A method and device for determining eyeball posture position
By segmenting the eyeball image, edge detection and attitude calculation, combined with the detection model, the problem of low accuracy in the determination of eyeball posture position in the prior art is solved, and the accuracy of attitude position determination is improved.
Patent Information
- Application Number
- CN202411026365.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-07-30
AI Technical Summary
In the prior art, the eyeball posture position is determined by setting up a special detection sensor and camera equipment to jointly identify the acquired images in a single way, resulting in a reduction in data accuracy.
By segmenting the eyeball images corresponding to the target eyeball of the person to be tested, edge detection is performed to determine the contour position, and combining the spindle rotation calculation and ellipse fitting calculation, combined with the pose prediction results of the eyeball image by the detection model, the eyeball posture position is comprehensively determined.
It improves the data accuracy of eyeball posture position determination, enhances the analysis ability in multiple scenarios, and avoids the risk of incorrect position analysis.
Smart Images

Figure CN118552608B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of image detection technology, and in particular to a method and device for determining the position of an eyeball posture. Background Art
[0002] The application of eye posture and position in various fields is very important. For example, in the construction of virtual reality scenes, the changes in the body position of the human body to be constructed may cause the virtual pupil center of the human body to be constructed to shift and the virtual eyeball to rotate, thus affecting the accuracy of scene construction. In addition, detecting the eye posture and position helps to timely discover and correct abnormal movement or position deviation of the eye image when analyzing images in various scenes, avoiding the risks caused by incorrect eye position analysis.
[0003] At present, the determination of the eyeball posture position is mainly carried out by setting up special detection sensors and camera equipment to jointly perform a single recognition method on the collected images, which reduces the data accuracy of the eyeball posture position determination. Summary of the invention
[0004] In view of this, the purpose of the present application is to provide a method and device for determining the eyeball posture position, by performing edge detection on the eyeball sub-image segmented from the eyeball image corresponding to the target eyeball of the person to be tested, to determine the contour position of the target eyeball, and performing principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image according to the contour position to determine the eyeball posture, combining the contour position and the eyeball posture with the posture prediction result of the eyeball image by the detection model, to comprehensively determine the eyeball posture position of the target eyeball, thereby improving the data accuracy of the eyeball posture position determination.
[0005] The present application provides a method for determining an eyeball posture position, the method comprising:
[0006] Segmenting the acquired eyeball image corresponding to the target eyeball of the person to be tested to obtain eyeball sub-images segmented from the eyeball image;
[0007] Performing edge detection on the eyeball sub-image to determine contour position data corresponding to the target eyeball;
[0008] Based on the contour position data, performing a principal axis rotation calculation and an ellipse fitting calculation on the eyeball sub-image respectively to obtain eyeball posture data corresponding to the target eyeball;
[0009] Based on the contour position data and the eyeball posture data, as well as the posture prediction data output by a preset detection model according to the eyeball image, the eyeball posture position of the target eyeball is determined.
[0010] Furthermore, the acquired eyeball image corresponding to the target eyeball of the person to be tested is segmented to obtain eyeball sub-images segmented from the eyeball image, including:
[0011] Obtaining an eyeball image corresponding to a target eyeball of the person to be tested captured by a depth camera device;
[0012] Performing image window segmentation on the eyeball image to obtain a plurality of sub-window images corresponding to the eyeball image;
[0013] Using a preset edge detector to perform iris contour segmentation on the eyeball image to obtain an iris contour image corresponding to the eyeball image;
[0014] The multiple sub-window images and the iris contour image are determined as the eyeball sub-image.
[0015] Furthermore, based on the contour position data, the principal axis rotation calculation and the ellipse fitting calculation are respectively performed on the eyeball sub-image to obtain the eyeball posture data corresponding to the target eyeball, including:
[0016] Based on the contour position data, a principal axis rotation calculation is performed on the plurality of sub-window images in the eyeball sub-image to obtain first eyeball posture data corresponding to the target eyeball;
[0017] Based on the contour position data, an ellipse fitting calculation is performed on the iris contour image in the eyeball sub-image to obtain second eyeball posture data corresponding to the target eyeball.
[0018] Further, based on the contour position data, the principal axis rotation calculation is performed on the multiple sub-window images in the eyeball sub-image to obtain the first eyeball posture data corresponding to the target eyeball, including:
[0019] Based on the contour position data, performing a principal axis rotation calculation on the plurality of sub-window images in the eyeball sub-image to determine a first rotation angle corresponding to the target eyeball;
[0020] Determine a plurality of pixel point coordinates and a window center point coordinate corresponding to each sub-window image in the plurality of sub-window images;
[0021] Based on the first rotation angle, the coordinates of the plurality of pixel points and the coordinates of the center point of the window, a preset center angle calculation formula is used to determine the first center position coordinates corresponding to the target eyeball;
[0022] The first rotation angle and the first center position coordinates are determined as first eyeball posture data corresponding to the target eyeball.
[0023] Further, based on the contour position data, performing a principal axis rotation calculation on the plurality of sub-window images in the eyeball sub-image to determine a first rotation angle corresponding to the target eyeball includes:
[0024] Decentralizing the contour position data to obtain a contour point set matrix;
[0025] Based on the contour point set matrix, a plurality of eigenvectors corresponding to the contour point set matrix are determined using a preset covariance matrix calculation formula;
[0026] Based on the number of eigenvalues respectively corresponding to the plurality of eigenvectors, determining a eigenvector whose eigenvalue number represents a maximum value as a first eigenvector indicating the main axis direction corresponding to the target eyeball;
[0027] Based on the first feature vector, a first rotation angle corresponding to the target eyeball is determined using a preset rotation angle calculation formula.
[0028] Further, the performing of ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to obtain second eyeball posture data corresponding to the target eyeball includes:
[0029] Based on the contour position data, performing an ellipse fitting calculation on the iris contour image in the eyeball sub-image to determine a second rotation angle corresponding to the target eyeball;
[0030] Performing coordinate mean calculation on the iris contour image to obtain the second center position coordinates corresponding to the target eyeball;
[0031] The second rotation angle and the second center position coordinates are determined as second eyeball posture data corresponding to the target eyeball.
[0032] Further, the performing ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to determine the second rotation angle corresponding to the target eyeball includes:
[0033] Performing partial derivative convolution processing on the contour position data to obtain an elliptic curve relationship equation for fitting the target eyeball in the iris contour image;
[0034] Based on the focal length parameter corresponding to the depth camera device and the elliptic curve relationship, an elliptical focal length matrix is established;
[0035] Based on the elliptical focal length matrix, a second eigenvector corresponding to the target eyeball is determined using a preset diagonalization calculation formula;
[0036] Based on the eigenvalue corresponding to the second eigenvector and the set normal vector parameters, a preset normal vector calculation formula is used to determine the normal vector corresponding to the target eyeball;
[0037] Based on the direction parameter corresponding to the normal vector, a rotation angle of the direction parameter relative to the iris contour image is determined as a second rotation angle corresponding to the target eyeball.
[0038] The embodiment of the present application also provides a device for determining the position of an eyeball posture, the device comprising:
[0039] An image segmentation module is used to segment the acquired eyeball image corresponding to the target eyeball of the person to be tested, and obtain eyeball sub-images segmented from the eyeball image;
[0040] A contour determination module, used for performing edge detection on the eyeball sub-image to determine contour position data corresponding to the target eyeball;
[0041] A posture calculation module, used to perform principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image based on the contour position data, so as to obtain eyeball posture data corresponding to the target eyeball;
[0042] A posture determination module is used to determine the eye posture position of the target eye based on the contour position data and the eye posture data, as well as the posture prediction data output by a preset detection model according to the eye image.
[0043] Furthermore, when the image segmentation module is used to segment the acquired eyeball image corresponding to the target eyeball of the person to be tested to obtain the eyeball sub-image segmented from the eyeball image, the image segmentation module is used to:
[0044] Obtaining an eyeball image corresponding to a target eyeball of the person to be tested captured by a depth camera device;
[0045] Performing image window segmentation on the eyeball image to obtain a plurality of sub-window images corresponding to the eyeball image;
[0046] Using a preset edge detector to perform iris contour segmentation on the eyeball image to obtain an iris contour image corresponding to the eyeball image;
[0047] The multiple sub-window images and the iris contour image are determined as the eyeball sub-image.
[0048] Further, when the posture calculation module is used to perform principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image based on the contour position data to obtain the eyeball posture data corresponding to the target eyeball, the posture calculation module is used to:
[0049] Based on the contour position data, a principal axis rotation calculation is performed on the plurality of sub-window images in the eyeball sub-image to obtain first eyeball posture data corresponding to the target eyeball;
[0050] Based on the contour position data, an ellipse fitting calculation is performed on the iris contour image in the eyeball sub-image to obtain second eyeball posture data corresponding to the target eyeball.
[0051] Further, when the posture calculation module is used to perform a principal axis rotation calculation on the multiple sub-window images in the eyeball sub-image based on the contour position data to obtain the first eyeball posture data corresponding to the target eyeball, the posture calculation module is used to:
[0052] Based on the contour position data, performing a principal axis rotation calculation on the plurality of sub-window images in the eyeball sub-image to determine a first rotation angle corresponding to the target eyeball;
[0053] Determine a plurality of pixel point coordinates and a window center point coordinate corresponding to each sub-window image in the plurality of sub-window images;
[0054] Based on the first rotation angle, the coordinates of the plurality of pixel points and the coordinates of the center point of the window, a preset center angle calculation formula is used to determine the first center position coordinates corresponding to the target eyeball;
[0055] The first rotation angle and the first center position coordinates are determined as first eyeball posture data corresponding to the target eyeball.
[0056] Further, when the posture calculation module is used to perform a principal axis rotation calculation on the multiple sub-window images in the eyeball sub-image based on the contour position data to determine a first rotation angle corresponding to the target eyeball, the posture calculation module is used to:
[0057] Decentralizing the contour position data to obtain a contour point set matrix;
[0058] Based on the contour point set matrix, a plurality of eigenvectors corresponding to the contour point set matrix are determined using a preset covariance matrix calculation formula;
[0059] Based on the number of eigenvalues respectively corresponding to the plurality of eigenvectors, determining a eigenvector whose eigenvalue number represents a maximum value as a first eigenvector indicating the main axis direction corresponding to the target eyeball;
[0060] Based on the first feature vector, a first rotation angle corresponding to the target eyeball is determined using a preset rotation angle calculation formula.
[0061] Further, when the posture calculation module is used to perform ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to obtain the second eyeball posture data corresponding to the target eyeball, the posture calculation module is used to:
[0062] Based on the contour position data, performing an ellipse fitting calculation on the iris contour image in the eyeball sub-image to determine a second rotation angle corresponding to the target eyeball;
[0063] Performing coordinate mean calculation on the iris contour image to obtain the second center position coordinates corresponding to the target eyeball;
[0064] The second rotation angle and the second center position coordinates are determined as second eyeball posture data corresponding to the target eyeball.
[0065] Further, when the posture calculation module is used to perform ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to determine the second rotation angle corresponding to the target eyeball, the posture calculation module is used to:
[0066] Performing partial derivative convolution processing on the contour position data to obtain an elliptic curve relationship equation for fitting the target eyeball in the iris contour image;
[0067] Based on the focal length parameter corresponding to the depth camera device and the elliptic curve relationship, an elliptical focal length matrix is established;
[0068] Based on the elliptical focal length matrix, a second eigenvector corresponding to the target eyeball is determined using a preset diagonalization calculation formula;
[0069] Based on the eigenvalue corresponding to the second eigenvector and the set normal vector parameters, a preset normal vector calculation formula is used to determine the normal vector corresponding to the target eyeball;
[0070] Based on the direction parameter corresponding to the normal vector, a rotation angle of the direction parameter relative to the iris contour image is determined as a second rotation angle corresponding to the target eyeball.
[0071] An embodiment of the present application also provides an electronic device, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the method for determining the eye posture position as described above are performed.
[0072] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for determining the eyeball posture position as described above are executed.
[0073] The embodiments of the present application provide a method and device for determining the eyeball posture position, and the method comprises: segmenting an acquired eyeball image corresponding to a target eyeball of a person to be tested to obtain an eyeball sub-image segmented from the eyeball image; performing edge detection on the eyeball sub-image to determine contour position data corresponding to the target eyeball; based on the contour position data, performing principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image to obtain eyeball posture data corresponding to the target eyeball; determining the eyeball posture position of the target eyeball based on the contour position data and the eyeball posture data, as well as posture prediction data output by a preset detection model according to the eyeball image.
[0074] Compared with the prior art method of jointly performing a single recognition on the captured image by setting up a special detection sensor and a camera device, the contour position of the target eye is determined by performing edge detection on the eye sub-image segmented from the eye image corresponding to the target eye of the person to be tested, and the eye posture is determined by performing principal axis rotation calculation and ellipse fitting calculation on the eye sub-image according to the contour position. The contour position and eye posture are combined with the posture prediction result of the eye image by the detection model to comprehensively determine the eye posture position of the target eye, thereby improving the data accuracy of the eye posture position determination.
[0075] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0076] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0077] Figure 1 A flow chart of a method for determining eyeball posture position provided in an embodiment of the present application;
[0078] Figure 2 A schematic diagram of the structure of a device for determining the position of an eyeball posture provided in an embodiment of the present application;
[0079] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0080] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work belongs to the scope of protection of the present application.
[0081] Research has found that, at present, the determination of the eyeball posture position is mainly carried out by setting up special detection sensors and camera equipment to jointly perform a single recognition method on the collected images, which reduces the accuracy of the data for determining the eyeball posture position.
[0082] Based on this, an embodiment of the present application provides a method for determining the position of the eyeball posture, by performing edge detection on the eyeball sub-image segmented from the eyeball image corresponding to the target eyeball of the person to be tested, to determine the contour position of the target eyeball, and performing principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image according to the contour position to determine the eyeball posture, combining the contour position and the eyeball posture with the posture prediction result of the eyeball image by the detection model, to comprehensively determine the eyeball posture position of the target eyeball, thereby improving the data accuracy of the eyeball posture position determination.
[0083] See also Figure 1 , Figure 1 Flow chart of a method for determining eyeball posture position provided in an embodiment of the present application. Figure 1 As shown in , the method for determining the eyeball posture position provided in the embodiment of the present application includes:
[0084] S101 , segmenting the acquired eyeball image corresponding to the target eyeball of the person to be tested to obtain eyeball sub-images segmented from the eyeball image.
[0085] It should be noted that before segmenting the eyeball image, the eyeball image corresponding to the target eyeball of the person to be tested is obtained by controlling the set depth camera device to aim at the target eyeball of the person to be tested; wherein the depth camera device may include a depth color camera (RGBD camera).
[0086] In an embodiment of the present application, an eye image corresponding to the target eye of the person to be tested can be obtained in a variety of scenarios, including but not limited to virtual reality scenarios, behavior analysis scenarios, and motor ability assessment scenarios.
[0087] Specifically, when the lens direction of the depth camera device is perpendicular to the head direction of the person to be tested, the depth camera device is positioned close to the eyes of the person to be tested, and the depth camera device is controlled to face the target eyeball of the person to be tested to completely capture the target eyeball of the person to be tested; the depth camera device is controlled to perform translational movement to adjust the position and height of the depth camera device to within a predetermined range; and an eye image corresponding to the target eyeball of the person to be tested is obtained.
[0088] Here, the eyeball image corresponding to the target eyeball of the person to be tested is an RGB-D image; the size of the eyeball image is H×W×4; where H and W are the height and width of the eyeball image, respectively, and 4 represents three color (RGB) channels and a depth (D) channel.
[0089] In this step, during the specific implementation, first, an eyeball image corresponding to the target eyeball of the person to be tested captured by the depth camera device is obtained; then, the eyeball image is segmented into image windows to obtain multiple sub-window images corresponding to the eyeball image, and at the same time, the eyeball image is segmented into iris contours using a preset edge detector to obtain an iris contour image corresponding to the eyeball image; finally, the multiple sub-window images and the iris contour image are determined as eyeball sub-images.
[0090] In one implementation of the present application, during specific implementation, step S101 may include:
[0091] S1011. Obtain an eyeball image corresponding to a target eyeball of the person to be tested, which is captured by a depth camera device.
[0092] In this step, when it is specifically implemented, first, the depth camera device is controlled to aim at the target eyeball of the person to be tested; then, the depth camera device collects the eyeball image corresponding to the target eyeball of the person to be tested; finally, the collected eyeball image is obtained.
[0093] Specifically, the depth camera device is controlled to aim at the target eyeball of the person to be tested through the following steps; specifically, first, the depth camera device is controlled to face the target eyeball of the person to be tested, and the camera field of view image captured by the depth camera device is obtained; then, edge detection is performed on the camera field of view image, and the eyeball contour data determined by edge detection in the camera field of view image is extracted; thereafter, it is determined whether the eyeball contour data meets the eyeball model standard corresponding to the target eyeball of the person to be tested that is pre-calibrated; if the eyeball contour data meets the eyeball model standard, it is determined that the depth camera device is aimed at the target eyeball of the person to be tested.
[0094] Furthermore, if the eyeball contour data does not meet the eyeball model standard, based on the eyeball contour data, a preset visual servo controller is used to obtain deviation data between the camera field of view image and the target eyeball of the person to be tested; based on the deviation data, a preset feedback controller is used to adjust the orientation of the depth camera device until the eyeball contour data meets the eyeball model standard.
[0095] S1012: performing image window segmentation on the eyeball image to obtain a plurality of sub-window images corresponding to the eyeball image.
[0096] In this step, the eyeball image is segmented into image windows according to the set window size and step size, and the image window is extracted from left to right and from top to bottom to obtain multiple sub-window images corresponding to the eyeball image.
[0097] Specifically, set the window size to w×w; the step size to s; the window I i,j is I[i:i+w,j:j+w]; for each window I i,j , the center point coordinates of each window are (c x , c y ); the specific representation of the center point coordinates is as follows.
[0098] c x = j + w / 2; j = 0, s, s+1, …, w−1.
[0099] c y = i + w / 2; i = 0, s+1, s+2, …, w−1.
[0100] In one possible implementation of the present application, the window size is set to 16x16 pixels or 32x32 pixels, and the step size is 1, according to the size of the actual image.
[0101] S1013: Use a preset edge detector to perform iris contour segmentation on the eyeball image to obtain an iris contour image corresponding to the eyeball image.
[0102] In this step, the iris contour in the eyeball image is determined by a preset edge detector; then, the eyeball image is segmented by the edge detector to obtain an iris contour image corresponding to the eyeball image.
[0103] In one possible implementation of the present application, the preset edge detector is a Canny edge detector; when the Canny edge detector performs iris contour segmentation on the eye image, first, the noise of the eye image is eliminated to smooth the image; then, the Sobel operator is used to calculate the gradient of the eye image in the horizontal and vertical directions; thereafter, non-maximum suppression processing is performed on the eye image according to the amplitude of the gradient; finally, according to the set high threshold and low threshold, the pseudo edges generated by the noise are further removed, and the broken edges are connected to form a complete edge contour.
[0104] Here, the Canny edge detector adopts a non-maximum suppression method, that is, if the edge is not obvious, it is ignored, and if the edge is obvious, it is enhanced. The extracted edge is relatively thin, and the extracted iris contour is relatively irregular.
[0105] S1014: Determine the multiple sub-window images and the iris contour image as the eyeball sub-image.
[0106] In this step, a plurality of sub-window images obtained by window segmentation and an iris contour image obtained by an edge detector are determined as eyeball sub-images obtained by eyeball image segmentation.
[0107] S102: Perform edge detection on the eyeball sub-image to determine contour position data corresponding to the target eyeball.
[0108] In this step, coordinate systems are respectively established in the multiple sub-window images and iris contour image included in the eyeball sub-image, and edge detection is performed on the eyeball sub-image to determine the contour of the target eyeball of the person to be tested; and the contour point coordinates of the target eyeball in the contour coordinate system are determined, and the contour point coordinates are determined as the contour position data corresponding to the target eyeball.
[0109] S103, based on the contour position data, performing principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image respectively to obtain eyeball posture data corresponding to the target eyeball.
[0110] It should be noted that the eyeball posture data refers to the imaging posture of the target eyeball of the person to be tested in the depth camera device, and the eyeball posture data also refers to the actual posture of the target eyeball of the person to be tested; wherein the eyeball posture data includes the rotation angle and center position coordinates corresponding to the target eyeball of the person to be tested.
[0111] Here, since the spatial distance and shooting angle of the depth camera device will cause different imaging in real scenes, the image perpendicular to the center of the eyeball is usually used as a reference to calculate the rotation angle of the eyeball, that is, the tilt direction and tilt angle of the eyeball.
[0112] In this step, during the specific implementation, first, based on the contour position data of the target eyeball, a main axis rotation calculation is performed on multiple sub-window images to obtain the first eyeball posture data corresponding to the target eyeball; then, an ellipse fitting calculation is performed on the iris contour image to obtain the second eyeball posture data corresponding to the target eyeball; finally, the first eyeball posture data and the second eyeball posture data corresponding to the target eyeball are determined as the eyeball posture data corresponding to the target eyeball.
[0113] In one implementation of the present application, during specific implementation, step S103 may include:
[0114] S1031. Based on the contour position data, perform principal axis rotation calculation on the multiple sub-window images in the eyeball sub-image to obtain first eyeball posture data corresponding to the target eyeball.
[0115] In this step, during the specific implementation, first, the first rotation angle corresponding to the target eyeball is determined by performing a main axis rotation calculation on multiple sub-window images; then, multiple pixel point coordinates and window center point coordinates corresponding to each sub-window image are determined; finally, based on the first rotation angle, multiple pixel point coordinates and window center point coordinates, the first center position coordinates corresponding to the target eyeball are determined using a preset center angle calculation formula.
[0116] The first eyeball posture data includes a first rotation angle and a first center position coordinate corresponding to the target eyeball.
[0117] In one implementation of the present application, during specific implementation, step S1031 may include:
[0118] S10311. Based on the contour position data, perform principal axis rotation calculation on the multiple sub-window images in the eyeball sub-image to determine a first rotation angle corresponding to the target eyeball.
[0119] In one implementation of the present application, during specific implementation, step S10311 may include:
[0120] S103111. Decentralize the contour position data to obtain a contour point set matrix.
[0121] In this step, during the specific implementation, first, the centroid coordinates corresponding to each sub-window image are calculated based on the contour position data; then, each contour position coordinate in the contour position data is decentralized to obtain multiple de-centroided coordinates; finally, the contour point set matrix is obtained based on the multiple de-centroided coordinates.
[0122] Specifically, the expression of the centroid coordinates corresponding to each sub-window image is as follows.
[0123] .
[0124] .
[0125] Among them, (x centroid ,y centroid ) is the centroid coordinate; N is the number of contour position coordinates in the contour position data; (x k ,y k ) is an expression for the coordinates of each contour position in the contour position data.
[0126] Furthermore, the expression of the centroid coordinates is as follows.
[0127] .
[0128] .
[0129] Among them, (x k ´,y k ´) is the centroid coordinate.
[0130] Furthermore, the expression of the contour point set matrix is as follows.
[0131] .
[0132] S103112. Based on the contour point set matrix, determine a plurality of eigenvectors corresponding to the contour point set matrix using a preset covariance matrix calculation formula.
[0133] In this step, the contour point set matrix is transformed by a preset covariance matrix calculation formula to obtain a covariance matrix corresponding to the contour point set matrix, and then a plurality of eigenvectors corresponding to the contour point set matrix are determined by a feature calculation formula.
[0134] Specifically, the expression of the preset covariance matrix calculation formula is as follows.
[0135] .
[0136] Among them, C is the covariance matrix corresponding to the contour point set matrix; X is the contour point set matrix.
[0137] Furthermore, the expression of the characteristic calculation formula is as follows.
[0138] .
[0139] Among them, v is a plurality of eigenvectors corresponding to the contour point set matrix; λ is a plurality of eigenvalues corresponding to the contour point set matrix.
[0140] S103113. Based on the numbers of eigenvalues corresponding to the multiple eigenvectors, determine a eigenvector whose eigenvalue number represents a maximum value as a first eigenvector indicating the main axis direction corresponding to the target eyeball.
[0141] In this step, an eigenvalue with a maximum value is determined among multiple eigenvalues corresponding to the contour point set matrix; and an eigenvector corresponding to the eigenvalue with the maximum value is determined as the first eigenvector of the main axis direction corresponding to the target eyeball.
[0142] S103114. Based on the first feature vector, use a preset rotation angle calculation formula to determine a first rotation angle corresponding to the target eyeball.
[0143] In this step, the first feature vector is input into a preset rotation angle calculation formula to determine a first rotation angle corresponding to the target eyeball.
[0144] Specifically, the preset rotation angle calculation formula is as follows.
[0145] .
[0146] Where θ is the first rotation angle corresponding to the target eyeball; (v y , v x ) is the inverse coordinate corresponding to the first eigenvector.
[0147] S10312. Determine a plurality of pixel point coordinates and a window center point coordinate corresponding to each of the plurality of sub-window images.
[0148] In this step, a plurality of pixel point coordinates and a window center point coordinate corresponding to each sub-window image are determined in a coordinate system established for each sub-window image in the plurality of sub-window images.
[0149] S10313. Based on the first rotation angle, the coordinates of the multiple pixel points and the coordinates of the window center point, use a preset center angle calculation formula to determine the first center position coordinates corresponding to the target eyeball.
[0150] In this step, the first rotation angle, the coordinates of the plurality of pixel points and the coordinates of the center point of the window are substituted into a preset center angle calculation formula to determine the first center position coordinates corresponding to the target eyeball.
[0151] Specifically, the expression of the preset center angle calculation formula is as follows.
[0152] .
[0153] .
[0154] Among them, (x´, y´) is the coordinate of the first center position corresponding to the target eyeball; (c x , c y ) is the coordinate of the center point of the window; (x, y) is the coordinate of multiple pixel points; θ is the first rotation angle corresponding to the target eyeball.
[0155] S10314. Determine the first rotation angle and the first center position coordinates as first eyeball posture data corresponding to the target eyeball.
[0156] In this step, according to the data content included in the eyeball posture data, the first rotation angle and the first center position coordinates corresponding to the target eyeball are determined as the first eyeball posture data corresponding to the target eyeball of the person to be tested.
[0157] S1032. Based on the contour position data, perform ellipse fitting calculation on the iris contour image in the eyeball sub-image to obtain second eyeball posture data corresponding to the target eyeball.
[0158] In this step, during the specific implementation, first, the contour position data of the eyeball image is used to perform an ellipse fitting calculation on the iris contour image to determine the second rotation angle corresponding to the target eyeball; then, the coordinate mean calculation is performed on the iris contour image to obtain the second center position coordinates corresponding to the target eyeball; finally, the second rotation angle and the second center position coordinates corresponding to the target eyeball are determined as the second eyeball posture data corresponding to the target eyeball.
[0159] In one implementation of the present application, during specific implementation, step S1032 may include:
[0160] S10321. Based on the contour position data, perform ellipse fitting calculation on the iris contour image in the eyeball sub-image to determine a second rotation angle corresponding to the target eyeball.
[0161] In one implementation of the present application, during specific implementation, step S10321 may include:
[0162] S103211. Perform partial derivative convolution processing on the contour position data to obtain an elliptic curve relationship equation for fitting the target eyeball in the iris contour image.
[0163] In this step, the contour position coordinates included in the contour position data in the iris contour image are partially derived using a preset partial derivative formula to fit the points in the two-dimensional space; then, the calculated partial derivative relationship is convolved using an approximation processing formula to obtain an elliptic curve relationship equation for fitting the target eyeball in the iris contour image.
[0164] Here, since the convolution block moves with a step size determined by the pixel block, the calculated partial derivative relationship is approximated.
[0165] Specifically, the expression of the preset partial derivative formula is as follows.
[0166] .
[0167] Furthermore, the expression of the approximate processing formula is as follows.
[0168] .
[0169] Wherein, (x, y) is the contour position data in the iris contour image.
[0170] Furthermore, the expression of the elliptic curve relationship of the target eyeball fitted in the iris contour image is obtained as follows.
[0171] .
[0172] Among them, A, B, C, D, E, and F are the elliptic coefficients of the elliptic curve relationship respectively.
[0173] S103212. Establish an elliptical focal length matrix based on the focal length parameter corresponding to the depth camera device and the elliptic curve relationship.
[0174] In this step, according to the establishment relationship of the preset elliptical focal length matrix, based on the focal length parameters corresponding to the depth camera device and the elliptical coefficients in the elliptical curve relationship, an elliptical focal length matrix for diagonalization calculation is established.
[0175] Specifically, the expression of the elliptical focal length matrix is as follows.
[0176] .
[0177] Among them, Q is the elliptical focal length matrix; f is the focal length parameter corresponding to the depth camera device; A, B, C, D, E, and F are the elliptic coefficients of the elliptic curve relationship respectively.
[0178] S103213. Based on the elliptical focal length matrix, use a preset diagonalization calculation formula to determine the second eigenvector corresponding to the target eyeball.
[0179] In this step, the elliptical focal length matrix is diagonalized using a preset diagonalization calculation formula to obtain a second eigenvector corresponding to the target eyeball.
[0180] Specifically, the expression of the preset diagonalization calculation formula is as follows.
[0181] .
[0182] .
[0183] in, is the diagonal matrix corresponding to the second eigenvector corresponding to the target eyeball; are the eigenvalues corresponding to the second eigenvector corresponding to the target eyeball; is the preset orthogonal matrix; Q is the elliptical focal length matrix.
[0184] S103214. Based on the eigenvalue corresponding to the second eigenvector and the set normal vector parameters, use a preset normal vector calculation formula to determine the normal vector corresponding to the target eyeball.
[0185] In this step, the eigenvalue corresponding to the second eigenvector and the set normal vector parameters are input into a preset normal vector calculation formula to calculate and determine the normal vector corresponding to the target eyeball.
[0186] Specifically, the expression of the preset normal vector calculation formula is as follows.
[0187] .
[0188] Among them, N is the normal vector corresponding to the target eyeball; is a preset orthogonal matrix; are the eigenvalues corresponding to the second eigenvector respectively; S1 and S2 are the set normal vector parameters respectively.
[0189] Here, the set normal vector parameters are used to determine the direction corresponding to the target eyeball. The normal vector corresponding to the target eyeball represents the position and direction of the target eyeball in space; the normal vector parameters S1 and S2 are represented by +1 or −1 respectively to indicate the direction of the normal vector; for example, the normal vector of an ellipse points in the positive direction, at this time S1=+1, S2=+1, so that the direction of the normal vector is consistent with the direction of the feature vector.
[0190] S103215. Based on the direction parameter corresponding to the normal vector, determine the rotation angle of the direction parameter relative to the iris contour image as a second rotation angle corresponding to the target eyeball.
[0191] In this step, the direction parameter corresponding to the normal vector represents the specific pointing direction of the normal vector, and the second rotation angle corresponding to the target eyeball is determined according to the rotation angle of the direction parameter corresponding to the normal vector relative to the iris contour image.
[0192] S10322. Calculate the coordinate mean of the iris contour image to obtain the second center position coordinates corresponding to the target eyeball.
[0193] In this step, for the contour position data in the iris contour image, a plurality of contour position coordinates included in the contour position data are averaged and the second center position coordinates corresponding to the target eyeball are obtained.
[0194] S10323. Determine the second rotation angle and the second center position coordinates as second eyeball posture data corresponding to the target eyeball.
[0195] In this step, the second rotation angle calculated by ellipse fitting and the second center position coordinates calculated by the coordinate mean are determined as the second eyeball posture data corresponding to the target eyeball.
[0196] Furthermore, in order to determine the three-dimensional position of the target eyeball, the three-dimensional position of the target eyeball is calculated by corresponding feature points based on the camera internal parameters and interface data corresponding to the iris contour image relative to the depth camera device.
[0197] S104, determining the eyeball posture position of the target eyeball based on the contour position data and the eyeball posture data, as well as the posture prediction data output by a preset detection model according to the eyeball image.
[0198] In this step, during the specific implementation, first, the eyeball image is taken as input using a preset detection model to obtain posture prediction data output by the detection model, and the posture prediction data is used as one of the eyeball posture data corresponding to the eyeball image; then, the first eyeball posture data, the second eyeball posture data and the posture prediction data are averaged, and the average calculation result is determined as the actual eyeball posture data corresponding to the target eyeball; finally, the eyeball posture position of the target eyeball is determined based on the contour position data and the actual eyeball posture data.
[0199] In one possible implementation of the present application, the preset detection model is a YOLO target detection model, which regards the target detection problem as a single regression problem and directly predicts the bounding box and category probability from the image pixels.
[0200] Specifically, the YOLO target detection model includes a priori boxes (Anchor, pre-defined boxes), convolutional layer (Backbone), feature processing layer (Neck), output layer (Head) and loss function (Loss).
[0201] Among them, the prior box (Anchor, pre-defined boxes) is used to help the model learn how to predict the bounding box of the target faster; the model first performs data enhancement on the received eyeball image and then inputs it into the convolution layer (Backbone). The convolution layer is the backbone network of the model, which contains multiple sub-convolution layers (such as conv1, conv2, conv3, etc.), which is responsible for extracting features from the image. These sub-convolution layers are stacked to gradually reduce the spatial dimensions (i.e., width and height) of the image while increasing the depth (i.e., number of channels) of the features; the feature processing layer (Neck) performs further feature extraction processing on the features extracted by the convolution layer, and generates multiple feature maps of different scales in the feature processing layer. These feature maps are used to detect targets of different sizes; the output layer is responsible for generating the final detection results based on the feature maps provided by the feature processing layer, including the coordinates of the bounding box, the confidence (indicating the probability of the existence of the target in the bounding box), and the category score; the loss function (Loss) is calculated for the output layer. The loss function is used to evaluate the difference between the model prediction and the true label. Through the back propagation algorithm, the gradient of the loss function is used to update the weights of the model, thereby optimizing the performance of the model.
[0202] The method for determining the eyeball posture position provided in the embodiment of the present application determines the contour position of the target eyeball by performing edge detection on the eyeball sub-image segmented from the eyeball image corresponding to the target eyeball of the person to be tested, and determines the eyeball posture by performing principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image according to the contour position. The contour position and eyeball posture are combined with the posture prediction result of the eyeball image by the detection model to comprehensively determine the eyeball posture position of the target eyeball, thereby improving the data accuracy of the eyeball posture position determination.
[0203] See also Figure 2 , Figure 2 This is a schematic diagram of the structure of a device for determining the position of an eyeball posture provided by an embodiment of the present application. Figure 2 As shown in , the eyeball posture position determination device 200 includes:
[0204] An image segmentation module 210 is used to segment the acquired eyeball image corresponding to the target eyeball of the person to be tested, and obtain eyeball sub-images segmented from the eyeball image;
[0205] A contour determination module 220 is used to perform edge detection on the eyeball sub-image to determine contour position data corresponding to the target eyeball;
[0206] A posture calculation module 230 is used to perform a principal axis rotation calculation and an ellipse fitting calculation on the eyeball sub-image based on the contour position data to obtain eyeball posture data corresponding to the target eyeball;
[0207] The posture determination module 240 is used to determine the eye posture position of the target eye based on the contour position data and the eye posture data, as well as the posture prediction data output by a preset detection model according to the eye image.
[0208] Furthermore, when the image segmentation module 210 is used to segment the acquired eyeball image corresponding to the target eyeball of the person to be tested to obtain the eyeball sub-image segmented from the eyeball image, the image segmentation module 210 is used to:
[0209] Obtaining an eyeball image corresponding to a target eyeball of the person to be tested captured by a depth camera device;
[0210] Performing image window segmentation on the eyeball image to obtain a plurality of sub-window images corresponding to the eyeball image;
[0211] Using a preset edge detector to perform iris contour segmentation on the eyeball image to obtain an iris contour image corresponding to the eyeball image;
[0212] The multiple sub-window images and the iris contour image are determined as the eyeball sub-image.
[0213] Further, when the posture calculation module 230 is used to perform the principal axis rotation calculation and the ellipse fitting calculation on the eyeball sub-image based on the contour position data to obtain the eyeball posture data corresponding to the target eyeball, the posture calculation module 230 is used to:
[0214] Based on the contour position data, a principal axis rotation calculation is performed on the plurality of sub-window images in the eyeball sub-image to obtain first eyeball posture data corresponding to the target eyeball;
[0215] Based on the contour position data, an ellipse fitting calculation is performed on the iris contour image in the eyeball sub-image to obtain second eyeball posture data corresponding to the target eyeball.
[0216] Further, when the posture calculation module 230 is used to perform a principal axis rotation calculation on the multiple sub-window images in the eyeball sub-image based on the contour position data to obtain the first eyeball posture data corresponding to the target eyeball, the posture calculation module 230 is used to:
[0217] Based on the contour position data, performing a principal axis rotation calculation on the plurality of sub-window images in the eyeball sub-image to determine a first rotation angle corresponding to the target eyeball;
[0218] Determine a plurality of pixel point coordinates and a window center point coordinate corresponding to each sub-window image in the plurality of sub-window images;
[0219] Based on the first rotation angle, the coordinates of the plurality of pixel points and the coordinates of the center point of the window, a preset center angle calculation formula is used to determine the first center position coordinates corresponding to the target eyeball;
[0220] The first rotation angle and the first center position coordinates are determined as first eyeball posture data corresponding to the target eyeball.
[0221] Further, when the posture calculation module 230 is used to perform a principal axis rotation calculation on the multiple sub-window images in the eyeball sub-image based on the contour position data to determine a first rotation angle corresponding to the target eyeball, the posture calculation module 230 is used to:
[0222] Decentralizing the contour position data to obtain a contour point set matrix;
[0223] Based on the contour point set matrix, a plurality of eigenvectors corresponding to the contour point set matrix are determined using a preset covariance matrix calculation formula;
[0224] Based on the number of eigenvalues respectively corresponding to the plurality of eigenvectors, determining a eigenvector whose eigenvalue number represents a maximum value as a first eigenvector indicating the main axis direction corresponding to the target eyeball;
[0225] Based on the first feature vector, a first rotation angle corresponding to the target eyeball is determined using a preset rotation angle calculation formula.
[0226] Further, when the posture calculation module 230 is used to perform ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to obtain the second eyeball posture data corresponding to the target eyeball, the posture calculation module 230 is used to:
[0227] Based on the contour position data, performing an ellipse fitting calculation on the iris contour image in the eyeball sub-image to determine a second rotation angle corresponding to the target eyeball;
[0228] Performing coordinate mean calculation on the iris contour image to obtain the second center position coordinates corresponding to the target eyeball;
[0229] The second rotation angle and the second center position coordinates are determined as second eyeball posture data corresponding to the target eyeball.
[0230] Further, when the posture calculation module 230 is used to perform ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to determine the second rotation angle corresponding to the target eyeball, the posture calculation module 230 is used to:
[0231] Performing partial derivative convolution processing on the contour position data to obtain an elliptic curve relationship equation for fitting the target eyeball in the iris contour image;
[0232] Based on the focal length parameter corresponding to the depth camera device and the elliptic curve relationship, an elliptical focal length matrix is established;
[0233] Based on the elliptical focal length matrix, a second eigenvector corresponding to the target eyeball is determined using a preset diagonalization calculation formula;
[0234] Based on the eigenvalue corresponding to the second eigenvector and the set normal vector parameters, a preset normal vector calculation formula is used to determine the normal vector corresponding to the target eyeball;
[0235] Based on the direction parameter corresponding to the normal vector, a rotation angle of the direction parameter relative to the iris contour image is determined as a second rotation angle corresponding to the target eyeball.
[0236] The device for determining the eyeball posture position provided in the embodiment of the present application determines the contour position of the target eyeball by performing edge detection on the eyeball sub-image segmented from the eyeball image corresponding to the target eyeball of the person to be tested, and determines the eyeball posture by performing principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image according to the contour position. The contour position and eyeball posture are combined with the posture prediction result of the eyeball image by the detection model to comprehensively determine the eyeball posture position of the target eyeball, thereby improving the data accuracy of the eyeball posture position determination.
[0237] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 3 As shown in , the electronic device 300 includes a processor 310 , a memory 320 and a bus 330 .
[0238] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 is running, the processor 310 communicates with the memory 320 via the bus 330. When the machine-readable instructions are executed by the processor 310, the above-mentioned Figure 1 The steps of the method for determining the eyeball posture position in the method embodiment shown, the specific implementation method can be found in the method embodiment, and will not be repeated here.
[0239] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 The steps of the method for determining the eyeball posture position in the method embodiment shown, the specific implementation method can be found in the method embodiment, and will not be repeated here.
[0240] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0241] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.
[0242] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0243] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0244] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.
[0245] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.
Claims
1. A method for determining eyeball posture position, characterized in that: The determination method comprises: Segmenting the acquired eyeball image corresponding to the target eyeball of the person to be tested to obtain eyeball sub-images segmented from the eyeball image; Performing edge detection on the eyeball sub-image to determine contour position data corresponding to the target eyeball; Based on the contour position data, performing a principal axis rotation calculation and an ellipse fitting calculation on the eyeball sub-image respectively to obtain eyeball posture data corresponding to the target eyeball; Perform a mean calculation on the eyeball posture data and the posture prediction data output by a preset detection model based on the eyeball image, determine the mean calculation result as the actual eyeball posture data corresponding to the target eyeball, and determine the eyeball posture position of the target eyeball based on the contour position data and the actual eyeball posture data.
2. The method according to claim 1, characterized in that The step of segmenting the acquired eyeball image corresponding to the target eyeball of the person to be tested to obtain eyeball sub-images segmented from the eyeball image includes: Obtaining an eyeball image corresponding to a target eyeball of the person to be tested captured by a depth camera device; Performing image window segmentation on the eyeball image to obtain a plurality of sub-window images corresponding to the eyeball image; Using a preset edge detector to perform iris contour segmentation on the eyeball image to obtain an iris contour image corresponding to the eyeball image; The multiple sub-window images and the iris contour image are determined as the eyeball sub-image.
3. The method according to claim 2, characterized in that The step of performing a principal axis rotation calculation and an ellipse fitting calculation on the eyeball sub-image based on the contour position data to obtain eyeball posture data corresponding to the target eyeball includes: Based on the contour position data, a principal axis rotation calculation is performed on the plurality of sub-window images in the eyeball sub-image to obtain first eyeball posture data corresponding to the target eyeball; Based on the contour position data, an ellipse fitting calculation is performed on the iris contour image in the eyeball sub-image to obtain second eyeball posture data corresponding to the target eyeball.
4. The method according to claim 3, characterized in that The step of performing a principal axis rotation calculation on the plurality of sub-window images in the eyeball sub-image based on the contour position data to obtain first eyeball posture data corresponding to the target eyeball includes: Based on the contour position data, performing a principal axis rotation calculation on the plurality of sub-window images in the eyeball sub-image to determine a first rotation angle corresponding to the target eyeball; Determine a plurality of pixel point coordinates and a window center point coordinate corresponding to each sub-window image in the plurality of sub-window images; Based on the first rotation angle, the coordinates of the plurality of pixel points and the coordinates of the center point of the window, a preset center angle calculation formula is used to determine the first center position coordinates corresponding to the target eyeball; The first rotation angle and the first center position coordinates are determined as first eyeball posture data corresponding to the target eyeball.
5. The method according to claim 4, characterized in that The step of performing a principal axis rotation calculation on the plurality of sub-window images in the eyeball sub-image based on the contour position data to determine a first rotation angle corresponding to the target eyeball includes: Decentralizing the contour position data to obtain a contour point set matrix; Based on the contour point set matrix, a plurality of eigenvectors corresponding to the contour point set matrix are determined using a preset covariance matrix calculation formula; Based on the number of eigenvalues respectively corresponding to the plurality of eigenvectors, determining a eigenvector whose eigenvalue number represents a maximum value as a first eigenvector indicating the main axis direction corresponding to the target eyeball; Based on the first feature vector, a first rotation angle corresponding to the target eyeball is determined using a preset rotation angle calculation formula.
6. The method according to claim 3, characterized in that The step of performing an ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to obtain second eyeball posture data corresponding to the target eyeball includes: Based on the contour position data, performing an ellipse fitting calculation on the iris contour image in the eyeball sub-image to determine a second rotation angle corresponding to the target eyeball; Performing coordinate mean calculation on the iris contour image to obtain the second center position coordinates corresponding to the target eyeball; The second rotation angle and the second center position coordinates are determined as second eyeball posture data corresponding to the target eyeball.
7. The method according to claim 6, characterized in that The step of performing an ellipse fitting calculation on the iris contour image in the eyeball sub-image based on the contour position data to determine a second rotation angle corresponding to the target eyeball includes: Performing partial derivative convolution processing on the contour position data to obtain an elliptic curve relationship equation for fitting the target eyeball in the iris contour image; Based on the focal length parameter corresponding to the depth camera device and the elliptic curve relationship, an elliptical focal length matrix is established; Based on the elliptical focal length matrix, a second eigenvector corresponding to the target eyeball is determined using a preset diagonalization calculation formula; Based on the eigenvalue corresponding to the second eigenvector and the set normal vector parameters, a preset normal vector calculation formula is used to determine the normal vector corresponding to the target eyeball; Based on the direction parameter corresponding to the normal vector, a rotation angle of the direction parameter relative to the iris contour image is determined as a second rotation angle corresponding to the target eyeball.
8. A device for determining eyeball posture position, characterized in that: The determining device comprises: An image segmentation module is used to segment the acquired eyeball image corresponding to the target eyeball of the person to be tested, and obtain eyeball sub-images segmented from the eyeball image; A contour determination module, used to perform edge detection on the eyeball sub-image to determine contour position data corresponding to the target eyeball; A posture calculation module, used to perform principal axis rotation calculation and ellipse fitting calculation on the eyeball sub-image based on the contour position data, so as to obtain eyeball posture data corresponding to the target eyeball; A posture determination module is used to calculate the mean of the eyeball posture data and the posture prediction data output by a preset detection model based on the eyeball image, determine the mean calculation result as the actual eyeball posture data corresponding to the target eyeball, and determine the eyeball posture position of the target eyeball based on the contour position data and the actual eyeball posture data.
9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to perform the steps of the method for determining the eye posture position as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for determining the eyeball posture position as described in any one of claims 1 to 7 are executed.
