An eye movement tracking model calibration method and device

Abstract

The present invention discloses a method and device for calibrating an eye movement tracking model. The method includes: displaying a current calibration point on a human-computer interaction interface, and prompting the user to fixate on the current calibration point, and then simultaneously capturing eye images of the user by using a plurality of cameras arranged in an array; constructing a virtual cone, using the area enclosed by the intersection line of the virtual cone and the captured eye image as a cropping area, and cropping a corresponding sub-image from the captured eye image according to the cropping area; using the sub-images corresponding to each camera to splice into a target eye image for the current calibration point; selecting one of the other calibration points in the calibration point set except for the calibration points that have been traversed as the current calibration point, and returning to execute the step of displaying the current calibration point on the human-computer interaction interface until all calibration points have been traversed; performing calibration according to the target eye images corresponding to all calibration points.

Description

Technical Field

[0001] The present invention relates to the technical field of eye movement tracking, and more particularly to a method and device for calibrating an eye movement tracking model. Background Art

[0002] Eye movement tracking is the process of measuring the movement of the eyes. The most concerned event in eye movement tracking research is to determine the fixation point or gaze point of the user. Currently, the popular eye movement tracking technology is mainly the "non-invasive" technology based on eye video analysis. Its basic principle is as follows: A beam of near-infrared light is irradiated onto the user's face, and the user's eyeball will reflect the near-infrared light, and then a light spot is formed at the reflection point. When the camera captures the eye image, the light spot can be captured; at the same time, the position of the pupil center is obtained by using image processing methods. Then, the corneal reflection point is used as the base point of the relative position between the eye camera and the eyeball. According to the position of the pupil center obtained by image processing, the line-of-sight vector coordinates can be obtained by using the offset of the pupil center position relative to the corneal reflection point, so as to determine the human eye fixation point. Based on video eye trackers, in addition to monitoring fixation, other useful measurement indicators can also be displayed, including pupil size and blink rate, etc.

[0003] Due to individual differences in the shape, size, and structure of the human eye, there is a non-linear relationship between the projection point position of the points on the eye sphere in the camera reference system and the eye rotation angle, and there is a model error between the line-of-sight estimation direction and the true line-of-sight direction. Therefore, the line-of-sight tracking system requires a calibration link. Therefore, before conducting an eye movement experiment, an eye movement calibration process is generally carried out. During this process, the algorithm will measure the characteristics of the subject's eyes, including the cornea, pupil, and reflection information, etc., to calculate the eye movement data process. During the calibration process, the subject needs to observe the points that appear at specific positions on the stimulus screen, and this point is called the calibration point. The reference position of the calibration point is associated with the eye movement characteristics, and then the parameters in the eye movement tracking algorithm are corrected to achieve more accurate eye movement tracking.

[0004] However, although the human eye movement tracking algorithm has been calibrated in the prior art, in the prior art, a single camera is used to capture the user's eye movement data. Since the user's facial posture changes dynamically, when the user gazes at different calibration points, the facial posture may be different, which leads to a change in the calibration basis. Therefore, the prior art has the technical problem of inaccurate calibration. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and device for calibrating an eye movement tracking model to improve the calibration accuracy.

[0006] The present invention solves the above technical problems through the following technical solutions:

[0007] The present invention provides a method for calibrating an eye movement tracking model, and the method includes:

[0008] Select a calibration point from the calibration point set as the current calibration point, display the current calibration point on the human-computer interaction interface, and prompt the user to fixate on the current calibration point. Then, use a plurality of cameras arranged in an array to simultaneously capture the user's eye images, and the number of cameras is two or more;

[0009] For each camera, with the optical axis direction of the camera as the central axis, preset a cone angle to construct a virtual cone, use the area surrounded by the intersection line of the virtual cone and the captured eye image as the cropping area, and crop the corresponding sub-image from the captured eye image according to the cropping area;

[0010] Use the sub-images corresponding to each camera to splice into a target eye image for the current calibration point; select one from the other calibration points in the calibration point set except the ones that have been traversed as the current calibration point, and return to execute the step of displaying the current calibration point on the human-computer interaction interface until all calibration points have been traversed; perform calibration according to the target eye images corresponding to all calibration points.

[0011] Optionally, the determination process of the calibration point set includes:

[0012] Filter out the peak hours when the user uses the current device according to the frequency of the user operating the current device, and filter out the target eye movement tracking data from the historical eye movement tracking data;

[0013] For each piece of eye movement tracking data, map the visual focus corresponding to the target eye movement tracking data to the current device screen to obtain a visual focus set;

[0014] Perform clustering processing on the visual foci included in the visual focus set to obtain several focus clusters; according to the distribution characteristics of the visual foci included in each focus cluster, obtain the corresponding number of calibration points, and obtain the corresponding calibration point distribution data according to the number of calibration points, and generate a calibration point set according to the calibration point distribution data corresponding to each focus cluster.

[0015] Optionally, the obtaining the corresponding number of calibration points according to the distribution characteristics of the visual foci included in each focus cluster, and obtaining the corresponding calibration point distribution data according to the number of calibration points includes:

[0016] For the current focus cluster among a number of focus clusters, obtain the center of the current focus cluster, and obtain the average value of the connecting line segments between all the included visual foci. Take the area enclosed by using the center of the current focus cluster as the center and the average value as the radius as the target area; obtain the number of calibration points corresponding to the current focus cluster according to the proportion of the target area in the total target area of all visual focus clusters; randomly distribute the calibration points corresponding to the current focus cluster in the corresponding target area to obtain the calibration point distribution data for the current focus cluster.

[0017] Optionally, when obtaining the number of calibration points corresponding to the current focus cluster according to the proportion of the target area in the total target area of all visual focus clusters, the method includes:

[0018] Judge whether the number of calibration points corresponding to the current focus cluster is less than one. If so, set the number of calibration points corresponding to the current cluster focus to one.

[0019] Optionally, the method for determining the several cameras includes:

[0020] Broadcast a shooting assistance request to the surroundings, where the shooting assistance request includes the first coordinate of the current device itself; so that the assisting device that receives the shooting assistance request calculates its own optimal shooting range according to its own second coordinate and the optimal shooting distance pre-calibrated by the manufacturer, and responds to the shooting assistance request when the first coordinate is within the optimal shooting range;

[0021] Take the camera carried by the assisting device that responds to the shooting assistance request as the camera for shooting the user's eye image.

[0022] Optionally, the process of obtaining the preset cone angle includes:

[0023] According to the distribution positions of the cameras, evenly divide the eye area corresponding to each camera;

[0024] Calculate the size of the preset cone angle according to the corresponding eye area and the distance from the camera to the eye.

[0025] Optionally, the process of stitching the sub-images corresponding to each camera into a target eye image for the current calibration point includes:

[0026] According to the coordinates of the corresponding camera, convert the pixel points of each sub-image into the same coordinate system. According to the eye area corresponding to each sub-image, stitch each sub-image into a complete eye image, and take the complete eye image as the target eye image.

[0027] Optionally, the calibration according to the target eye images corresponding to all calibration points includes:

[0028] Obtain the reflection spot in the target eye image, smooth the edge of the reflection spot to obtain a smoothed contour; fit the lines of the smoothed contour into an ellipse, take the center point of the ellipse as the corneal reflection center, and perform correction according to the corneal reflection center and the target eye image.

[0029] The present invention also provides an eye movement tracking model correction device, and the device includes:

[0030] A display module, configured to select a correction point from the correction point set as the current correction point, display the current correction point on the human-computer interaction interface, and prompt the user to fixate on the current correction point, and then use a plurality of cameras arranged in an array to simultaneously capture the eye images of the user, and the number of cameras is two or more;

[0031] A cropping module, for each camera, taking the optical axis direction of the camera as the central axis, presetting a cone angle to construct a virtual cone, using the area surrounded by the intersection line of the virtual cone and the captured eye image as the cropping area, and cropping the corresponding sub-image from the captured eye image according to the cropping area;

[0032] A splicing module, configured to splice the sub-images corresponding to each camera into a target eye image for the current correction point; select one of the other correction points in the correction point set except the correction points that have been traversed as the current correction point, and return to execute the step of displaying the current correction point on the human-computer interaction interface until all correction points have been traversed; perform correction according to the target eye images corresponding to all correction points.

[0033] Optionally, the determination process of the correction point set includes:

[0034] Screen the peak time period when the user uses the current device according to the frequency of the user operating the current device, and screen the target eye movement tracking data from the historical eye movement tracking data;

[0035] For each piece of eye movement tracking data, map the visual focus corresponding to the target eye movement tracking data to the current device screen to obtain a visual focus set;

[0036] Perform clustering processing on the visual foci included in the visual focus set to obtain several focus clusters; according to the distribution characteristics of the visual foci included in each focus cluster, obtain the corresponding number of correction points, and obtain the corresponding correction point distribution data according to the number of correction points, and generate a correction point set according to the correction point distribution data corresponding to each focus cluster.

[0037] The present invention has the following advantages compared with the prior art:

[0038] In the present invention, multiple cameras respectively capture sub-images within a very small range in the direction of their respective optical axes, and then the sub-images are stitched together to form a target eye image. Since the distortion of the image captured by the camera along the optical axis direction is the smallest, the overall distortion of the target research image is smaller than the total distortion captured by a single camera, thereby improving the calibration accuracy.

[0039] In addition, the sub-images corresponding to each camera are used to stitch together a target eye image for the current calibration point, and the cooperation among multiple cameras can eliminate the errors and external interferences existing during the shooting of a single camera.

[0040] Taking the optical axis direction of the camera as the central axis, a virtual cone is preset with a cone angle, and the area enclosed by the intersection line of the virtual cone and the captured eye image is used as the cropping area, and the corresponding sub-image is cropped from the captured eye image according to the cropping area. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic flowchart of an eye movement tracking model calibration method provided by an embodiment of the present invention;

[0042] Figure 2 It is a schematic diagram of the distribution of calibration points in an eye movement tracking model calibration method provided by an embodiment of the present invention;

[0043] Figure 3 It is a schematic diagram of the position of the reflected light spot relative to the pupil in an eye movement tracking model calibration method provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] The embodiments of the present invention will be described in detail below. The embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation manners and specific operation processes are given. However, the protection scope of the present invention is not limited to the following embodiments.

[0045] Embodiment 1:

[0046] Figure 1 It is a schematic flowchart of an eye movement tracking model calibration method provided by an embodiment of the present invention. As Figure 1 described, the method includes:

[0047] S101: Select a calibration point from the calibration point set as the current calibration point, display the current calibration point on the human-computer interaction interface, and prompt the user to fixate on the current calibration point, and then use a plurality of cameras arranged in an array to simultaneously capture the eye image of the user, and the number of cameras is two or more;

[0048] Specifically, first, the peak period of the user using the current device can be filtered according to the frequency of the user operating the current device. For example, the period when the frequency of the user operating the current device is greater than 5 times / 10 min can be used as the peak period, and then the eye movement tracking data in the peak period is used as the target eye movement tracking data. Using the eye movement tracking data in the peak period can improve the accuracy of tracking.

[0049] Then, for each piece of eye movement tracking data, map the visual focus corresponding to the target eye movement tracking data onto the screen of the current device to obtain a set of visual foci.

[0050] Perform clustering processing on the visual foci included in the set of visual foci to obtain several focus clusters. Taking the current focus cluster A among the several focus clusters as an example, obtain the center point a of the current focus cluster, combine each pair of visual foci in the focus cluster A, calculate the distance between the visual foci included in each combination, and then calculate the average value of all the distances. At the same time, obtain the center points of each visual focus in the focus cluster A, draw a circle with the center point as the center and the average value as the radius. The area within this circle is used as the target area 1 corresponding to the focus cluster A. Similarly, the target area 2 corresponding to the focus cluster B and the target area 3 corresponding to the focus cluster C can be obtained. The total area of the target areas of all visual focus clusters is the sum of the total areas of the target area 1, the target area 2, and the target area 3.

[0051] Obtain the number of calibration points corresponding to the current focus cluster A according to the proportion of the target area 1 in the total area of the target areas of all visual focus clusters. For example, if the number of calibration points corresponding to the current focus cluster A is 3.4, then round up to obtain the number of calibration points corresponding to the current focus cluster A as 4. Applying the above embodiments of the present invention, rounding up can increase the number of calibration points, thereby improving the calibration accuracy. Further, if it is determined that the number of calibration points corresponding to the current focus cluster is less than one, set the number of calibration points corresponding to the current focus cluster to one.

[0052] Randomly distribute the 4 calibration points corresponding to the current focus cluster in the corresponding target area to obtain the calibration point distribution data for the current focus cluster.

[0053] Take the current focus cluster A as the keyword, and use the coordinates of each calibration point and the characteristic information of the calibration point, such as cross-shaped or rectangular information, as the values to generate the calibration point set corresponding to the current focus cluster A. Applying the above embodiments of the present invention, there are more calibration points in the place where the user's visual focus is densely distributed, and there are fewer calibration points in the place where the user's visual focus is sparsely distributed. Therefore, a calibration point set with appropriate accuracy can be obtained according to the distribution characteristics of the user's visual focus. Figure 2Schematic diagram of the distribution of calibration points in an eye movement tracking model calibration method provided by an embodiment of the present invention, as follows Figure 2 As shown, the visual focus of the user is evenly distributed on the display screen. Therefore, the calibration points are also evenly distributed. The calibration points can be one point, or 3 points, 5 points, 9 points, and 13 points. The number of calibration points varies according to different experimental tasks. This algorithm creates a mathematical transformation between the eye position (subtracting CR) and the fixation position of each target, and then creates a matrix to cover the entire calibration area and interpolates between each point. The more calibration points are used, the higher the accuracy and uniformity in the entire field of view.

[0054] Then, broadcast a shooting assistance request to the surroundings, where the shooting assistance request includes the first coordinate of the current device itself. The assisting device that receives the shooting assistance request calculates its own optimal shooting range according to its own second coordinate and the optimal shooting distance pre-calibrated by the manufacturer. When the first coordinate is within the optimal shooting range, it responds to the shooting assistance request;

[0055] Use the camera carried by the assisting device that responds to the shooting assistance request as the camera for shooting the user's eye image. By applying the above embodiments of the present invention, the camera in the model can be used to shoot the user's eye image, and at the same time, the camera of the assisting device can also be used to shoot the user's eye image. In this way, using two cameras to shoot the eye image can eliminate the error caused by the perspective effect of a single camera, thereby improving the accuracy.

[0056] S102: For each camera including one or more cameras set by the device itself and the camera of the assisting device, with the optical axis direction of the camera as the central axis, preset a cone angle to construct a virtual cone, and use the area surrounded by the intersection line of the virtual cone and the captured eye image as the cropping area, and crop the corresponding sub-image from the captured eye image according to the cropping area.

[0057] Specifically, according to the distribution positions of camera 1 and camera 2, it is necessary to evenly divide the eye area corresponding to each camera; for example, if camera 1 is on the left, camera 2 is in the middle, and camera 3 is on the right, then the three cameras respectively correspond to one-third of the eye area. Measure the left one-third of the eye area from the first facial image captured by camera 1, obtain the middle one-third of the eye area from camera 2 in the middle, and obtain the right one-third of the eye area from camera 3.

[0058] In practical applications, the pattern formed by the connection lines of the projection points of the distribution positions of each camera on the vertical plane can be a triangle, a quadrilateral, a pentagon, a hexagon, etc. Therefore, the eye areas corresponding to each camera should also be in the corresponding positions in the eye.

[0059] To briefly describe the technical process for achieving the above effects, the embodiments of the present invention take a camera 1 and a camera 2 as examples for a principle description. Camera 1 is located on the left relative to camera 2. Therefore, the corresponding eyeball area of camera 1 is the left half, and the corresponding eyeball area of camera 2 is the right half. That is, half of the eyeball image is obtained from the first facial image captured by camera 1 and the second facial image captured by camera 2 respectively. Therefore, taking half of the diameter of the eyeball image in the first facial image as the arc length and the distance from camera 1 to the eyeball as the radius, the cone angle corresponding to camera 1 is calculated. Then, with the line connecting camera 1 to the eyeball, that is, the optical axis direction of camera 1 as the central axis, a second virtual cone is obtained. Similarly, the method for obtaining the second virtual cone of camera 2 is similar to the above method.

[0060] Then, the area enclosed by the intersection line of the first virtual cone and the eyeball image captured by camera 1 is used as the cropping area, and the corresponding sub-image 1 is cropped from the captured eyeball image according to the cropping area.

[0061] Then, the area enclosed by the intersection line of the second virtual cone and the eyeball image captured by camera 2 is used as the cropping area, and the corresponding sub-image 2 is cropped from the captured eyeball image according to the cropping area.

[0062] S103: Use the sub-images corresponding to each camera to stitch together the target eyeball image for the current calibration point; select one of the other calibration points in the calibration point set except the ones that have been traversed as the current calibration point, and return to execute the step of displaying the current calibration point on the human-computer interaction interface until all calibration points have been traversed; perform calibration according to the target eyeball images corresponding to all calibration points.

[0063] First, the method of iris feature point matching can be used to stitch sub-image 1 and sub-image 2 into the target eyeball image.

[0064] Furthermore, according to the coordinates of the corresponding camera, the pixel points in the sub-images captured by each camera can be converted into the same coordinate system, and the sub-images corresponding to camera 1 and the sub-images corresponding to camera 2 are scaled to the same size. The sub-image corresponding to camera 1 is placed on the left, and the sub-image corresponding to camera 2 is placed on the right. The method of feature point coincidence is used to stitch each sub-image into a complete eyeball image, and the complete eyeball image is used as the target eyeball image.

[0065] Then, select one of the other calibration points except the current calibration point as the current calibration point, and execute the above embodiments of the present invention again until the user has viewed all calibration points.

[0066] Figure 3Schematic diagram of the position of the reflected light spot relative to the pupil in an eye movement tracking model calibration method provided by an embodiment of the present invention, as Figure 3 shown, Figure 3 In Figure 3 , the black area is the pupil, and the circle is the reflected light spot. The reflected light spot in the target eye image is obtained, and the edge of the reflected light spot is smoothed to obtain a smoothed contour; the line of the smoothed contour is fitted into an ellipse, and the center point of the ellipse is used as the corneal reflection center, and calibration is performed according to the corneal reflection center and the target eye image.

[0067] The specific calibration method is prior art, and the embodiments of the present invention will not be elaborated herein.

[0068] The present invention respectively captures sub-images within a very small range in the direction of its own optical axis through multiple cameras, and then stitches the sub-images into a target eye image. Since the distortion of the image captured by the camera along the optical axis direction is the smallest, therefore, the overall distortion of the target research image is smaller than the total distortion captured by a single camera, and thus the calibration accuracy can be improved.

[0069] Embodiment 2:

[0070] Corresponding to Embodiment 1 of the present invention, Embodiment 2 of the present invention further provides an eye movement tracking model calibration device, and the device includes:

[0071] A display module, configured to select a calibration point from the calibration point set as the current calibration point, display the current calibration point on the human-computer interaction interface, and prompt the user to fixate on the current calibration point, and then simultaneously capture the eye image of the user by using a plurality of cameras arranged in an array, and the number of cameras is two or more;

[0072] A cropping module, for each camera, taking the optical axis direction of the camera as the central axis, presetting a cone angle to construct a virtual cone, using the area surrounded by the intersection line of the virtual cone and the captured eye image as the cropping area, and cropping out the corresponding sub-image from the captured eye image according to the cropping area;

[0073] A stitching module, configured to stitch the sub-images corresponding to each camera into a target eye image for the current calibration point; select one of the other calibration points in the calibration point set except the calibration points that have been traversed as the current calibration point, and return to execute the step of displaying the current calibration point on the human-computer interaction interface until all calibration points have been traversed; perform calibration according to the target eye images corresponding to all calibration points.

[0074] In a specific implementation manner of the embodiment of the present invention, the determination process of the calibration point set includes:

[0075] Filter out the peak period when the user uses the current device according to the frequency of the user operating the current device, and filter out the target eye movement tracking data from the historical eye movement tracking data;

[0076] For each piece of eye movement tracking data, map the visual focus corresponding to the target eye movement tracking data to the screen of the current device to obtain a set of visual foci;

[0077] Perform clustering processing on the visual foci included in the set of visual foci to obtain several focus clusters; according to the distribution characteristics of the visual foci included in each focus cluster, obtain the corresponding number of calibration points, and obtain the corresponding calibration point distribution data according to the number of calibration points, and generate a set of calibration points according to the calibration point distribution data corresponding to each focus cluster.

[0078] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. An eye movement tracking model calibration method, characterized in that The method includes: Select a calibration point from the set of calibration points as the current calibration point, display the current calibration point on the human-computer interaction interface, and prompt the user to fixate on the current calibration point. Then, use a plurality of cameras arranged in an array to simultaneously capture the user's eye images, and the number of cameras is two or more; For each camera, construct a virtual cone with the optical axis direction of the camera as the central axis and a preset cone angle. Use the area enclosed by the intersection line of the virtual cone and the captured eye image as the cropping area, and crop the corresponding sub-image from the captured eye image according to the cropping area; Use the sub-images corresponding to each camera to splice into a target eye image for the current calibration point; select one of the other calibration points in the set of calibration points except for the ones that have been traversed as the current calibration point, and return to execute the step of displaying the current calibration point on the human-computer interaction interface until all calibration points have been traversed; perform calibration according to the target eye images corresponding to all calibration points; The determination process of the set of calibration points includes: Filter out the peak hours when the user uses the current device according to the frequency of the user's operation of the current device, and filter out the target eye movement tracking data from the historical eye movement tracking data; For each eye movement tracking data, map the visual focus corresponding to the target eye movement tracking data to the screen of the current device to obtain a set of visual foci; Perform clustering processing on the visual foci included in the set of visual foci to obtain several focus clusters; according to the distribution characteristics of the visual foci included in each focus cluster, obtain the corresponding number of calibration points, and obtain the corresponding calibration point distribution data according to the number of calibration points, and generate a set of calibration points according to the calibration point distribution data corresponding to each focus cluster.

2. The eye movement tracking model calibration method according to claim 1, characterized in that The step of obtaining the corresponding number of calibration points according to the distribution characteristics of the visual foci included in each focus cluster and obtaining the corresponding calibration point distribution data includes: For the current focus cluster among several focus clusters, obtain the center of the current focus cluster and obtain the average value of the connecting line segments between all the included visual foci. Use the center of the current focus cluster as the center of the circle and the average value as the radius to enclose the area as the target area; obtain the number of calibration points corresponding to the current focus cluster according to the proportion of the target area in the total area of the target areas of all visual focus clusters; randomly distribute the calibration points corresponding to the current focus cluster in the corresponding target area to obtain the calibration point distribution data for the current focus cluster.

3. The method for calibrating an eye movement tracking model according to claim 2, wherein When obtaining the number of calibration points corresponding to the current focus cluster according to the proportion of the target area in the total area of the target areas of all visual focus clusters, the method further includes: Judge whether the number of calibration points corresponding to the current focus cluster is less than one. If so, set the number of calibration points corresponding to the current cluster focus to one.

4. A method for calibrating an eye movement tracking model according to claim 1, characterized in that The determination method of the plurality of cameras includes: Broadcast a shooting assistance request to the surroundings, where the shooting assistance request includes the first coordinate of the current device itself; so that the assisting device that receives the shooting assistance request calculates its own optimal shooting range according to its own second coordinate and the optimal shooting distance pre-calibrated by the manufacturer, and responds to the shooting assistance request when the first coordinate is within the optimal shooting range; Use the camera carried by the assisting device that responds to the shooting assistance request as the camera for shooting the user's eye image.

5. A method for calibrating an eye movement tracking model according to claim 4, characterized in that, The process of obtaining the preset cone angle includes: According to the distribution positions of the cameras, evenly divide the eye area corresponding to each camera; Calculate the size of the preset cone angle according to the corresponding eye area and the distance between the camera and the eye.

6. The eye movement tracking model calibration method according to claim 5, characterized in that, The process of stitching the sub-images corresponding to each camera into the target eye image for the current calibration point includes: According to the coordinates of the corresponding camera, convert the pixel points of each sub-image into the same coordinate system, and stitch each sub-image into a complete eye image according to the eye area corresponding to each sub-image, and use the complete eye image as the target eye image.

7. A method for calibrating an eye movement tracking model according to claim 1, characterized in that The process of performing calibration according to the target eye images corresponding to all calibration points includes: Obtain the reflected light spot in the target eye image, smooth the edge of the reflected light spot to obtain a smoothed contour; fit the line of the smoothed contour into an ellipse, use the center point of the ellipse as the corneal reflection center, and perform calibration according to the corneal reflection center and the target eye image.

8. An eye movement tracking model calibration device, characterized in that, The device includes: A display module, configured to select a calibration point from the calibration point set as the current calibration point, display the current calibration point on the human-computer interaction interface, and prompt the user to fixate on the current calibration point, and then use a plurality of cameras arranged in an array to simultaneously shoot the user's eye image, and the number of cameras is two or more; A cropping module, configured to, for each camera, construct a virtual cone with the optical axis direction of the camera as the central axis and a preset cone angle, use the area surrounded by the intersection line of the virtual cone and the captured eye image as the cropping area, and crop the corresponding sub-image from the captured eye image according to the cropping area; A stitching module, configured to stitch the sub-images corresponding to each camera into the target eye image for the current calibration point; select one of the other calibration points in the calibration point set except the calibrated calibration points as the current calibration point, and return to execute the step of displaying the current calibration point on the human-computer interaction interface until all calibration points are traversed; perform calibration according to the target eye images corresponding to all calibration points; the determination process of the calibration point set includes: Filter out the peak period when the user uses the current device according to the frequency of the user operating the current device, and filter out the target eye movement tracking data from the historical eye movement tracking data; For each eye movement tracking data, map the visual focus corresponding to the target eye movement tracking data to the screen of the current device to obtain a visual focus set; Cluster the visual foci included in the set of visual foci to obtain a number of focus clusters; according to the distribution characteristics of the visual foci included in each focus cluster, obtain the corresponding number of calibration points, and obtain the corresponding calibration point distribution data according to the number of calibration points, and generate a calibration point set according to the calibration point distribution data corresponding to each focus cluster.

Images