An artificial intelligence-based eye behavior analysis method and system

Abstract

This invention relates to an artificial intelligence-based method and system for eye behavior analysis. The method includes: radially segmenting two eye images from two horizontal markers to divide the eye images into multiple sub-regions; constructing a sequence of UX tuples based on the sub-regions and comparing it with samples to determine eye behavior. This invention enables real-time identification and response to eye behavior through rapid, continuous decision-making, making it suitable for applications requiring fast response.

Description

[Technical Field]

[0001] This invention belongs to the field of posture retrieval technology based on artificial intelligence, and particularly relates to an eye behavior analysis method and system based on artificial intelligence. [Background Technology]

[0002] With the rapid development of computer hardware and software technologies, multi-modal human-computer interaction (MMHCl) has become a very popular field in human-computer interaction software engineering methods both domestically and internationally. Because MMHCl better aligns with human psychological and physiological habits, it is considered a more natural way of interacting with computers. Artificial intelligence (AI) is a crucial driving force leading a new round of technological revolution and industrial transformation, profoundly changing people's production, lifestyles, and learning methods. AI technology has moved from theoretical research to practical applications, profoundly altering production and lifestyles. In recent years, large-scale AI model technology has developed rapidly, relying on "big data + high computing power + strong algorithms" for training, enhancing the ability to handle complex tasks. AI is shifting from single-point applications to diversified applications, penetrating specific industry scenarios, and accelerating towards a new stage of comprehensive application. Breakthroughs in large-scale models and the rise of generative AI have brought advanced tools for product and process innovation to various industries. The rapid development and widespread application of AI technology are propelling society into a new era of intelligence and digitalization. With continuous technological progress and strong policy support, AI is expected to play a key role in more fields, promoting high-quality economic and social development.

[0003] The concept of eye tracking originated in the 1930s, initially designed for commercial purposes. However, due to limitations in sensor technology and the computing speed of information processors, eye tracking technology encountered many difficulties in its development, resulting in very slow progress. Eye movements accompany almost all human activities, and eye movement information contains important clues that can reveal different human behaviors. Eye tracking is a scientific application technology that allows users to send and receive information without physical contact with the interactive device. In principle, eye tracking mainly studies the acquisition, modeling, and simulation of eye movement information, with wide applications. However, current analysis methods are mainly based on traditional analysis of eye images or methods that do not distinguish the specific categories of each part of the eye image, lacking the expression of eye image information, or lacking an understanding of the characteristics of eye image composition when expressing it. When preprocessing eye images, there is a lack of utilization of eye characteristics, leading to high complexity or poor scalability. Based on the above problems, this invention can identify and respond to eye behaviors in real time by rapidly making continuous eye behavior decisions, and is suitable for application scenarios requiring rapid response. [Summary of the Invention]

[0004] To address the aforementioned problems in the prior art, this invention proposes an artificial intelligence-based eye behavior analysis method and system, the method comprising:

[0005] Step S1: Acquire the eye image and preprocess it; specifically: perform parabolic shape preprocessing and area-based size preprocessing.

[0006] Step S11: Obtain the horizontal marker and the vertical marker based on the horizontal marker; divide the eye image into four regions based on the vertical and horizontal markers, and perform parabolic shape preprocessing and area-based size preprocessing for each region.

[0007] Step S12: Obtaining the horizontal markers specifically involves: segmenting the eye image to obtain the eye region; obtaining the two endpoints of the longest diameter of the eye region as two horizontal markers; constructing an X-axis based on the horizontal markers; and using the point with the furthest vertical distance from the X-axis as two vertical markers; and constructing a Y-axis based on the two vertical markers.

[0008] Preferred method: When the Y-axis formed by the two vertical markers is not perpendicular to the X-axis, select the center point of the projection of the two vertical markers onto the X-axis to construct a Y-axis perpendicular to the X-axis, and take the intersection of the Y-axis and the edge of the eye region as the vertical marker; Of course, there are various ways to adjust and set the vertical markers, which will not be elaborated here.

[0009] Step S13: Divide the eye image into four regions based on the X-axis and Y-axis; connect the horizontal and vertical markers in the same region to form stretching auxiliary lines; determine the first projection height L of the horizontal markers on the Y-axis; determine the second projection height C of the horizontal markers on the X-axis;

[0010] Step S14: Perform parabolic shape preprocessing for each region; specifically: stretch the region along the first or second direction and perform pixel interpolation until the stretching condition is met to complete the shape preprocessing.

[0011] Step S15: Stretch the region along the X-axis toward the outer side of the eye region so that the first projection height L reaches the preset value of the first projection height, so as to complete the area-based size preprocessing.

[0012] Step S2: Starting from the left horizontal marker AL and the right horizontal marker AR in the two horizontal markers A respectively, set U1 left rays and U1 right rays in the part between the upper tangent line formed by them and the lower tangent line formed by them and the upper part of the eye region and the lower part of the eye region; the U1 left rays and U1 right rays divide the eye image into UX sub-regions.

[0013] Step S3: Trialize each sub-region to obtain a triplet corresponding to each sub-region. Arrange all triplets according to the distribution position of the sub-regions to obtain the UX tuple corresponding to the eye region ((I i，0 I i，1 I i，2 )); where: i is the sub-region number; I i，0 I is the number of pixels in sub-region i that belong to the scleral region; i，1 I is the number of pixels in sub-region i that belong to the iris region; i，2 This represents the number of pixels in sub-region i that belong to the pupil region; i = 1 to UX;

[0014] Step S4: Use the N most recent consecutive human eye regions to form the current UX tuple sequence of length N, compare it with the UX tuple sample sequence in the eye behavior sample library, and determine the similarity between the current UX tuple sequence and the UX tuple sample sequence; select the eye behavior corresponding to the UX tuple sample sequence with the smallest similarity as the determined eye behavior.

[0015] Furthermore, the stretching condition is that any point on the edge of the region coincides with a preset parabola passing through the horizontal and vertical markers, or the second projection height C after stretching reaches a preset value for the second projection height.

[0016] Furthermore, the first direction is the direction of the Y-axis toward the outer side of the eye region.

[0017] Furthermore, the second direction is perpendicular to the stretching guide line and towards the outer side of the eye area.

[0018] Furthermore, depending on the similar orientation of the four regions, either the first or second direction can be selected for stretching.

[0019] Furthermore, the left and right rays of the U-shaped strip are evenly distributed.

[0020] Furthermore, an artificial intelligence model is used to classify each pixel in the eye image, so that the pixel belongs to the sclera, iris or pupil respectively.

[0021] An artificial intelligence-based eye behavior analysis system includes: the system being used to implement the artificial intelligence-based eye behavior analysis method.

[0022] An artificial intelligence-based eye behavior analysis platform includes a processor coupled to a memory, the memory storing program instructions, and the artificial intelligence-based eye behavior analysis method is implemented when the program instructions stored in the memory are executed by the processor.

[0023] A computer-readable storage medium includes a program that, when run on a computer, causes the computer to perform the aforementioned AI-based eye behavior analysis method.

[0024] The beneficial effects of this invention include:

[0025] (1) By using parabolic shape preprocessing and area-based size preprocessing, the shape and size of the eye image can be adjusted more accurately to adapt to eye images of different shapes, states and sizes. This allows each segmented sub-region to concentrate more information in the horizontal direction and differentiates the amount of information that different eye positions can bring, so that the information in the eye image can be distributed to different sub-regions as much as possible. This achieves an effective balance between complexity and not losing important information.

[0026] (2) By utilizing the characteristics of eye images, a UX tuple based on triples and its tuple sequence is constructed. Based on existing image segmentation and pixel classification counting, the understanding of the characteristics of eye image composition is used in the expression, thereby transforming the representation of these characteristics into a quantifiable basis, providing a fast basis for continuous eye behavior judgment, and enabling real-time recognition and response to eye behavior. It is suitable for application scenarios that require rapid response.

[0027] (3) The one-way discrete sequence based on spatial information can sensitively detect changes in real-time eye behavior in real-time environment. When it is impossible to obtain effective eye behavior judgment using extensive information, the two-way discrete sequence based on time information reduces the fixed information factor caused by environmental factors. By being sensitive to possible complex factors, it can improve the possibility of matching with samples and improve the stability and effectiveness of eye behavior judgment. [Attached Image Description]

[0028] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to unduly limit the invention. In the drawings:

[0029] Figure 1 This is a schematic diagram of the artificial intelligence-based eye behavior analysis method provided by the present invention.

[0030] Figure 2 This is a schematic diagram of the method for segmenting eye images provided by the present invention.

Detailed Implementation Methods

[0031] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions are merely for explaining the present invention and are not intended to limit the scope of the invention.

[0032] This invention proposes an artificial intelligence-based method and system for analyzing eye behavior, as shown in the attached figure. Figure 1 As shown, the method includes the following steps:

[0033] Step S1: Acquire the eye image and preprocess it; specifically: perform parabolic shape preprocessing and area-based size preprocessing.

[0034] As attached Figure 2 As shown, the steps specifically include the following steps:

[0035] Step S11: Obtain the horizontal marker and the vertical marker based on the horizontal marker; divide the eye image into four regions based on the vertical and horizontal markers, and perform parabolic shape preprocessing and area-based size preprocessing for each region.

[0036] Step S12: Obtaining the horizontal markers specifically involves: segmenting the eye image to obtain the eye region; obtaining the two endpoints of the longest diameter of the eye region as two horizontal markers (A); constructing an X-axis based on the horizontal markers; and using the point with the furthest vertical distance from the X-axis as two vertical markers (B); and constructing a Y-axis based on the two vertical markers. In other words, the axis of the coordinate system created here does not necessarily pass through the center of the eye, but is related to the eye behavior at that time. Figure 2 The image shows the lower right area of ​​the eye region;

[0037] Step S13: Divide the eye image into four regions based on the X-axis and Y-axis; connect the horizontal and vertical markers in the same region to form a stretching guide line (TT′); determine the first projection height L of the horizontal marker on the Y-axis; determine the second projection height C of the horizontal marker on the X-axis;

[0038] Step S14: Perform parabolic shape preprocessing for each region; specifically: stretch the region along the first or second direction and perform pixel interpolation until the stretching condition is met to complete the shape preprocessing.

[0039] Preferably, the stretching condition is that any point on the edge of the region coincides with a preset parabola passing through the horizontal and vertical markers, or the second projection height C after stretching reaches a preset value for the second projection height.

[0040] Preferably, the first direction is the direction from the Y-axis toward the outer side of the eye area, as shown in the attached figure. Figure 2 The first direction is indicated by the middle arrow; the second direction is perpendicular to the stretching guide line and points outwards towards the outer edge of the eye area.

[0041] Preferred method: Select either the first or second direction for stretching based on the similar orientation of the four regions;

[0042] Step S15: Stretch the region along the X-axis toward the outer side of the eye region so that the first projection height L reaches the preset value of the first projection height, so as to complete the area-based size preprocessing.

[0043] Step S2: Starting from the left horizontal marker AL and the right horizontal marker AR in the two horizontal markers A, respectively, set U1 left rays and U1 right rays in the area between the upper tangent line formed by them and the lower tangent line formed by them and the upper part of the eye region, and the lower tangent line formed by them and the lower part of the eye region. The U1 left rays and U1 right rays divide the eye image into UX sub-regions. When the U1 left ray or the U1 right ray acts independently, it divides the eye region into U1+1 sub-regions. After the two intersect, these regions will be further subdivided to form UX irregular regions. The number of these irregular regions is related to the shape of the eye region, the preset value of the projection height, and the division method. The irregular division method allows each segmented sub-region to concentrate more information in the horizontal direction and differentiates the amount of information that different eye positions can bring, so that the information in the eye image can be distributed as much as possible to different sub-regions.

[0044] Preferably, the left and right U-shaped rays are evenly distributed, and the included angle between adjacent rays is the same;

[0045] Preferably, the left and right rays of the U-shaped strip are not uniformly arranged, and the included angles between adjacent rays are not the same;

[0046] Preferably, the steps further include setting U2 upper rays and U2 lower rays respectively, starting from the upper horizontal marker BU and the lower horizontal marker BB in the two vertical marker positions B, in the portion between them and the left tangent line of the left part of the eye region and the right tangent line of the right part of the eye region; the U2 upper rays and U2 lower rays divide the eye image into UY sub-regions; of course, it is also possible to start from the upper horizontal marker BU and the lower horizontal marker BB in the two vertical marker positions B, and start from the left horizontal marker AL and the right horizontal marker AR in the two horizontal marker positions A as the division center, thereby forming a more complex division method, which will not be described in detail; considering that for the eye image, not every region is equally important, therefore, uniform grid division cannot achieve an effective balance between complexity and not losing importance. This invention uses a radial segmentation method based on uniform angle to divide the grid, ultimately forming UX or UY sub-regions; the following uses UX sub-regions as an example;

[0047] Preferred values: U1 and U2 are preset values;

[0048] Preferably, the upper or lower rays are spaced at a predetermined uniform angle; the uniform angle is 0.1 to 1°.

[0049] Step S3: Trialize each sub-region to obtain a triplet corresponding to each sub-region. Arrange all triplets according to the distribution position of the sub-regions to obtain the UX tuple corresponding to the eye region ((I i，0 I i，1 I i，2 )); where: i is the sub-region number; I i，0 I is the number of pixels in sub-region i that belong to the scleral region; i，1 I is the number of pixels in sub-region i that belong to the iris region; i，2 This represents the number of pixels in sub-region i that belong to the pupil region; i = 1 to UX;

[0050] Preferably, the distribution positions are arranged in a top-to-bottom and left-to-right order.

[0051] Preferably, the distribution location is a fixed distribution location;

[0052] Preferred method: Use an artificial intelligence model to segment the eye image into sclera, iris and pupil regions; during the segmentation process, each pixel is classified, so that the pixel belongs to the sclera (the white part of the eye, i.e. the white area around the outer edge of the eyeball), the iris (the colored part around the pupil, responsible for controlling the size of the pupil to regulate the amount of light entering the eye) or the pupil (the black circular hole in the center of the iris through which light enters the inside of the eyeball).

[0053] Preferably, the artificial intelligence model mainly consists of an encoder and a decoder, such as a U-Net network; the eye image is converted into a single-channel grayscale image through a 1×1 convolution and then fed into the encoder;

[0054] Alternatively: Step S3 specifically involves: binarizing each sub-region to obtain a tuple corresponding to each sub-region, and arranging all the tuples according to the distribution position of the sub-regions to obtain the UX tuple corresponding to the eye image ((I i，0 I i，1 )); where: i is the sub-region number; I i，0 I is the number of pixels in sub-region i that belong to the iris region; i，1 It is the number of pixels in sub-region i that belong to the pupil region; converting ternary to binary can reduce computational complexity;

[0055] Step S4: Use the N most recent consecutive human eye regions to form the current UX tuple sequence of length N, compare it with the UX tuple sample sequence in the eye behavior sample library, and determine the similarity between the current UX tuple sequence and the UX tuple sample sequence; select the eye behavior corresponding to the UX tuple sample sequence with the smallest similarity as the determined eye behavior;

[0056] Preferably, continuous human eye images are obtained from human eye image videos, and a human eye region including only the eye area is obtained based on the human eye images;

[0057] Preferred method: Pre-construct UX tuple sample sequences corresponding to different eye behaviors under different viewpoints, environments, and user eye categories to build an eye behavior sample library;

[0058] Preferred method: Use Euclidean distance to calculate the similarity between the current UX tuple sequence and the UX tuple sample sequence;

[0059] Step S4 specifically includes the following steps:

[0060] Step S41: The current sequence of UX tuples is represented as The UX tuple sample sequence is represented as

[0061] Step S42: Calculate the intra-sequence discreteness of the current UX tuple sequence and the UX tuple sample sequence respectively to obtain the current one-way discreteness sequence. The unidirectional discreteness sequence of the sample sequence of UX tuples Specifically, the current unidirectional discreteness sequence and the sample unidirectional discreteness sequence are calculated using the following formulas (1) and (2); where: i, j = 3 ~ UX;

[0062]

[0063]

[0064] Step S43: Normalize the entropy of the current one-way discrete sequence and the sample one-way discrete sequence; specifically, use the following formulas (3) and (4) to obtain the current one-way discrete sequence after entropy normalization.

[0065] and sample unidirectional discrete sequence in: These are the mean values ​​of the 1st, 2nd, and 3rd elements in the triplet within the current unidirectional discreteness sequence; These are the mean values ​​of the 1st, 2nd, and 3rd elements in the triplet within the sample unidirectional dispersion sequence, respectively.

[0066]

[0067]

[0068] Step S44: Calculate the current one-way discreteness sequence after entropy normalization. and sample unidirectional discrete sequence The similarity between them is used as the similarity between the current UX tuple sequence and the UX tuple sample sequence; specifically, the similarity sim is calculated using the following formula (5). I，J ;

[0069]

[0070] Alternative: The similarity sim is calculated using the following formula (6). I，J Where: ε is the minimum value, for example: ε = 0.001;

[0071]

[0072] Preferably, step S4 further includes step S45: selecting the eye behavior corresponding to the UX tuple sample sequence with the lowest similarity, and repeating step S4 to make continuous eye behavior judgments; when all similarities are greater than the similarity threshold, proceed to step S5; otherwise, repeat step S4 to make continuous eye behavior judgments.

[0073] Alternative: When all similarities are greater than the similarity threshold and the latest eye behavior decision time is reached, proceed to step S5; at this time, when the eye behavior decision time is reached and the corresponding decision must be made, proceed to the next step to perform a bidirectional discrete sequence based on time information, reduce the fixed information factor caused by environmental factors, and improve the possibility of matching with the sample by reducing the sensitivity to possible complex environmental factors.

[0074] Step S5: Using the current UX tuple sequence of length N and the UX tuple sample sequence, construct the current bidirectional discreteness sequence based on time information, respectively. and sample two-way discrete sequence The similarity between the two is taken as the similarity between the current UX tuple sequence and the UX tuple sample sequence; the eye behavior corresponding to the UX tuple sample sequence with the smallest similarity is selected as the determined eye behavior; unlike spatial information, when considering temporal information, the environmental factor information contained in the time information before and after the eye behavior is the most complete. Therefore, determining the discrete sequence based on bidirectional factors can most effectively improve stability.

[0075] Step S5 specifically includes the following steps:

[0076] Step S51: The current sequence of UX tuples of length N is represented as follows A sequence of N-length UX tuple samples is represented as follows:

[0077] Step S52: Calculate the time dispersion of the current UX tuple sequence and the UX tuple sample sequence respectively to obtain the current bidirectional dispersion sequence. The sample bidirectional discreteness sequence of UX tuple samples. Specifically, the current bidirectional dispersion sequence and the sample bidirectional dispersion sequence are calculated using the following formulas (7) and (8); where: i, j = 1 ~ UX, n = 2 ~ N-1;

[0078]

[0079]

[0080] Step S53: Normalize the entropy of the current bidirectional discrete sequence and the sample bidirectional discrete sequence; specifically, use the following formulas (9) and (10) to obtain the current bidirectional discrete sequence after entropy normalization. and sample two-way discrete sequence in:

[0081] These are the mean values ​​of the 1st, 2nd, and 3rd elements in the triplet within the current bidirectional dispersion sequence; These are the mean values ​​of the 1st, 2nd, and 3rd elements in the triplet within the sample bidirectional dispersion sequence, respectively.

[0082]

[0083]

[0084]

[0085] Step S54: Calculate the current bidirectional discreteness sequence after entropy normalization. and sample two-way discrete sequence The similarity between them is used as the similarity between the current UX tuple sequence and the UX tuple sample sequence; specifically, the similarity sim is calculated using the following formula (11). I，J ;

[0086]

[0087] Step S55: Select the eye behavior corresponding to the UX tuple sample sequence with the lowest similarity as the determined eye behavior; repeat steps S4 and / or S5 to make continuous eye behavior decisions.

[0088] Based on the same inventive concept, the present invention also provides an artificial intelligence-based eye behavior analysis system, the system being used to implement the above-mentioned artificial intelligence-based eye behavior analysis method;

[0089] Based on the same inventive concept, the present invention also provides an artificial intelligence-based eye behavior analysis server, the server being used to implement the above-mentioned artificial intelligence-based eye behavior analysis method;

[0090] Based on the same inventive concept, the present invention also provides an artificial intelligence-based eye behavior analysis device, the device being used to implement the above-mentioned artificial intelligence-based eye behavior analysis method;

[0091] Based on the same inventive concept, the present invention also provides an artificial intelligence-based eye behavior analysis platform, which is used to implement the above-mentioned artificial intelligence-based eye behavior analysis method;

[0092] A computer program (also referred to as a program, software, software application, script, or code) can be written in any form of programming language, including assembly or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment. A computer program may, but does not necessarily, correspond to a file in a file system. A program can be stored as part of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to said program, or in multiple co-located files (e.g., a file storing one or more modules, subroutines, or code portions). A computer program can be deployed to execute on a single computer or on multiple computers located at a single site or distributed across multiple sites and interconnected by a communications network.

[0093] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0094] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0095] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0096] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. An artificial intelligence-based eye behavior analysis method, characterized in that, The method includes: Step S1: Acquire the eye image and preprocess it; specifically: perform parabolic shape preprocessing and area-based size preprocessing. Step S11: Obtain the horizontal flag bit, and obtain the vertical flag bit based on the horizontal flag bit; The eye image is divided into four regions based on vertical and horizontal markers. For each region, parabolic shape preprocessing and area-based size preprocessing are performed. Step S12: Obtaining the horizontal markers specifically involves: segmenting the eye image to obtain the eye region; obtaining the two endpoints of the longest diameter of the eye region as two horizontal markers; constructing an X-axis based on the horizontal markers; and using the point with the furthest vertical distance from the X-axis as two vertical markers; and constructing a Y-axis based on the two vertical markers. Step S13: Divide the eye image into four regions based on the X and Y axes; connect the horizontal and vertical markers in the same region to form stretching guide lines; Determine the first projection height L of the horizontal marker on the Y-axis; Determine the second projection height C of the vertical marker position on the X-axis; Step S14: Perform parabolic shape preprocessing for each region; specifically: stretch the region along the first or second direction and perform pixel interpolation until the stretching condition is met to complete the shape preprocessing; the stretching condition is that any point on the edge of the region coincides with a preset parabola passing through the horizontal and vertical markers. Step S15: Stretch the region along the X-axis toward the outer side of the eye region so that the first projection height L reaches the preset value of the first projection height, so as to complete the area-based size preprocessing. Step S2: Starting from the left horizontal marker AL and the right horizontal marker AR in the two horizontal markers A respectively, set U1 left rays and U1 right rays in the part between the upper tangent line formed by them and the lower tangent line formed by them and the upper part of the eye region and the lower part of the eye region; the U1 left rays and U1 right rays divide the eye image into UX sub-regions. Step S3: Ternarize each sub-region to obtain a triplet corresponding to each sub-region. Arrange all triplets according to the distribution position of the sub-regions to obtain the UX tuple ((Ii, 0, Ii, 1, Ii, 2)) corresponding to the eye region; where: i is the sub-region number; Ii, 0 is the number of pixels in sub-region i belonging to the sclera region; Ii, 1 is the number of pixels in sub-region i belonging to the iris region; Ii, 2 is the number of pixels in sub-region i belonging to the pupil region; i = 1 to UX; Step S4: Use the N most recent consecutive human eye regions to form the current UX tuple sequence of length N, compare it with the UX tuple sample sequence in the eye behavior sample library, and determine the similarity between the current UX tuple sequence and the UX tuple sample sequence; select the eye behavior corresponding to the UX tuple sample sequence with the smallest similarity as the determined eye behavior.

2. The eye behavior analysis method based on artificial intelligence according to claim 1, characterized in that, The first direction is the direction of the Y-axis toward the outer side of the eye area.

3. The eye behavior analysis method based on artificial intelligence according to claim 2, characterized in that, The second direction is perpendicular to the stretching guide line and towards the outer side of the eye area.

4. The eye behavior analysis method based on artificial intelligence according to claim 3, characterized in that, Depending on the similar orientation of the four areas, choose either the first or second direction for stretching.

5. The eye behavior analysis method based on artificial intelligence according to claim 4, characterized in that, The left and right rays of U1 are evenly distributed.

6. The eye behavior analysis method based on artificial intelligence according to claim 5, characterized in that, An artificial intelligence model is used to classify each pixel in an eye image, so that each pixel belongs to the sclera, iris, or pupil.

7. An artificial intelligence-based eye behavior analysis system, characterized in that, include: The system is used to implement the artificial intelligence-based eye behavior analysis method according to any one of claims 1-4.

8. An artificial intelligence-based eye behavior analysis platform, characterized in that, The system includes a processor coupled to a memory, the memory storing program instructions, which, when executed by the processor, implement the artificial intelligence-based eye behavior analysis method according to any one of claims 1-4.

9. A computer-readable storage medium, characterized in that, Includes a program that, when run on a computer, causes the computer to perform an artificial intelligence-based eye behavior analysis method as described in any one of claims 1-4.

Images