mxc, Myc) ((Mxc ', Myc').
5 shows the gaze difference for the object-1 and the object-2 when the
Since the
Likewise, when the
Based on this principle, a first transformation matrix CM1 is applied to the entity-1 31 and a second transformation matrix CM2 is applied to the entity- 32) of the eyes of both eyes. At this time, the object with the smallest gaze difference is the object that the user looks at.
The calculated binocular eye difference value has no significance in the presence of a sign because both determine only the length or the distance, and therefore the absolute value of the binocular eye difference value is applied.
In FIG. 6, three
6 (a) shows a gathering of eyes and a widening when gazing at the object-1 31 having the first transformation matrix CM1. That is, in the
6 (b) shows the gaze of both eyes flaring in the front and
FIG. 6C shows a group of eye lines when gazing at the object-3 33 having the third transformation matrix CM3. That is, in the
The following table illustrates the binocular vision difference for three individuals (31, 32, 33) located at a distance (Z = 80 cm, 210 cm, 340 cm) in the Z direction.
In Table 1 above, the conversion matrix for
The present invention for tracking the Z position of an object to which a line of sight is focused by the above method, particularly the distance from the user, can be arranged as shown in FIG.
In the process of FIG. 7, a conversion matrix (CM1, CM2, CM3) for a plurality of distances is prepared. The number of distance-specific transformation matrices can be increased according to the required accuracy, and the measurement accuracy can be increased more finely.
S71: The calculation of both eyes is started when the transformation matrix for several Z distances is ready.
S72: When the user watches any one of several objects arranged at an arbitrary Z distance, a binocular image at this time is acquired using the first camera.
S73: Obtaining the forward image using the second camera facing the entities arranged in the Z direction corresponding to the binocular image acquisition.
S74: The coordinates of both eyes and the reference coordinates of the forward image are obtained through the procedure described above, and the gaze of both eyes is calculated by using this.
S75: The difference between the lines of sight of both eyes calculated using a plurality of transformation matrices is calculated, and its absolute value is taken.
S76: The object of arbitrary distance that the absolute value of the gaze difference of both eyes is the smallest is judged as the object to which the current user strikes.
8 is a photograph of a scene in which the method of the present invention is being experimented. In front of the user, there is a camera for photographing the face of the user in front of the user and a camera for photographing the screen of the monitor as an object beyond the user's front. The white point on the monitor is used as a reference point to show the eye coordinates and the gaze difference of the eyes according to the change of the user's gaze.
In the embodiment of the present invention as described above, it is necessary to individually calibrate individual entities existing in the space in which the user resides, that is, to calculate a transformation matrix for each entity, and the transformation matrix obtained therefrom, And the like.
According to the present invention, since the distances between the objects can be estimated by the user, precise pointing can be performed on the objects installed at different distances in the space where the user resides, and in particular, The position or distance of the gazing position can be tracked, so that the gazing object can be accurately pointed without any crosstalk.
Also, by preparing a plurality of transformation matrices with different distances for one entity, the user can precisely point to one entity at various positions. If a plurality of transformation matrices are provided for each object as described above, the range of the user's fidelity is widened, and thus the object can be pointed more conveniently.
In the foregoing, exemplary embodiments have been described and shown in the accompanying drawings to facilitate understanding of the present invention. It should be understood, however, that such embodiments are merely illustrative of the present invention and not limiting thereof. And it is to be understood that the invention is not limited to the details shown and described. Since various other modifications may occur to those of ordinary skill in the art.
10: User
11a and 11b: the eyes of the user
20: Camera device
30, 31, 32, 33: object
40: Analysis system
Claims (10)
Obtaining gaze information of both eyes of a user gazing at an object;
Obtaining binocular disparity differences for each of the individuals using the binocular visual information;
And comparing the gaze difference of each of the individuals to determine an object representing the minimum value as an object to which the user is to gaze.
Wherein the object is an electronic product located in a space in which the user resides.
Wherein the conversion information includes information on a single object arranged at a different distance from the user.
Wherein the transformation information is a calibration matrix or a transformation matrix obtained by a geometric transform.
The gaze information of the user is acquired from a camera that photographs the user's eyes,
Wherein eye center coordinates are obtained by binarization based on histogram analysis and component labeling of eye images acquired from the camera.
A first camera for photographing both eyes of the user;
A second camera for photographing an object to be looked at by the user;
And an analysis system for determining the direction of the user's gaze from the image information from the first camera and the second camera.
Wherein the object is an electronic product located in a space in which the user resides.
Wherein the conversion information includes information on one entity arranged at a different distance from the user.
Wherein the transformation information is obtained by the analysis system and is a calibration matrix or transformation matrix obtained by a geometric transform.
The gaze information of the user is acquired from a camera that photographs the user's eyes,
Wherein the analysis system obtains pupil center coordinates by binarization based on histogram analysis and component labeling of eye images acquired from a camera.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020140115680A KR20160026565A (en) | 2014-09-01 | 2014-09-01 | method for 3-D eye-gage tracking |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020140115680A KR20160026565A (en) | 2014-09-01 | 2014-09-01 | method for 3-D eye-gage tracking |
Publications (1)
Publication Number | Publication Date |
---|---|
KR20160026565A true KR20160026565A (en) | 2016-03-09 |
Family
ID=55536928
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
KR1020140115680A KR20160026565A (en) | 2014-09-01 | 2014-09-01 | method for 3-D eye-gage tracking |
Country Status (1)
Country | Link |
---|---|
KR (1) | KR20160026565A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11216974B2 (en) | 2017-12-14 | 2022-01-04 | Samsung Electronics Co., Ltd. | Staring distance determination method and device |
US11249305B2 (en) | 2019-04-11 | 2022-02-15 | Samsung Electronics Co., Ltd. | Head-mounted display device and operating method of the same for determining a measurement parameter |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101343875B1 (en) | 2011-12-06 | 2013-12-23 | 경북대학교 산학협력단 | Analysis device of user cognition and method for analysis of user cognition |
-
2014
- 2014-09-01 KR KR1020140115680A patent/KR20160026565A/en not_active Application Discontinuation
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101343875B1 (en) | 2011-12-06 | 2013-12-23 | 경북대학교 산학협력단 | Analysis device of user cognition and method for analysis of user cognition |
Non-Patent Citations (5)
Title |
---|
Cheol Woo Cho, Ji Woo Lee, Eui Chul Lee, Kang Ryoung Park, "A Robust Gaze Tracking Method by Using Frontal Viewing and Eye Tracking Cameras", Optical Engineering, Vol. 48, No. 12, 127202, Dec. 2009. |
R. C. Gonzalez and R. E. Woods, Digital Image Processing, 2nd ed., Prentice-Hall, Englewood Cliffs, NJ, 2002. |
W. Doyle, Operation useful for similarity-invariant pattern recognition, J. Assoc. Comput. Mach, 1962, 9: 259~267 |
Y. J. Ko, E. C. Lee, and K. R. Park, "A robust gaze detection method by compensating for facial movements based on corneal specularities," Pattern Recogn. Lett. 29, 10, 1474?1485 2008. |
허환, 이지우, 이원오, 이의철, 박강령, '장애인을 위한 새로운 감성 인터페이스 연구', 2011.2., 대한인간공학회지 제30권 제1호 pp229-23 |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11216974B2 (en) | 2017-12-14 | 2022-01-04 | Samsung Electronics Co., Ltd. | Staring distance determination method and device |
US11249305B2 (en) | 2019-04-11 | 2022-02-15 | Samsung Electronics Co., Ltd. | Head-mounted display device and operating method of the same for determining a measurement parameter |
US11526004B2 (en) | 2019-04-11 | 2022-12-13 | Samsung Electronics Co., Ltd. | Head-mounted display device and operating method of the same |
US11809623B2 (en) | 2019-04-11 | 2023-11-07 | Samsung Electronics Co., Ltd. | Head-mounted display device and operating method of the same |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR102212209B1 (en) | Method, apparatus and computer readable recording medium for eye gaze tracking | |
CN107688391B (en) | Gesture recognition method and device based on monocular vision | |
US10007336B2 (en) | Apparatus, system, and method for mobile, low-cost headset for 3D point of gaze estimation | |
CN107004275B (en) | Method and system for determining spatial coordinates of a 3D reconstruction of at least a part of a physical object | |
JP6295645B2 (en) | Object detection method and object detection apparatus | |
CN104380338B (en) | Information processor and information processing method | |
JP6417702B2 (en) | Image processing apparatus, image processing method, and image processing program | |
KR101486543B1 (en) | Method and apparatus for recognition and segmentation object for 3d object recognition | |
US9659408B2 (en) | Mesh reconstruction from heterogeneous sources of data | |
KR101471488B1 (en) | Device and Method for Tracking Sight Line | |
EP3243162A1 (en) | Gaze detection offset for gaze tracking models | |
KR101255219B1 (en) | Method of eye-gaze tracking and system adopting the method | |
CN104834901A (en) | Binocular stereo vision-based human face detection method, device and system | |
Cho et al. | Gaze Detection by Wearable Eye‐Tracking and NIR LED‐Based Head‐Tracking Device Based on SVR | |
Chansri et al. | Hand gesture recognition for Thai sign language in complex background using fusion of depth and color video | |
CA3207099A1 (en) | Methods and systems for registering preoperative image data to intraoperative image data of a scene, such as a surgical scene | |
Benalcazar et al. | A 3D iris scanner from multiple 2D visible light images | |
KR101226668B1 (en) | 3 Dimensional Motion Recognition System and Method Using Stereo Camera | |
JP5336325B2 (en) | Image processing method | |
KR20160026565A (en) | method for 3-D eye-gage tracking | |
Heo et al. | Object recognition and selection method by gaze tracking and SURF algorithm | |
EP3074844A1 (en) | Estimating gaze from un-calibrated eye measurement points | |
KR101868520B1 (en) | Method for hand-gesture recognition and apparatus thereof | |
JP2012003724A (en) | Three-dimensional fingertip position detection method, three-dimensional fingertip position detector and program | |
KR101581586B1 (en) | Compensation method for noise of depth image |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
A201 | Request for examination | ||
E902 | Notification of reason for refusal | ||
E601 | Decision to refuse application |