Computer screen control method and system

Abstract

The invention relates to the technical field of screen control, in particular to a computer screen control method and system. The method comprises the following steps: collecting a computer lens image to carry out scene modeling, generating a three-dimensional usage scene model, carrying out scene privacy evaluation and generating a privacy degree parameter, analyzing a peep-proof demand according to the privacy degree parameter, generating a screen peep-proof demand, carrying out portrait positioning when peep-proof is needed, generating a portrait positioning set, and carrying out image processing according to the portrait positioning set. And calculating a virtual contact distance based on the portrait positioning set, performing contact sight simulation to obtain a figure sight track, analyzing a screen visible range according to the sight track and the virtual contact distance, and dynamically adjusting the screen brightness to realize the peep-proof control method. According to the invention, a safer computer screen control method is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of screen control, and particularly to a computer screen control method and system. Background Art

[0002] Computers have become an indispensable tool in people's daily life and work. However, while being widely used, the problem of users' privacy and security has gradually emerged. Especially in public places and shared office environments, users' screen information is easily stolen by others, resulting in the risk of information leakage and privacy infringement. Traditional privacy protection measures often fail to meet the requirements in a dynamic environment. There is an urgent need for a more intelligent and flexible screen control method to address this issue. Existing technologies mainly rely on physical occlusion or fixed anti-peeping filters for screen anti-peeping. The limitation of this method is that it cannot be adjusted in real time according to the changes in the actual scenario and the needs of users, and cannot effectively respond to different usage scenarios and user behaviors. Especially in a complex multi-user environment, it is difficult to ensure that the privacy of each user is fully protected, resulting in still existing information security risks. Summary of the Invention

[0003] Based on this, it is necessary to provide a computer screen control method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, a computer screen control method includes the following steps:

[0005] Step S1: Collect computer lens images; perform scene modeling on the computer lens images to obtain a three-dimensional usage scene model; perform scene privacy evaluation on the three-dimensional usage scene model to generate a scene privacy degree parameter;

[0006] Step S2: Analyze the anti-peeping requirements of the three-dimensional usage scene model according to the scene privacy degree parameter to generate a screen anti-peeping requirement; perform human portrait detection on the computer lens images based on the screen anti-peeping requirement, extract the spatial position and line-of-sight direction of the person, and generate a human portrait positioning set;

[0007] Step S3: Calculate the virtual contact distance of the person for the three-dimensional usage scene model based on the human portrait positioning set to generate a person's virtual contact distance; perform contact line-of-sight simulation according to the person's virtual contact distance to obtain the person's line-of-sight trajectory;

[0008] Step S4: Analyze the visible screen range of the computer screen according to the person's line-of-sight trajectory and the person's virtual contact distance to generate a person's visible screen range; dynamically adjust the brightness of the computer screen based on the screen anti-peeping requirement and the person's visible screen range to perform computer screen anti-peeping control.

[0009] By collecting computer lens images and monitoring the user's surrounding environment in real time, the present invention can timely obtain changes in the usage scenario. The created three-dimensional usage scenario model provides a profound understanding of the spatial layout. The privacy degree parameter generated by the scenario privacy evaluation provides a scientific basis for subsequent anti-peeping requirement analysis, ensuring a more accurate judgment of anti-peeping requirements. The process of generating the portrait positioning collection enables the system to identify and track the specific position and dynamics of the user, laying a foundation for further contact distance calculation. The virtual contact distance calculation is based on the actual distance between the user and the screen, optimizing the user's visual experience. The contact line of sight simulation clearly depicts the user's line of sight trajectory, providing clear data support for the screen visible range analysis. In the screen visible range analysis, by combining the user's line of sight trajectory and the virtual contact distance, potential leakage points of privacy information can be accurately defined. The function of dynamically adjusting the computer screen brightness enables the screen to automatically adapt to different environments and requirements, effectively reducing the risk of others peeping at the screen content. The operation of the entire system improves the intelligent level of computer use, enabling users to enjoy higher privacy protection and information security in various environments, promoting the wide popularization and implementation of intelligent screen control technology in practical applications, effectively addressing the increasingly severe privacy protection challenges in modern society, and enhancing the user's sense of information security and usage experience in a changing environment.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: Collect computer lens images; perform high-precision image segmentation on the computer lens images to obtain a set of scene elements;

[0012] Step S12: Perform scene depth mapping through the set of scene elements to obtain scene depth parameters;

[0013] Step S13: Based on the scene depth parameters, perform spatial point cloud reconstruction on the set of scene elements to generate a three-dimensional spatial point cloud; apply grid processing to the three-dimensional spatial point cloud to obtain a three-dimensional usage scenario model;

[0014] Step S14: Perform spatial density analysis on the three-dimensional usage scenario model to obtain scene activity hotspots; identify scene identifiers in the scene activity hotspots;

[0015] Step S15: Based on the scene identifiers, distinguish the types of the three-dimensional usage scenario model to obtain the computer usage scenario type; perform privacy evaluation according to the computer usage scenario type to generate a scene privacy degree parameter.

[0016] Through high-precision image segmentation, the present invention can accurately identify and extract scene elements. The generated scene depth parameters provide an important basis for subsequent spatial point cloud reconstruction. The reconstruction and meshing of the spatial point cloud improve the fineness and realism of the 3D model. Spatial density analysis can effectively identify active hotspots in the scene, thereby optimizing the computer's usage environment. Identifying the identifiers in the scene activity hotspots provides data support for subsequent scene type differentiation. The differentiation of scene types enables the system to perform more accurate privacy assessment. The generated scene privacy level parameters provide a scientific basis for user privacy protection, overall enhancing the intelligent level of computer screen control, ensuring effective information security management in different usage scenarios, and meeting the increasing demand for privacy protection in modern society.

[0017] Preferably, step S15 includes:

[0018] Step S151: Perform topological deconstruction on the 3D usage scene model based on the scene identifier to obtain a scene topological network;

[0019] Step S152: Perform multi-scale convolutional transformation on the scene topological network to obtain scene feature mapping data;

[0020] Step S153: Identify the computer usage scene type according to the preset application scene atlas for the scene feature mapping data;

[0021] Step S154: Perform a privacy risk score on the computer usage scene class to obtain a scene risk score;

[0022] Step S155: Perform a privacy assessment on the scene risk score based on the preset scene risk score threshold to generate a scene privacy level parameter.

[0023] The scene topological network obtained by topological deconstruction in the present invention provides a deep understanding of spatial relationships. The generation of scene feature mapping data allows capturing subtle changes in the scene at multiple scales. The combination of the application scene atlas enables efficient scene type identification. The clear classification of the computer usage scene type provides a basis for privacy risk scoring. The generation of the scene risk score can quantify the privacy risk level. Performing a privacy assessment based on the preset risk score threshold ensures the scientificity and accuracy of the assessment. The generated scene privacy level parameters provide targeted information protection measures for users, overall enhancing the intelligence and adaptability of the computer screen control system in privacy management and meeting the privacy needs of users in complex usage environments.

[0024] Preferably, step S2 includes the following steps:

[0025] Step S21: Compare the scene privacy level parameter with the threshold based on the preset privacy level threshold to obtain a privacy determination result;

[0026] Step S22: Map the anti-peeping requirement to the three-dimensional usage scenario model according to the privacy determination result to generate the screen anti-peeping requirement;

[0027] Step S23: Perform human face detection on the computer lens image based on the screen anti-peeping requirement, correct the illumination factor of the computer lens image to obtain an evenly illuminated image; Detect the human face area of the evenly illuminated image; Perform key point positioning on the human face area to generate human face key point data;

[0028] Step S24: Estimate the initial human face positioning according to the human face pose data to obtain the initial human face positioning; Perform positioning integration on the initial human face positioning to generate a human face positioning set.

[0029] The present invention realizes the accurate judgment of the scene privacy degree by comparing with a preset privacy degree threshold. The privacy determination result provides a basis for the subsequent anti-peeping requirement mapping. The generation of the screen anti-peeping requirement ensures the adoption of effective privacy protection measures when necessary. The correction of the illumination factor improves the accuracy of human face area detection. The generation of the evenly illuminated image optimizes the environment condition for face recognition. The key point positioning of the human face area provides detailed data support for face recognition and tracking. The estimation of the initial human face positioning ensures the rapid response of the human face positioning. The application of the positioning integration technology improves the accuracy and stability of the human face positioning. Overall, it improves the intelligence and response ability of the computer screen control system in terms of privacy protection, and enhances the user's privacy sense of security in different environments.

[0030] Preferably, step S23 includes the following steps:

[0031] Step S231: Perform human face detection on the computer lens image based on the screen anti-peeping requirement, and perform brightness histogram equalization on the computer lens image to obtain a brightness-equalized image;

[0032] Step S232: Enhance the contrast of the brightness-uniform image to generate an enhanced lens image; Perform illumination compensation on the enhanced lens image to obtain an evenly illuminated image;

[0033] Step S233: Perform multi-scale human face detection on the evenly illuminated image to generate a human face candidate area; Perform boundary redrawing processing on the human face candidate area to obtain the human face area;

[0034] Step S234: Extract feature points from the human face area to obtain human face feature point data; Perform grid deformation mapping on the human face feature point data to obtain grid deformation data;

[0035] Step S235: Perform key point regression on the grid deformation data to obtain key point positioning data; Perform coordinate system conversion on the key point positioning data to obtain human face key point data.

[0036] The present invention optimizes the image brightness through brightness histogram equalization. The generated brightness - equalized image improves the basis for subsequent processing. The enhancement of the generated lens image improves the detail expressiveness of the image. The contrast enhancement makes the facial features more prominent, which is helpful for subsequent light compensation operations. The generation of the evenly - lit image ensures the face detection effect under different lighting conditions. The adoption of multi - scale face detection technology improves the flexibility and accuracy of face recognition. The boundary redrawing process of the face candidate region optimizes the face positioning accuracy. The extraction of feature points provides important data support for face recognition. The application of grid deformation mapping enhances the adaptability of facial features. The key - point regression technology improves the accuracy of face key - point positioning. Through coordinate system conversion, the consistency of data is achieved. Overall, the intelligent level of the computer screen control system in privacy protection and face recognition is improved, and the adaptability of the system to complex environments is enhanced.

[0037] Preferably, the virtual contact distance calculation for the three - dimensional usage scenario model based on the portrait positioning set in step S3 includes:

[0038] Convert the three - dimensional usage scenario model into a positioning coordinate system;

[0039] Project the portrait positioning set into the positioning coordinate system to obtain user space coordinates;

[0040] Estimate the neighbor volume based on the user space coordinates to generate an interaction space volume;

[0041] Perform gradient projection conversion on the interaction space volume to generate multi - level contact probability parameters;

[0042] Perform boundary - constraint filtering on the multi - level contact probability parameters to obtain contact boundary data;

[0043] Calculate the contact distance for the contact boundary data to generate the virtual contact distance of the person.

[0044] The present invention realizes the standardization of space positioning by converting the three - dimensional usage scenario model into a positioning coordinate system. The generation of user space coordinates provides an accurate basis for subsequent interaction analysis. The neighbor volume estimation improves the understanding depth of the interaction space. The generation of the interaction space volume provides a quantitative basis for the relationship between the user and the environment. The application of gradient projection conversion enables the multi - level analysis of contact probability parameters to be realized. The boundary - constraint filtering technology ensures the accuracy of contact boundary data. The implementation of contact distance calculation realizes the accurate evaluation of the virtual contact distance of the person. Overall, the intelligent level of the computer screen control system in user interaction and privacy protection is improved, and the adaptability of the system to dynamic interaction scenarios is enhanced.

[0045] Preferably, the contact line-of-sight simulation according to the virtual contact distance of the person in step S3 includes:

[0046] Derive the actual interval of the person based on the virtual contact distance of the person to obtain the actual contact distance of the person;

[0047] Update the three-dimensional distance of the portrait positioning set based on the actual contact distance of the person to generate the three-dimensional contact distance of the person;

[0048] Perform line-of-sight projection simulation according to the three-dimensional contact distance of the person. During the line-of-sight projection process, set the projection angle to 25 - 35 degrees and the maximum projection distance to 2.5 - 3.5 meters to obtain the initial line-of-sight vector;

[0049] Perform difference smoothing processing on the initial line-of-sight vector to generate a continuous line-of-sight curve;

[0050] Perform three-dimensional trajectory reconstruction on the continuous line-of-sight curve to obtain the line-of-sight trajectory of the person.

[0051] The present invention realizes an accurate understanding of the user's interval by deriving the actual contact distance of the person. The three-dimensional distance update enhances the spatial expressiveness of the portrait positioning set. The generation of the three-dimensional contact distance of the person provides a real basis for line-of-sight projection. The obtaining of the initial line-of-sight vector lays a foundation for subsequent line-of-sight simulation. The difference smoothing processing improves the smoothness of the line of sight. The generated continuous line-of-sight curve reflects the natural behavior of the person's dynamics. The application of three-dimensional trajectory reconstruction enables the visualization of the person's line-of-sight trajectory, overall enhancing the realism and immersion of the computer screen control system in the user interaction experience and enhancing the system's adaptability to line-of-sight simulation in complex scenarios.

[0052] Preferably, the screen visible range analysis of the computer screen according to the line-of-sight trajectory of the person and the virtual contact distance of the person in step S4 includes:

[0053] Perform spatial projection conversion on the line-of-sight trajectory of the person to obtain the screen plane projection, where the width of the screen is set to 1900 - 1940 pixels, the height is 1060 - 1100 pixels, and the parameter settings during the projection process are a focal length of 45mm - 55mm and a projection angle of 55 - 65 degrees;

[0054] Perform cross-correlation on the screen plane projection and the virtual contact distance of the person to generate a viewing distance combination index;

[0055] Construct a frustum model based on the viewing distance combination index; perform field-of-view generalization processing on the frustum model to obtain the three-dimensional field-of-view range;

[0056] Perform plane intercept mapping on the three-dimensional field-of-view range to generate a screen field-of-view coverage matrix;

[0057] The character's perspective is inferred based on the screen field of view coverage matrix to generate the character's visible screen range.

[0058] The present invention realizes the effective association between the sight line trajectory of the character and the screen through spatial projection conversion, the generation of the screen plane projection provides basic data for the sight distance combination index, the application of cross-correlation promotes the formation of the sight distance combination index, the construction of the cone model provides a three-dimensional framework for the analysis of the screen visible range, the field of view generalization processing improves the understanding depth of the field of view range, the generation of the three-dimensional field of view range ensures a comprehensive evaluation of the screen coverage, the plane interception mapping optimizes the visualization effect of the screen field of view coverage matrix, the character perspective inversion technology realizes the accurate recognition of the visible screen range, and the overall improvement of the computer screen control system in terms of user interaction and information display. The intelligent level, and the ability of the system to adapt to dynamic usage scenarios are enhanced.

[0059] Preferably, the step S4 of dynamically adjusting the brightness of the computer screen based on the screen anti-peeping requirements and the visible screen range of the person includes:

[0060] The visible screen range of the characters is screened to the maximum extent to obtain the main character's perspective, where the main character's perspective is a viewing angle range of 90-110° and a maximum viewing distance of 1.8-2.2 meters; based on the main character's perspective, the visible screen range of the characters is extracted to obtain the visible screen range of other characters, where the visible screen range of other characters is 70-90° and a maximum viewing distance of 1.6-2.0 meters;

[0061] Based on the screen anti-peeping requirements, the visible screen range of other characters is cut to expose the screen, with a cutting accuracy of 5-10 pixels, to generate exposed screen units;

[0062] Perform brightness gradient design on the exposed screen unit to obtain a brightness adjustment gradient;

[0063] Performing brightness adjustment simulation on a computer screen based on a brightness adjustment gradient to generate brightness of the computer screen at various levels;

[0064] The exposed screen unit is gradually controlled by adjusting the brightness of the computer screen at different levels. The maximum brightness change is set between 0-255. The unit brightness near the edge of the screen is set to 80-120cd, and the unit brightness in the center area is set to 180-220cd to implement computer screen anti-peeping control.

[0065] The present invention realizes the clear positioning of the main character's perspective through the maximum range screening, provides a basis for the subsequent analysis of the screen visible range, enhances the understanding of information dissemination in the environment during the process of extracting the visible screen ranges of other characters, generates exposed screen units based on the screen anti-peeping requirements, the brightness gradient design ensures the reasonable adjustment of the brightness in different regions, the generated brightness adjustment gradient provides a basis for detailed brightness control, the implementation of the brightness adjustment simulation enables the computer screen brightness to adapt to the changing environmental requirements, the gradual control of the exposed screen units by the brightness of each level of the computer screen achieves a dynamic anti-peeping effect, overall improving the intelligent level of the computer screen control system in protecting user privacy and information security, and enhancing the system's adaptability to complex interaction scenarios.

[0066] The present invention also provides a computer screen control system for executing the computer screen control method as described above. The computer screen control system includes:

[0067] An image acquisition module for acquiring computer lens images; performing scene modeling on the computer lens images to obtain a three-dimensional usage scene model; performing scene privacy assessment on the three-dimensional usage scene model to generate a scene privacy degree parameter;

[0068] An anti-peeping analysis module for performing anti-peeping requirement analysis on the three-dimensional usage scene model according to the scene privacy degree parameter to generate a screen anti-peeping requirement; performing portrait detection on the computer lens images based on the screen anti-peeping requirement to extract the spatial positions and line-of-sight directions of the characters, and generating a portrait positioning collection;

[0069] A contact calculation module for performing virtual contact distance calculation on the three-dimensional usage scene model based on the portrait positioning collection to generate a character virtual contact distance; performing contact line-of-sight simulation according to the character virtual contact distance to obtain the character line-of-sight trajectory;

[0070] A brightness control module for performing screen visible range analysis on the computer screen according to the character line-of-sight trajectory and the character virtual contact distance to generate a character visible screen range; dynamically adjusting the brightness of the computer screen based on the screen anti-peeping requirement and the character visible screen range to perform computer screen anti-peeping control.

[0071] Through the real-time monitoring of the image acquisition module, the system can timely obtain and analyze the surrounding environment. The created three-dimensional usage scenario model provides a comprehensive perspective for understanding the spatial layout. The privacy degree parameters generated by the scene privacy assessment provide a reliable basis for the subsequent anti-peeping requirement analysis. The anti-peeping analysis module can accurately identify when anti-peeping is needed. The function of generating a portrait positioning collection enables the system to effectively track the user's position. The virtual contact distance calculation of the contact calculation module optimizes the interaction between the user and the screen. The contact line-of-sight simulation provides a clear description of the user's line-of-sight trajectory. The brightness control module combines the line-of-sight trajectory and the virtual contact distance to accurately analyze the visible range of the screen. The ability to dynamically adjust the screen brightness enables the system to automatically adapt to different environments, reducing the risk of being peeped by others. The overall system improves the intelligent level of computer use, ensures that users enjoy higher privacy protection and information security in a changing environment, promotes the wide application of intelligent screen control technology, meets the growing demand for information security in modern society, and enhances the sense of security and trust of users during computer use. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a schematic flowchart of the steps of a computer screen control method;

[0073] Figure 2 It is a schematic flowchart of the detailed implementation steps of step S2;

[0074] The realization, functional features and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0075] The technical method of the present invention will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0076] In addition, the drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, so the repeated description of them will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0077] It should be understood that although terms such as "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0078] To achieve the above object, please refer to Figures 1 to 2 , a computer screen control method, comprising the following steps:

[0079] Step S1: Collect computer lens images; perform scene modeling on the computer lens images to obtain a three-dimensional usage scene model; perform scene privacy evaluation on the three-dimensional usage scene model to generate a scene privacy degree parameter;

[0080] Step S2: Analyze the anti-peeping requirements of the three-dimensional usage scene model according to the scene privacy degree parameter to generate a screen anti-peeping requirement; perform human portrait detection on the computer lens images based on the screen anti-peeping requirement, extract the human spatial position and line of sight direction, and generate a human portrait positioning set;

[0081] Step S3: Calculate the virtual contact distance of the three-dimensional usage scene model based on the human portrait positioning set to generate a human virtual contact distance; perform contact line of sight simulation according to the human virtual contact distance to obtain the human line of sight trajectory;

[0082] Step S4: Analyze the visible screen range of the computer screen according to the human line of sight trajectory and the human virtual contact distance to generate a human visible screen range; dynamically adjust the brightness of the computer screen based on the screen anti-peeping requirement and the human visible screen range to perform computer screen anti-peeping control.

[0083] The present invention monitors the user's surrounding environment in real time by collecting computer lens images, can timely obtain changes in the usage scenario, the created three-dimensional usage scenario model provides a profound understanding of the spatial layout, the privacy degree parameter generated by the scenario privacy evaluation provides a scientific basis for the subsequent anti-peeping requirement analysis, ensures a more accurate judgment of the anti-peeping requirements, the process of generating the portrait positioning collection enables the system to identify and track the specific position and dynamics of the user, lays a foundation for further contact distance calculation, the virtual contact distance calculation is based on the actual distance between the user and the screen, optimizes the user's visual experience, the contact line of sight simulation clearly depicts the user's line of sight trajectory, provides clear data support for the screen visible range analysis, in the screen visible range analysis, combining the user's line of sight trajectory and the virtual contact distance can accurately define potential leakage points of privacy information, the function of dynamically adjusting the computer screen brightness enables the screen to automatically adapt under different environments and requirements, effectively reduces the risk of peeping at the screen content by others, the operation of the entire system improves the intelligent level of computer use, enables the user to enjoy higher privacy protection and information security guarantee in various environments, promotes the wide popularization and implementation of intelligent screen control technology in practical applications, effectively responds to the increasingly severe privacy protection challenges in modern society, and improves the user's sense of information security and usage experience in a changing environment.

[0084] In an embodiment of the present invention, the computer screen control method includes the following steps:

[0085] Step S1: Collect computer lens images; perform scene modeling on the computer lens images to obtain a three-dimensional usage scenario model; perform scene privacy evaluation on the three-dimensional usage scenario model to generate a scene privacy degree parameter;

[0086] In this embodiment, image data around the computer camera lens is captured in real time by a high-definition camera. The image capture frequency is set to 30 frames per second to ensure that the images have sufficient details for subsequent processing. For each captured frame of image, image preprocessing algorithms are applied, including denoising and grayscale conversion, to facilitate the accuracy of subsequent scene modeling. Then, a deep learning model, especially an image recognition algorithm based on a convolutional neural network (CNN), is used to process each frame of image to identify the objects and environmental information in the image. Next, 3D reconstruction technology is utilized to generate a 3D scene model of the computer camera lens through multi-view image synthesis. This model can accurately reconstruct the physical space where the lens is located and provide a basis for subsequent scene privacy evaluation. During scene privacy evaluation, based on the results of object and face recognition, a specific algorithm is applied to calculate the probability distribution of private areas in the space, generating a scene privacy degree parameter, which reflects the visible range size and privacy characteristics of sensitive areas in the scene. For example, in an office environment, if there are many desks or people in the area facing the computer camera lens, the parameter value is relatively high; on the contrary, if the lens only faces an empty area, the privacy degree is relatively low.

[0087] Step S2: Analyze the anti-peeping requirements for the 3D usage scene model according to the scene privacy degree parameter to generate screen anti-peeping requirements; perform human portrait detection on the computer camera lens image based on the screen anti-peeping requirements, extract the spatial positions and line-of-sight directions of the people, and generate a human portrait positioning set.

[0088] In this embodiment, after the scene privacy degree parameter is generated, the anti-peeping requirements analysis link is entered next. At this time, according to the privacy degree parameter, it is judged whether screen anti-peeping control is required through a set threshold. If the privacy degree is high, it is considered that anti-peeping is required, and the human portrait positioning processing flow is entered. In this process, image segmentation technology and facial feature extraction methods are used to detect the human face or portrait in the computer camera lens image, and combined with the pose estimation algorithm of deep learning, the position and line-of-sight direction of each person are accurately determined. Combining the existing 3D usage scene model, a complete human portrait positioning set is generated, which includes all the people appearing in the scene and their corresponding spatial positions. Then, based on this positioning set, by establishing a contact distance model in the virtual space, the virtual contact distance between each person and the computer screen is calculated, and a virtual contact distance list of each person is further generated. At this time, if a person is relatively close to the screen, the contact distance is smaller and the line-of-sight coverage range is larger; on the contrary, the contact distance is farther and the line-of-sight coverage range is smaller.

[0089] Step S3: Calculate the virtual contact distance for the 3D usage scene model based on the human portrait positioning set to generate the virtual contact distance of the person; perform contact line-of-sight simulation according to the virtual contact distance of the person to obtain the line-of-sight trajectory of the person.

[0090] In this embodiment, after the virtual contact distance is calculated, a contact line of sight simulation is performed to infer the line of sight trajectory of each character. First, using 3D modeling technology, the line of sight direction of each character is simulated through the connection line between the character positioning point and the screen position. Then, based on the calculation of the line of sight trajectory, the intersection points of the line of sight and the screen and the visible area are generated. These intersection points form the line of sight trajectory of the character, and the screen visual field range of each character can be obtained. Combining the virtual contact distance of each character, the line of sight trajectory is further optimized to make the line of sight simulation result more accurate. Next, based on the simulation result, the interaction area between the line of sight trajectory and the screen is analyzed, and the size of the visible area at different positions on the screen is calculated, so as to provide a basis for screen brightness adjustment.

[0091] Step S4: Analyze the visible screen range of the computer screen according to the character's line of sight trajectory and the virtual contact distance of the character, and generate the visible screen range of the character; dynamically adjust the brightness of the computer screen based on the screen anti-peeping requirement and the visible screen range of the character to perform computer screen anti-peeping control.

[0092] In this embodiment, based on the visible screen range of the character analyzed according to the line of sight trajectory and the virtual contact distance, combined with the anti-peeping requirement of the screen, the screen brightness is dynamically adjusted. First, according to the light intensity and scene privacy degree parameter within the visible screen range of the character, it is judged whether the current screen brightness meets the anti-peeping requirement. If the brightness is too high, the screen content will be visible to others, so the brightness is adjusted to a suitable range. The brightness adjustment process adopts a real-time adjustment method, and based on the current ambient light intensity of the screen and the peeping angle in the surrounding environment, the optimal screen brightness value is calculated to ensure that the screen display content is only visible to the users closer to the computer screen. If the contact distance of this user is far, the system will reduce the brightness to reduce the risk of external peeping. At the same time, the system will update the visible range and anti-peeping requirement in real time according to the dynamic position change of the character, and keep the screen brightness within a reasonable range.

[0093] Preferably, step S1 includes the following steps:

[0094] Step S11: Collect the computer lens image; perform high-precision image segmentation on the computer lens image to obtain a set of scene elements;

[0095] Step S12: Perform scene depth mapping through the set of scene elements to obtain scene depth parameters;

[0096] Step S13: Based on the scene depth parameters, perform spatial point cloud reconstruction on the set of scene elements to generate a three-dimensional spatial point cloud; apply grid processing to the three-dimensional spatial point cloud to obtain a three-dimensional usage scene model;

[0097] Step S14: Conduct a spatial density analysis on the three-dimensional usage scenario model to obtain scene activity hotspots; identify the scene identifiers in the scene activity hotspots;

[0098] Step S15: Based on the scene identifiers, distinguish the scene types of the three-dimensional usage scenario model to obtain computer usage scenario types; conduct a privacy assessment according to the computer usage scenario types to generate scene privacy level parameters.

[0099] In this embodiment, a high-resolution camera is used to collect continuous image data of the area in front of the computer screen at a speed of 30 frames per second. The resolution of the collected images is set to 1920×1080 pixels, and real-time image preprocessing is performed, including denoising and grayscale processing, to make the images clearer and easier for subsequent analysis. Then, a deep learning model is used to perform high-precision image segmentation on the images. The U-Net (a convolutional neural network architecture) is used for semantic segmentation of the images to identify and extract the main scene elements in the images, such as desktops, walls, items, and people, etc., to obtain a set of scene elements. This set contains the bounding boxes and corresponding semantic labels of each scene element. The accuracy of the segmentation result is controlled above 90%. When performing scene depth mapping through the set of scene elements, first, stereo vision technology is used. By collecting images from multiple angles and combining the known camera parameters, structured light technology is used to estimate the depth of the images. A depth estimation algorithm is used to calculate the depth information of each pixel in the image, so as to obtain the depth value of each point in the scene. The depth mapping algorithm uses a multi-view stereo matching method based on tensor decomposition. By combining multiple image views, the depth information is made more accurate, and finally, the depth parameters of the scene are generated. Among them, the depth parameters include the relative distance and spatial position of each scene element, and the depth accuracy is controlled at the centimeter level. When performing spatial point cloud reconstruction based on the scene depth parameters, first, the scene depth parameters are combined with the position information in the set of scene elements, and a point cloud generation algorithm is used to convert the depth information into spatial point cloud data. Each point in the point cloud data represents a specific spatial position in the scene. Then, a point cloud processing method based on the Poisson reconstruction algorithm is used to process the point cloud to generate a continuous three-dimensional point cloud model. The density of the point cloud model is controlled at 5000 points per cubic meter to ensure that the point cloud model is dense enough to reflect the real scene. After that, the three-dimensional space point cloud is meshed. The octree segmentation method is used to divide the point cloud data into small pieces, and a meshing algorithm is applied to connect these small pieces into a three-dimensional mesh model, and finally, a three-dimensional usage scene model is obtained. The meshing accuracy is 1 mm. When performing spatial density analysis on the three-dimensional usage scene model, first, multiple grid regions are defined in the three-dimensional model, and the number and distribution density of objects in each grid region are calculated. A clustering algorithm (such as the DBSCAN density clustering algorithm) is used to analyze these grid regions to identify the regions with frequent activities and relatively dense object distributions, and generate scene activity hotspots. The formation of the hotspots is based on the threshold of spatial density. When the density of a certain region exceeds the set threshold, that region is marked as an activity hotspot. The setting of the threshold depends on the scene type and environmental factors. If the scene is an office environment, the activity hotspots usually appear near the desks and work areas. When identifying the scene identifiers in the scene activity hotspots, first, by identifying the people or items in the hotspots and combining the object recognition algorithm, it is determined whether the region is a specific functional area, such as a desk, a meeting area, a rest area, etc.,Use YOLOv5 (an object detection algorithm) to classify objects in the image, and combine deep learning algorithms to analyze activities within the region. Finally, determine the scene identification of the hot zone. The identifications include functional areas such as desks and conference tables. When differentiating scene types based on the scene identification, utilize the identified scene identification to distinguish the type of the 3D usage scene model through a classification model. Employ the K-means clustering algorithm to divide the scene into different regional categories, such as office areas, rest areas, conference areas, etc. The classification model differentiates scene types according to the activity frequency and functions within each region. The determination of the scene type is based on the activity frequency and regional identification. After the scene type is determined, further conduct a privacy assessment. Based on the scene type, the distribution of activity hot zones, and the visual visibility in the usage scene, use a weighted assessment model to calculate the privacy level of the scene. The privacy assessment model assigns different weights according to the characteristics of different scene types. For example, the privacy weight of the office area is relatively high, while that of the rest area is relatively low. Finally, obtain a quantified scene privacy level parameter.

[0100] Preferably, step S15 includes:

[0101] Step S151: Perform topological deconstruction on the 3D usage scene model based on the scene identification to obtain a scene topological network;

[0102] Step S152: Perform multi-scale convolutional transformation on the scene topological network to obtain scene feature mapping data;

[0103] Step S153: Identify the scene type of the computer usage scene according to the preset application scene atlas for the scene feature mapping data to obtain the computer usage scene type;

[0104] Step S154: Conduct a privacy risk score for the computer usage scene class to obtain a scene risk score;

[0105] Step S155: Conduct a privacy assessment on the scene risk score based on the preset scene risk score threshold to generate a scene privacy level parameter.

[0106] In this embodiment, when topologically deconstructing the three-dimensional usage scenario model based on the scenario identifier, first, using the scenario identifier information, analyze the spatial relationship between scenario elements. Consider all scenario elements and their connection relationships as a graph model, where each scenario element is a node in the graph, and the connection relationship between nodes is an edge in the graph. Use a depth graph convolutional network (DGCN) to model these nodes to obtain a complete scenario topology network. The structure of the topology network takes into account the layout and activity density in the physical space. For example, in an office scenario, the connection relationships between elements such as desks, chairs, and computers will be mapped as edges in the topology network, where the weight of the edge reflects the distance and interaction frequency between these elements. The generation of the scenario topology network does not depend on any preset geometric shape but is automatically constructed based on the actual scenario identifier and the relationship between elements. When performing multi-scale convolutional transformation on the scenario topology network, use a multi-scale convolutional neural network (MSCNN) to perform layer-by-layer convolutional processing on the scenario topology network. By setting different convolutional kernel sizes, extract the spatial features of the scenario at multiple scales. First, process the low-level features of the scenario topology network to capture relatively fine-grained spatial layout features. For example, identify the close relationship between the office desk and the chair. Then, perform convolutional processing at the medium scale to further extract regional spatial patterns, such as the interaction relationship between the office area and the meeting area. Finally, through high-scale convolutional processing, obtain the global scenario features and identify the overall structure of the scenario. For example, the layout pattern of the entire office. Through this multi-scale convolutional transformation, the obtained scenario feature mapping data can comprehensively reflect the different levels of spatial structure and functional partitions in the scenario. The convolutional kernel sizes of the convolutional network are 3×3, 5×5, and 7×7 respectively. When performing scenario type recognition on the scenario feature mapping data according to the preset application scenario atlas, first, through the trained classification model, combined with the application scenario atlas, which contains the markings and feature information of different types of scenarios, perform classification recognition on the scenario feature mapping data. Use a convolutional neural network (CNN) for feature extraction and map the features to the preset scenario type space through a fully connected layer. The scenario type space includes multiple categories such as the office area, meeting area, and rest area. The goal of scenario type recognition is to determine which category the current computer usage scenario belongs to based on the feature data extracted from the scenario topology network. The classification result is output through Softmax (normalized exponential function), and a probability value is assigned to each scenario type. Finally, select the scenario type with the largest probability value as the type of the computer usage scenario. During the training process, use 10,000 labeled images are used to achieve a training accuracy of over 95%. When performing privacy risk scoring for computer usage scenario classes, first, using the identified scenario types, based on the characteristics of different scenario types and in combination with privacy risk assessment rules, different privacy risk weights are set for each scenario type. For example, the privacy risk score for the office area is relatively high because there is a large amount of confidential information and personnel operations in this area, while the privacy risk score for the rest area is relatively low because the use of this area usually does not involve sensitive data. Through these weight values, weighted scoring is performed on the scenario types. The weighted average method is used in the scoring process, combining the risk weights of each scenario type with their occurrence frequencies to obtain the scenario risk score. The scoring range is set from 0 to 100, where 0 indicates no privacy risk and 100 indicates extremely high privacy risk. When performing privacy assessment on the scenario risk score based on a preset scenario risk score threshold, first, a predefined risk score threshold is set, and this threshold is adjusted according to different usage scenarios. The threshold for the office area is relatively low, while the thresholds for the meeting area and the sensitive data operation area are relatively high. If the scenario risk score exceeds this threshold, then this scenario is considered a high-risk scenario and further privacy protection measures need to be taken, such as adjusting the screen brightness or activating the anti-peeping mode. If the risk score is lower than the threshold, no adjustment is required. Finally, according to the privacy assessment results, a scenario privacy degree parameter is generated, and this parameter ranges from 0 to 1, where 0 indicates no privacy risk and 1 indicates the highest privacy risk.,

[0107] Preferably, step S2 includes the following steps:

[0108] Step S21: Perform threshold comparison on the scenario privacy degree parameter based on a preset privacy degree threshold to obtain a privacy determination result;

[0109] Step S22: Map the anti-peeping requirements to the three-dimensional usage scenario model according to the privacy determination result to generate screen anti-peeping requirements;

[0110] Step S23: Perform human face detection on the computer camera image based on the screen anti-peeping requirements, correct the lighting factors of the computer camera image to obtain a balanced lighting image; detect the human face area of the balanced lighting image; perform key point localization on the human face area to generate human face key point data;

[0111] Step S24: Estimate the initial human face localization based on the human face pose data to obtain the initial human face localization; perform localization integration on the initial human face localization to generate a human face localization set.

[0112] In this embodiment, when comparing the scene privacy level parameter with a preset privacy level threshold, first, a privacy level threshold range is set, which is adjusted according to the privacy requirements of different usage scenarios. The privacy level threshold for the office scenario is set to 0.7, while the privacy level thresholds for the conference room and sensitive data operation scenarios are set to 0.9. By comparing the scene privacy level parameter with this threshold, privacy determination is performed. If the scene privacy level parameter is greater than or equal to the preset threshold, it is determined as a high-privacy scene; if it is less than the threshold, it is determined as a low-privacy scene. The setting of the privacy evaluation threshold depends on multiple factors, including the scene type and usage environment. For example, in an open office, the privacy threshold is relatively low, while in a conference room with a higher security level, the threshold is higher. The privacy level parameter is obtained through multi-level scene analysis. When mapping the anti-peeping requirements to the 3D usage scene model according to the privacy determination result, first, for the part with a high-privacy scene as the privacy determination result, an anti-peeping requirement mapping diagram is generated. This diagram marks different areas as "requiring anti-peeping" or "not requiring anti-peeping" according to the privacy requirements in the scene. For example, in an open office scenario, the work positions facing the screen will be marked as "requiring anti-peeping", while the areas facing the wall or not displaying the screen content to the outside will be marked as "not requiring anti-peeping". During the generation of the mapping, the privacy requirements of each area in the scene are comprehensively analyzed by combining the scene privacy level parameter and the spatial structure of the usage scene. The anti-peeping requirement mapping result is dynamically adjusted according to different scenes to ensure that the privacy of each area is fully protected. The generated screen anti-peeping requirement diagram includes multiple marked areas, accurately indicating which areas require anti-peeping protection. When the anti-peeping requirement for the screen is not to require anti-peeping, no processing is performed; when the anti-peeping requirement for the screen is to require anti-peeping, the computer camera is called according to the screen anti-peeping requirement to obtain the image data within the current field of view, and the timestamp is extracted to mark the current detection cycle. The dynamic parameter threshold of the human face detection is adjusted according to the environmental parameters such as the light intensity and the device placement angle at the time of detection trigger to adapt to different usage scenarios. The hardware acceleration module on the device side is used to quickly process the image data, and the human face recognition algorithm is called to match and identify the human target in the current camera frame, generating a list of human identification information. The identified human identification information for each person is associated with its spatial coordinates in the frame, and the usage status parameters of the device such as the screen brightness and the mouse movement trajectory are combined to determine whether the current person is the user of the device or a potential peeper. All the detected personnel information and their spatial relationship relative to the screen are recorded as the human face detection result. Based on the human face detection result, the lighting factor of the computer camera image is corrected. The histogram equalization algorithm is used to adjust the exposure and brightness of the image so that all areas in the image have a consistent brightness distribution, avoiding overly dark or overly bright image areas. The correction process is based on the Adaptive Histogram Equalization method in computer vision. First,Calculate the histogram of the input image, divide the image into multiple local regions, and perform equalization processing on each region separately to avoid over-adjustment of large areas. The image obtained through illumination equalization can provide clear and uniform image quality under different illumination conditions. After obtaining the evenly illuminated image, use a face detection algorithm to identify the face region in the image. When detecting the face region of the evenly illuminated image, apply existing face detection algorithms and adopt a face detection model based on a convolutional neural network (CNN). For example, use a Haar feature cascade classifier or a deep convolutional neural network to perform face recognition on the image. During the recognition process, first scan different regions in the image through the sliding window technique to identify the face region, and then further accurately locate the position of the face through the neural network model. Finally, output a rectangular box containing the position of the face region. After this process, the face region in the image is successfully detected. When performing key point localization on the face region, use a facial key point detection algorithm, a face landmark localization model based on deep learning, and train it through a calibrated landmark dataset. The model can identify the positions of key points on the face, including the positions of parts such as eyes, nose, and mouth. Through these key points, further construct the pose information of the face. During the key point localization process, first perform image preprocessing on the face region, including grayscale conversion and normalization processing, and then extract features through a convolutional neural network to locate the coordinate values of each facial key point and generate face key point data. The data includes the two-dimensional coordinate information of each key point. When estimating the initial positioning of the human portrait based on the face pose data, combine the face key point data and estimate the spatial position and orientation of the face through a pose estimation algorithm (such as the PnP algorithm). Through the three-dimensional reconstruction of the face key points, obtain the initial positioning of the human portrait. This positioning value is the position in the three-dimensional coordinate system and determine the orientation of the face, and then calculate the position and angle of the human portrait. Further evaluate the relative position with the screen to obtain the result of the initial human portrait positioning. When performing positioning integration on the initial human portrait positioning, adopt a weighted fusion strategy for multiple detection results, and merge the human portrait positioning information obtained at different time points or different angles to reduce the positioning error and ensure the accuracy of the positioning result. During integration, first perform standardization processing on the positioning data from different sources, and then fuse each positioning data according to the weighted average method. The weight distribution is based on the reliability of the positioning data. Finally, generate a human portrait positioning collection, which contains multiple processed and optimized human portrait positioning data.,

[0113] Preferably, step S23 includes the following steps:

[0114] Step S231: Perform human portrait detection on the computer lens image based on the screen anti-peeping requirement, and perform brightness histogram equalization on the computer lens image to obtain a brightness equalized image;

[0115] Step S232: Enhance the contrast of the luminance-uniform image to generate an enhanced lens image; perform illumination compensation on the enhanced lens image to obtain an evenly illuminated image;

[0116] Step S233: Perform multi-scale face detection on the evenly illuminated image to generate face candidate regions; perform boundary redrawing on the face candidate regions to obtain face regions;

[0117] Step S234: Extract feature points from the face regions to obtain face feature point data; perform grid deformation mapping on the face feature point data to obtain grid deformation data;

[0118] Step S235: Perform key point regression on the grid deformation data to obtain key point localization data; perform coordinate system conversion on the key point localization data to obtain face key point data.

[0119] In this embodiment, according to the screen anti-peeping requirement, the computer lens is called to obtain the image data within the current field of view, and the timestamp is extracted to mark the current detection cycle. The dynamic parameter threshold of human face detection is adjusted according to the environmental parameters such as light intensity and device placement angle at the time of detection trigger to adapt to different usage scenarios. The hardware acceleration module on the device is used to quickly process the image data, and the face recognition algorithm is called to match and recognize the human target in the current lens picture, generating a list of human identification information. The identified human identification information of each person is associated with its spatial coordinates in the picture, and combined with the usage status parameters of the device such as screen brightness and mouse movement trajectory, it is judged whether the current person belongs to the user or potential peeper of the device. All detected personnel information and their spatial relationship relative to the screen are recorded as the face detection result. Based on the face detection result, histogram equalization is performed on the computer lens image. The Adaptive Histogram Equalization algorithm is used. First, the image is divided into several small regions, and then the brightness histogram of each region is independently processed. By adjusting the brightness and contrast of each region, the global brightness distribution of the image becomes more uniform, avoiding over-bright or over-dark regions. The brightness equalized image obtained by this method has a uniform brightness distribution, enabling images taken in different environments to maintain visual consistency under large lighting differences. The threshold and region size in image processing can be determined through experiments. Generally, the size of the region is 20x20 pixels, and the contrast gain of the histogram is set to 1.5. Contrast enhancement is performed on the brightness uniform image. The local contrast enhancement algorithm is used. First, edge detection is performed on the brightness equalized image to identify the key edge regions in the image, and then the contrast of these regions is enhanced to highlight the details. During the enhancement process, the overall brightness of the image remains unchanged, while the gray range of the low-contrast regions is stretched to improve the local contrast of the image. The edges of the enhanced image are clearer, the details are more prominent, and the high-contrast regions in the image will be more obvious. The parameter used during the enhancement process is a contrast enhancement factor of 1.2. Under different environmental lighting conditions, light compensation is performed on the enhanced lens image. The light compensation algorithm is used to equalize the image. The light compensation method based on Gaussian Filter is mainly used. First, the enhanced image is Gaussian blurred to obtain a blurred light image, and then the difference between the original image and the blurred image is calculated to obtain the light compensation map. Finally, the compensation map is combined with the enhanced image to adjust the light brightness of the image, making the brightness of the entire image more uniform and avoiding local over-bright or over-dark regions. The size of the Gaussian filter kernel used during the compensation process is 15x15, and the standard deviation is set to 2.5. Perform multi-scale face detection on the evenly illuminated image. First, use a Convolutional Neural Network (CNN) model to perform multi-scale analysis of the image using multi-layer feature extraction technology, scan different-sized regions in the image, and detect faces at different distances and angles. The input image size used by the network is 224x224 pixels. Perform face detection through a pre-trained model, divide the image into multiple sub-regions of different sizes, and detect each region one by one through the sliding window technology to output face candidate regions. These candidate regions are the positions where faces exist in the image. During detection, compare the feature points of each region with the trained template to determine whether they conform to face features. The learning rate used by the model is 0.001. The training cycle is 50 epochs, and the detection accuracy reaches over 95%. When performing boundary redrawing on the face candidate region, first, the position of the face is framed by the detected candidate region. Then, morphological operations are used to refine the boundary. Specifically, the boundary lines are optimized through dilation and erosion operations to ensure that the boundary more accurately outlines the true shape of the face region. The dilation operation expands the boundary line, and the erosion operation shrinks the boundary, finally obtaining a more accurate face boundary. The accuracy of the boundary is affected by the parameters of the dilation and erosion operations. The number of dilation and erosion operations is set to 3 times. When extracting feature points from the face region, a classic facial feature point detection algorithm is used, based on the facial landmark detection model in the Dlib (a toolkit for machine learning) library. First, a grayscale image of the face region is obtained, and then through Dlib's 68-point facial feature point model, the key landmark points of the face are extracted. These landmark points include the key coordinates of regions such as eyes, eyebrows, nose, and mouth. During the extraction process, the image is first preprocessed, grayscaled, and noise is removed. Then, the landmark point coordinates of the face are calculated through the feature descriptors of local regions. The extracted feature point data includes the two-dimensional coordinate values of each landmark point. When performing grid deformation mapping on the face feature point data, first, a grid deformation model is constructed to map the facial feature points to a standard grid. The standard grid consists of a set of predetermined fixed coordinates. Through the deformation algorithm, the feature points in the actual image are adjusted to the standard grid. During the process, affine transformation is used to correct the feature points to ensure that facial features can be accurately matched to the standard grid at different scales. During the grid deformation process, the interpolation method used is bilinear interpolation. The grid size is set to a 20x20 grid cell, and each cell includes four control points. When performing key point regression on the grid deformation data, a regression model (e.g., based on support vector machine (SVM) regression) is used to perform regression processing on the deformed grid data. During the regression process, first, the grid deformation data is labeled through the training set to establish the mapping relationship between the input and output. The model learns through regression to convert the grid deformation data into accurate face key point data. During the regression process, the loss function used is mean squared error (MSE). This regression process ensures that the finally obtained key point data has high accuracy. During the regression process, the learning rate of the model is set to 0.0005. The training cycle is 100 times. When performing coordinate system conversion on the key point positioning data, by defining a three-dimensional coordinate system, the two-dimensional human face key point data is mapped into the three-dimensional space. Through affine transformation and perspective transformation, the two-dimensional key point data in the image is mapped into the actual three-dimensional space. During the conversion process, first, the image is geometrically corrected according to the known camera calibration parameters to obtain the coordinate positions of each key point in the three-dimensional space, and finally, the three-dimensional coordinate data of each facial key point is obtained.

[0120] Preferably, the virtual contact distance calculation for the three-dimensional usage scenario model based on the portrait positioning collection in step S3 includes:

[0121] Converting the three-dimensional usage scenario model into a positioning coordinate system;

[0122] Projecting the portrait positioning collection into the positioning coordinate system to obtain user space coordinates;

[0123] Based on the user space coordinates, estimating the nearest neighbor volume to generate an interaction space volume;

[0124] Performing gradient projection conversion on the interaction space volume to generate multi-level contact probability parameters;

[0125] Performing boundary constraint filtering on the multi-level contact probability parameters to obtain contact boundary data;

[0126] Calculating the contact distance for the contact boundary data to generate the virtual contact distance of the person.

[0127] In this embodiment, a standard coordinate system is selected as a reference. This coordinate system is generally set as a Cartesian coordinate system in a three-dimensional space, where the origin is usually set as the center point of the scene, and the X, Y, and Z axes represent the horizontal, vertical, and depth directions in the scene respectively. Then, the coordinates of each key point of the three-dimensional scene model are transformed. The rotation matrix is used to map all points in the scene model from the scene coordinate system to a new positioning coordinate system. During the transformation process, the rotation angle and scale factor are adjusted according to the specific size of the scene and the user's requirements. The transformed coordinate system is used for subsequent precise estimation and analysis of the user's spatial position. The setting of the transformation parameters is debugged according to the specific experimental scene. Obtain the coordinate data of each person in the three-dimensional space obtained from the previous steps, map these data to the preset positioning coordinate system according to their spatial positions, and use the projection transformation technology to transform the three-dimensional space coordinates of each person into two-dimensional screen coordinates. During the process, the depth information (Z-axis) of the person is ignored, and the horizontal and vertical coordinates (X, Y) are retained, so as to obtain the projected user space coordinates. The projection operation uses the perspective projection algorithm, and the projection matrix during the transformation process is set according to the resolution of the screen and the physical size of the display device to ensure that the projected coordinates are accurately mapped onto the screen. Each projected coordinate corresponds to the spatial position of the user in front of the screen, and usually, the screen projection coordinate accuracy of each user is set to reach 0.5 millimeters. Using the method based on Voronoi Diagram, the user space coordinates are estimated for nearest neighbors. By calculating the neighboring areas of each person's spatial position, the interaction space is determined. First, with the spatial coordinates of each person as the center, a neighboring volume area is divided, and this area is defined as the influence range of that person. By calculating the interaction ranges of all persons, an interaction space volume containing all users is obtained. The estimation process uses spatial segmentation technology to divide each user's space into non-overlapping neighboring volume areas. The boundaries of these areas are calculated by the Voronoi algorithm. The volume calculation is based on the spatial distances between each person and other persons. The estimation accuracy of the volume is affected by the scene resolution and calculation accuracy. The neighboring volume range for each user is set as a 5-meter cube. The gradient calculation method is used to process the volume of the interaction space. The boundary of the interaction volume is mapped into a gradient map. By performing differential calculations on the volume data, the gradient value of each point is obtained. Then, using the Gradient Descent algorithm, the boundary of the interaction volume is optimized to generate a more accurate contact area. During the gradient projection conversion, first, the interaction volume data is mapped onto a two-dimensional plane, and the gradient change of each volume point is calculated. Then, through the optimization algorithm, the contact point data is adjusted to a shape and position more suitable for user interaction. The finally obtained gradient map will reflect the contact probabilities of different interaction areas. The gradient value is adjusted through experiments and set as a step size of 0.1. Gaussian Filtering is used to smooth the contact area after gradient projection conversion. By blurring the boundary, the influence of noise is reduced, and the boundary of the contact area is ensured to be smoother. In the filtering operation, a Gaussian filter kernel with a standard deviation of 1.0 is used for convolution calculation. The parameter adjustment during the filtering process is based on the size of the user space. The contact probability data obtained after filtering will reflect the contact probabilities of different areas. Then, by setting boundary constraint conditions (Boundary Constraints), the effective contact areas are further screened out. The boundary constraint conditions are set according to the actual distances of the users in the scene. The contact area within 1 meter of the boundary is considered valid. Using the obtained contact boundary data, the distance between each user's space coordinate and the contact boundary is calculated. The Euclidean Distance formula is used for the calculation of the contact distance. By calculating the distances between each user's space coordinate and every point on the boundary, the virtual contact distance of each contact point is obtained. When calculating the contact distance, the spatial coordinates of each user and the interaction volume boundary are calculated multiple times to ensure the accuracy of the calculation results. The finally generated virtual contact distance data will be used to determine the anti-peeping and interaction optimization settings of the screen. The distance calculation method used during the calculation is three-dimensional Euclidean distance calculation. The set distance accuracy is 0.1 millimeter, and each calculation is adjusted based on the real-time data of the interaction space.

[0128] Preferably, the contact line-of-sight simulation according to the virtual contact distance of the person in step S3 includes:

[0129] Deriving the actual interval of the person based on the virtual contact distance of the person to obtain the actual contact distance of the person;

[0130] Updating the three-dimensional distance of the portrait positioning set based on the actual contact distance of the person to generate the three-dimensional contact distance of the person;

[0131] Performing line-of-sight projection simulation according to the three-dimensional contact distance of the person. During the line-of-sight projection process, setting the projection angle to 25 - 35 degrees and the maximum projection distance to 2.5 - 3.5 meters to obtain the initial line-of-sight vector;

[0132] Performing difference smoothing processing on the initial line-of-sight vector to generate a continuous line-of-sight curve;

[0133] Performing three-dimensional trajectory reconstruction on the continuous line-of-sight curve to obtain the line-of-sight trajectory of the person.

[0134] In this embodiment, when deriving the actual interval between people based on the virtual contact distance of people, first, the virtual contact distance data of people is obtained. This data is the distance value obtained by calculating the contact area between the user and the scene in the previous step. These virtual contact distances reflect the interaction range between the user and surrounding objects. Then, a geometric model of three-dimensional space is used to map the virtual contact distance to the actual spatial interval, and the linear interpolation method (Linear Interpolation) is used to refine the contact distance to ensure that the virtual contact distance is accurate to the actual contact space of each user. After the interpolation process, the obtained actual contact distance will be adjusted according to the relationship between the user's position and the physical size of the virtual scene. When updating the stereo distance of the portrait positioning collection based on the actual contact distance of people, first, the actual contact distance data of each person needs to be combined with the previously obtained portrait positioning collection. Through coordinate transformation, the two-dimensional position coordinates of the people are mapped into three-dimensional space. Then, combined with the actual contact distance of each person, the original portrait position is adjusted, and the spatial position of each person is updated to a new stereo coordinate system according to the contact distance. The updated stereo coordinates reflect the actual distance information of the user. This process is achieved through projection transformation (Projection Transformation), which converts the two-dimensional projection image into stereo space coordinates. The updated stereo contact distance will directly affect the accuracy of the line-of-sight projection simulation. During the update process, a three-dimensional geometric space model is used for coordinate operations to ensure that the accuracy of the stereo contact distance reaches 0.1 mm. By calculating the stereo contact distance of each person, the actual spatial relationship between the person and the screen is used as the input, and the ray casting algorithm (Ray Casting Algorithm) is used to project the line of sight from the viewpoint of each person according to the stereo contact distance to generate an initial line-of-sight vector. The line-of-sight vector extends forward from the position where the person is located and finally reaches a certain target point in the interaction scene after a series of spatial transformations. Each line-of-sight vector in the projection simulation process needs to be accurately calculated based on the spatial coordinates and contact distance of the person. The accuracy and stability of the line-of-sight vector directly affect the subsequent generation of the line-of-sight trajectory. During the line-of-sight projection process, the projection angle is set to 25 - 35 degrees, and the maximum projection distance is 2.5 - 3.5 meters. For each generated initial line-of-sight vector, the Cubic Spline Interpolation method is used to smooth the path of the line of sight, reducing the sharp turns generated during the straight-line projection process. Using this method, through interpolation calculation, the transition curve between multiple line-of-sight points is optimized to generate a continuous smooth curve. The process of interpolation smoothing will calculate the interpolation coefficients of the smooth curve based on the spatial position data and projection angle of each line-of-sight point. The finally obtained line-of-sight curve will be very smooth, avoiding the discontinuity and jitter existing in the original line-of-sight vector. After smoothing, the continuous line-of-sight curve will be used for subsequent three-dimensional trajectory reconstruction. The parameters of the smoothing process are generally set such that the distance between every two points does not exceed 1 centimeter, and the curve smoothness reaches 0.01 degrees. Map each smoothed line-of-sight vector into a three-dimensional coordinate system, and use the 3D Interpolation method to reconstruct each line-of-sight point according to the time series to ensure the smooth transition of the line-of-sight trajectory in space. Convert the line-of-sight curve into actual trajectory data through a reconstruction algorithm. These data will reflect the change trajectory of the person's line of sight. During the reconstruction process, the Kalman Filtering technology is used to process the noise data to further optimize the accuracy of the trajectory. The trajectory data will be dynamically updated according to the user's real-time spatial changes. The reconstructed three-dimensional trajectory map shows the person's line-of-sight trajectory, and this trajectory will be mapped to the display content on the screen.

[0135] Preferably, the analysis of the visible range of the computer screen according to the person's line-of-sight trajectory and the person's virtual contact distance in step S4 includes:

[0136] Perform a spatial projection conversion on the person's line-of-sight trajectory to obtain a screen plane projection, where the width of the screen is set to 1900 - 1940 pixels and the height is 1060 - 1100 pixels. The parameter settings during the projection process are a focal length of 45mm - 55mm and a projection angle of 55 - 65 degrees;

[0137] Perform a cross-correlation on the screen plane projection and the person's virtual contact distance to generate a viewing distance combination index;

[0138] Construct a frustum model based on the viewing distance combination index; perform a field-of-view generalization process on the frustum model to obtain a three-dimensional field-of-view range;

[0139] Perform a plane truncation mapping on the three-dimensional field-of-view range to generate a screen field-of-view coverage matrix;

[0140] Perform a reverse inference of the person's viewing angle based on the screen field-of-view coverage matrix to generate the visible screen range of the person.

[0141] In this embodiment, when performing spatial projection conversion on the gaze trajectory of a person, first, gaze trajectory data of the person is obtained, and the trajectory points are projected and converted according to the coordinate system of the three-dimensional space. The perspective projection method is used to map the three-dimensional trajectory points onto the two-dimensional plane of the screen. The camera matrix is used to convert the gaze trajectory of the person from the three-dimensional space to the coordinate system of the screen plane. During this process, the relationship between the projection ratio of the trajectory and the screen size is ensured to be accurate. Usually, the width of the screen is set to 1900 - 1940 pixels, and the height is set to 1060 - 1100 pixels. The spatial conversion is performed using the projection matrix, and the obtained planar projection data will accurately reflect the position of the person's gaze on the screen. The parameters during the projection process are set as the focal length being 45mm - 55mm and the projection angle being 55 - 65 degrees. When performing cross-correlation between the screen planar projection and the virtual contact distance of the person, first, based on the virtual contact distance data of the person, the contact range is defined and matched with the screen planar projection data. Combining the virtual contact distance of the person and the gaze trajectory on the projection plane, the cross-correlation analysis method is used for data matching to calculate the actual contact range corresponding to each gaze point. During this process, according to the spatial relationship between the screen resolution and the virtual contact distance of the person, the combined information of each projection point and the contact range is obtained. The generated visual distance combination index will reflect the visible area of each gaze point, and according to the different distances of the person from the screen, the visual distance value of each gaze point is adjusted. Based on the generated visual distance combination index, a frustum model of the person is constructed. The frustum geometry model is used, and this model consists of three parts: the viewpoint of the person, the projection plane, and the visible range. Combining the contact distance information of each projection point, the visible angle of the screen is deduced. The frustum is modeled using the three-dimensional space coordinate system, and the visual distance combination index is converted into the boundary parameters of the frustum. The near clipping plane and the far clipping plane of the frustum are defined. In actual calculations, the near clipping plane is usually set as the minimum distance between the person and the screen, and the far clipping plane is the maximum visible distance between the person and the screen. The frustum angle is set to 45 degrees to 60 degrees. The constructed frustum model is simplified, and the clustering algorithm is used to generalize the visible area within the frustum. During the clustering process, the space of the frustum is subdivided according to the visual distance combination index to determine the field of view range of each area. After processing, multiple sub-fields of view are obtained, and the size of each sub-field of view is related to the actual contact distance of the person. The total coverage range of the field of view is the total volume of the frustum in the three-dimensional space.The maximum likelihood estimation is adopted to optimize the calculation of the field of view range, ensuring that in practical applications, the distribution of the field of view range can more accurately reflect the actual changes in the visible area on the screen. The obtained three-dimensional field of view range data after processing will contain the complete information of the screen area that the person can see. The field of view range is controlled within 80% of the screen. According to the obtained three-dimensional field of view range, the plane clipping technology is used to perform plane mapping on the field of view range. Based on the two-dimensional size of the screen, the three-dimensional field of view range is mapped onto the two-dimensional plane of the screen. The mapped data after clipping represents the screen area that can be seen from the perspective of the person. The projection matrix is used to correspond the field of view range in the three-dimensional space with the plane coordinates of the screen, ensuring that the field of view range can be completely mapped to the display area of the screen. After the field of view coverage matrix of the screen is generated, the visible range of each area will be displayed. Each value in the matrix indicates whether each pixel on the screen is covered by the person's line of sight. The size of the generated matrix is the same as the screen resolution. Each value in the matrix is 0 indicating an invisible area and 1 indicating a visible area. Using the generated screen field of view coverage matrix as the input, the inverse reasoning algorithm is adopted to reverse infer the person's perspective. Based on each value in the field of view coverage matrix, the actual perspective and position of the person in space are deduced, and the reverse perspective related to the screen coverage area is calculated. The three-dimensional space coordinate system is used for reverse inference, and the field of view matrix is mapped back to the three-dimensional space where the person is located to obtain the actual position and perspective data of the person in space. Based on these data, the visible screen range of the person is further generated. The recursive algorithm is used to optimize the accuracy of the deduction during the reverse inference process, ensuring that the reverse inference result of the person's perspective can be consistent with the visible area on the screen. The perspective range during the reverse inference process is usually controlled within the person's field of view, and the maximum reverse inference distance is set to 2 meters.

[0142] Preferably, the dynamically adjusting the brightness of the computer screen based on the screen anti-peeping requirement and the visible screen range of the person in step S4 includes:

[0143] Perform maximum range screening on the visible screen range of the person to obtain the main person's perspective, where the main person's perspective is a perspective range of 90 - 110°, and the maximum viewing distance is 1.8 - 2.2 meters; based on the main person's perspective, extract the visible screen ranges of other people from the visible screen range of the person, where the visible screen ranges of other people are 70 - 90°, and the maximum viewing distance is 1.6 - 2.0 meters;

[0144] Perform exposed screen cutting on the visible screen ranges of other people based on the screen anti-peeping requirement, with a cutting accuracy of 5 - 10 pixels, to generate exposed screen units;

[0145] Perform brightness gradient design on the exposed screen units to obtain the brightness adjustment gradient;

[0146] Based on the brightness adjustment gradient, perform brightness adjustment simulation on the computer screen to generate computer screen brightness levels at all levels;

[0147] Perform gradient brightness control on the exposed screen units through the computer screen brightness levels at all levels. Set the maximum amplitude of the brightness change to be between 0 and 255. Set the brightness of the units near the screen edge to 80 - 120 cd, and set the brightness of the units in the central area to 180 - 220 cd to perform anti-peeping control on the computer screen.

[0148] In this embodiment, when performing maximum range screening on the visible screen range of a person, first, based on the determined screen size and the position of the person, the line-of-sight direction and distance information of the person are mapped onto the two-dimensional plane of the computer screen using Spatial Projection. The Convolutional Neural Network (CNN) in the deep learning algorithm is used to precisely calculate the visible area on the screen, identify the intersection points of the person's line of sight and the screen, calculate the maximum visible range according to the actual distance, eye height, line-of-sight angle of the person and the position of the screen, generate the regional data of the main person's perspective. The boundary of the main perspective is set as the visible points at the maximum viewing distance, and the width and height of the perspective are dynamically adjusted according to the relative position of the person and the screen. Each pixel point on the screen will be judged whether it is within the main perspective range. The main perspective range is usually set as a perspective range of 90 - 110°, and the maximum viewing distance is 1.8 - 2.2 meters. When extracting the visible screen range of other persons based on the main person's perspective, first, according to the boundary information of the main perspective, combined with the distance of the person and the angle of the screen, the Ray Tracing Algorithm is used to determine the line-of-sight area of other persons, project the line of sight of other persons onto the screen, obtain their visible area on the screen, calculate the intersection points of each line of sight and their relative positions with the screen, further screen out the specific positions of each intersection point on the screen, define the line-of-sight range of other persons according to the spatial relationship and relative angle between the person and the screen. The distance and angle of the intersection points will be used as constraint conditions to extract the screen area seen by other persons. The line-of-sight angle range of the person is usually set as a perspective range of 70 - 90°, and the maximum viewing distance is 1.6 - 2.0 meters. When performing exposed screen cutting on the visible screen range of other persons based on the screen anti-peeping requirement, first, the visible area of the screen is cut using the Occlusion Detection Algorithm, and the cutting accuracy is 5 - 10 pixels. According to the perspective of other persons, determine which parts of the screen will be exposed. The exposed area is obtained by screening the intersection part in the visible area. The selected exposed screen part is an independent screen unit. The calculation of the exposed area combines the size of the screen and the relative position of each exposed area. According to the cutting algorithm, the exposed part of the screen is divided into multiple small areas, and the size of each area is determined by the angle and distance of the person's line of sight. The exposed screen unit is divided according to the screen resolution, and the size of each unit area is 50x50 pixels. When designing the brightness gradient for the exposed screen unit, first, according to the position of the exposed screen unit and the visible range of the screen, design the brightness gradient model, combine the LightIntensity Decay Formula and the visible range information of the person to determine the law of brightness change.The brightness of each exposed unit is gradually changed using the Weighted Average Method. The brightness of the exposed units closer to the center of the screen is higher, and the brightness of the units farther from the center is lower. When designing, the Color Model is used to adjust the brightness value of each unit. Usually, the maximum amplitude of the brightness change is set between 0 and 255. The brightness gradient is set according to the actual anti-peeping requirements of the screen. The brightness of the units near the edge of the screen is set to 80 - 120 cd, and the brightness of the units in the central area is set to 180 - 220 cd. When simulating the brightness adjustment of the computer screen based on the brightness adjustment gradient, first, according to the calculated brightness gradient model, the brightness adjustment value of each exposed unit is integrated with the overall brightness of the screen. Using the Per-Pixel Adjustment Method, the brightness gradient is applied to each screen pixel point to simulate the brightness change of the screen, ensuring that within the exposed area of the screen, the brightness decreases smoothly from the center to the edge. During the adjustment process, the brightness is controlled by adjusting the RGB (red, green, blue) values of each pixel. The Interpolation Algorithm is used during the adjustment process to generate a smooth brightness transition, ensuring that the brightness change of each screen unit conforms to the predetermined gradient rule. During the simulation process, the brightness is adjusted in units of 100x100 pixels to obtain an accurate control effect. The adjustment result generates multiple brightness levels. Combining the generated brightness levels, according to the brightness gradient data of each screen unit, the brightness of each exposed screen unit is gradually controlled. Using the Layer-by-Layer Processing method, different brightness control values are set for units with different brightness levels. The brightness of the exposed units closer to the center is 180 - 220 cd, and the brightness of the edge units is 80 - 120 cd. The Gradient Algorithm is used to smoothly transition these brightness values. The brightness of the screen is adjusted under the preset anti-peeping requirements to ensure that the brightness change in the visible area of the screen meets the anti-peeping requirements. By simulating the brightness change, the exposed parts within the visible range of the screen will present different brightness levels in different areas of the screen. After the gradient brightness control is completed, a dynamic brightness adjustment scheme suitable for anti-peeping is generated.

[0149] The present invention also provides a computer screen control system for executing the computer screen control method as described above. The computer screen control system includes:

[0150] An image acquisition module, used for acquiring computer lens images; performing scene modeling on the computer lens images to obtain a three-dimensional usage scene model; performing scene privacy evaluation on the three-dimensional usage scene model to generate scene privacy degree parameters;

[0151] An anti-peeping analysis module, which is used to perform anti-peeping requirement analysis on the three-dimensional usage scenario model according to the scene privacy degree parameter to generate a screen anti-peeping requirement; perform human portrait detection on the computer lens image based on the screen anti-peeping requirement, extract the human spatial position and line of sight direction, and generate a human portrait positioning set;

[0152] A contact calculation module, which is used to perform virtual contact distance calculation on the three-dimensional usage scenario model based on the human portrait positioning set to generate a human virtual contact distance; perform contact line-of-sight simulation according to the human virtual contact distance to obtain the human line-of-sight trajectory;

[0153] A brightness control module, which is used to perform screen visible range analysis on the computer screen according to the human line-of-sight trajectory and the human virtual contact distance to generate a human visible screen range; dynamically adjust the brightness of the computer screen based on the screen anti-peeping requirement and the human visible screen range to perform computer screen anti-peeping control.

[0154] Through the real-time monitoring of the image acquisition module, the system can timely obtain and analyze the surrounding environment. The created three-dimensional usage scenario model provides a comprehensive perspective for understanding the spatial layout. The privacy degree parameter generated by the scene privacy evaluation provides a reliable basis for the subsequent anti-peeping requirement analysis. The anti-peeping analysis module can accurately identify when anti-peeping is needed. The function of generating the human portrait positioning set enables the system to effectively track the user's position. The virtual contact distance calculation of the contact calculation module optimizes the interaction between the user and the screen. The contact line-of-sight simulation provides a clear description of the user's line-of-sight trajectory. The brightness control module combines the line-of-sight trajectory and the virtual contact distance to accurately analyze the visible range of the screen. The ability to dynamically adjust the screen brightness enables the system to automatically adapt in different environments, reducing the risk of being peeked by others. The overall system improves the intelligent level of computer use, ensures that users enjoy higher privacy protection and information security in a changing environment, promotes the wide application of intelligent screen control technology, meets the growing demand for information security in modern society, and enhances the sense of security and trust of users during computer use.

[0155] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the application document within the present invention.

[0156] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A computer screen control method, characterized in that: The following steps are involved: Step S1: collecting computer lens images; performing scene modeling on the computer lens images to obtain a three-dimensional usage scene model; performing scene privacy evaluation on the three-dimensional usage scene model to generate scene privacy degree parameters; Step S2: performing anti-peeping demand analysis on the three-dimensional usage scene model according to the scene privacy parameter, and generating screen anti-peeping demand; performing portrait detection on the computer lens image based on the screen anti-peeping demand, extracting the spatial position and sight direction of the person, and generating a portrait positioning collection; Step S3: Calculate the virtual contact distance of the three-dimensional usage scene model based on the portrait positioning collection to generate the virtual contact distance of the character; perform contact sight line simulation according to the virtual contact distance of the character to obtain the sight line trajectory of the character; Step S4: analyzing the screen visible range of the computer screen according to the character's sight line trajectory and the character's virtual contact distance, and generating the character's visible screen range; The brightness of the computer screen is dynamically adjusted based on the screen anti-peeping requirements and the visible screen range of the person to perform computer screen anti-peeping control.

2. The computer screen control method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting computer lens images; performing high-precision image segmentation on the computer lens images to obtain a scene element set; Step S12: performing scene depth mapping through the scene element set to obtain a scene depth parameter; Step S13: reconstructing the space point cloud of the scene element set based on the scene depth parameter to generate a three-dimensional space point cloud; applying gridding processing to the three-dimensional space point cloud to obtain a three-dimensional usage scene model; Step S14: performing spatial density analysis on the three-dimensional usage scene model to obtain scene activity hotspots; identifying scene identifiers in the scene activity hotspots; Step S15: Based on the scene identifier, the three-dimensional usage scene model is distinguished by scene type to obtain the computer usage scene type; and a privacy assessment is performed according to the computer usage scene type to generate a scene privacy degree parameter.

3. The computer screen control method according to claim 2, characterized in that: Step S15 includes: Step S151: topologically deconstructing the three-dimensional usage scenario model based on the scenario identifier to obtain a scenario topology network; Step S152: performing a multi-scale convolution transformation on the scene topology network to obtain scene feature mapping data; Step S153: performing scene type recognition on the scene feature mapping data according to a preset application scene map to obtain a computer usage scene type; Step S154: Performing privacy risk scoring on the computer usage scenario category to obtain a scenario risk score; Step S155: Perform a privacy assessment on the scenario risk score based on a preset scenario risk score threshold to generate a scenario privacy parameter.

4. The computer screen control method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing threshold comparison on the scene privacy parameters based on a preset privacy threshold to obtain a privacy determination result; Step S22: mapping the anti-peeping requirement to the three-dimensional usage scenario model according to the privacy determination result to generate the screen anti-peeping requirement; Step S23: based on the screen anti-peeping requirement, perform portrait detection on the computer lens image, correct the lighting factor of the computer lens image, and obtain a balanced lighting image; detect the face area of ​​the balanced lighting image; locate the key points of the face area, and generate face key point data; Step S24: estimating the initial portrait location according to the face posture data to obtain an initial portrait location; integrating the initial portrait location to generate a portrait location collection.

5. The computer screen control method according to claim 4, characterized in that: Step S23 includes the following steps: Step S231: performing portrait detection on the computer lens image based on the screen anti-peeping requirement, and performing brightness histogram equalization on the computer lens image to obtain a brightness balanced image; Step S232: performing contrast enhancement on the uniform brightness image to generate an enhanced lens image; performing illumination compensation on the enhanced lens image to obtain a balanced illumination image; Step S233: performing multi-scale face detection on the balanced illumination image to generate a candidate face region; performing boundary redrawing processing on the candidate face region to obtain a face region; Step S234: extracting feature points from the face area to obtain face feature point data; performing grid deformation mapping on the face feature point data to obtain grid deformation data; Step S235: performing key point regression on the mesh deformation data to obtain key point positioning data; performing coordinate system conversion on the key point positioning data to obtain face key point data.

6. The computer screen control method according to claim 1, characterized in that: The step S3 of calculating the virtual contact distance of the three-dimensional usage scene model based on the portrait positioning collection includes: Convert the three-dimensional usage scene model into a positioning coordinate system; Project the portrait positioning collection into the positioning coordinate system to obtain the user space coordinates; Estimating the neighbor volume based on the user space coordinates to generate the interactive space volume; Perform gradient projection transformation on the interaction space volume to generate multi-level contact probability parameters; Perform boundary constraint filtering on multi-level contact probability parameters to obtain contact boundary data; The contact distance is calculated for the contact boundary data to generate the character's virtual contact distance.

7. The computer screen control method according to claim 1, characterized in that: The contact sight simulation according to the virtual contact distance of the character in step S3 includes: The actual distance between characters is deduced based on the virtual contact distance between characters, so as to obtain the actual contact distance between characters; Based on the actual contact distance of the person, the stereo distance of the portrait positioning collection is updated to generate the stereo contact distance of the person; Perform line of sight projection simulation according to the three-dimensional contact distance of the character. During the line of sight projection process, set the projection angle to 25-35 degrees and the maximum projection distance to 2.5-3.5 meters to obtain the initial line of sight vector; Perform difference smoothing on the initial sight line vector to generate a continuous sight line curve; The continuous sight line curve is reconstructed into three-dimensional trajectory to obtain the sight line trajectory of the character.

8. The computer screen control method according to claim 1, characterized in that: The step S4 of analyzing the screen visible range of the computer screen according to the person's sight line trajectory and the person's virtual contact distance includes: Perform spatial projection conversion on the sight line of the character to obtain the screen plane projection, wherein the screen width is set to 1900-1940 pixels, the height is set to 1060-1100 pixels, the parameters in the projection process are set to the focal length of 45mm-55mm, and the projection angle of 55-65 degrees; Cross-correlate the screen plane projection and the character's virtual contact distance to generate a combined viewing distance index; Construct a frustum model based on the combined index of viewing distance; perform field of view generalization processing on the frustum model to obtain a three-dimensional field of view range; Perform plane interception mapping on the three-dimensional field of view to generate a screen field of view coverage matrix, where each value in the matrix is ​​0 for an invisible area and 1 for a visible area; The character's perspective is inferred based on the screen field of view coverage matrix to generate the character's visible screen range.

9. The computer screen control method according to claim 1, characterized in that: The step S4 of dynamically adjusting the brightness of the computer screen based on the screen anti-peeping requirements and the visible screen range of the person includes: The visible screen range of the characters is screened to the maximum extent to obtain the main character's perspective, where the main character's perspective is a viewing angle range of 90-110° and a maximum viewing distance of 1.8-2.2 meters; based on the main character's perspective, the visible screen range of the characters is extracted to obtain the visible screen range of other characters, where the visible screen range of other characters is 70-90° and a maximum viewing distance of 1.6-2.0 meters; Based on the screen anti-peeping requirements, the visible screen range of other characters is cut to expose the screen, with a cutting accuracy of 5-10 pixels, to generate exposed screen units; Perform brightness gradient design on the exposed screen unit to obtain a brightness adjustment gradient; Performing brightness adjustment simulation on a computer screen based on a brightness adjustment gradient to generate brightness of the computer screen at various levels; The exposed screen unit is gradually controlled by adjusting the brightness of the computer screen at different levels. The maximum brightness change is set between 0-255. The unit brightness near the edge of the screen is set to 80-120cd, and the unit brightness in the center area is set to 180-220cd to implement computer screen anti-peeping control.

10. A computer screen control system, characterized in that: Used to execute the computer screen control method as claimed in claim 1, the computer screen control system comprises: An image acquisition module is used to acquire computer lens images; perform scene modeling on the computer lens images to obtain a three-dimensional usage scene model; perform scene privacy evaluation on the three-dimensional usage scene model to generate scene privacy degree parameters; The anti-peeping analysis module is used to analyze the anti-peeping requirements of the three-dimensional usage scene model according to the scene privacy parameters and generate the screen anti-peeping requirements; based on the screen anti-peeping requirements, the computer lens image is subjected to portrait detection, the spatial position and sight direction of the person are extracted, and a portrait positioning collection is generated; The contact calculation module is used to calculate the virtual contact distance of the three-dimensional usage scene model based on the portrait positioning collection to generate the virtual contact distance of the character; and to simulate the contact line of sight according to the virtual contact distance of the character to obtain the line of sight trajectory of the character; The brightness control module is used to analyze the screen visible range of the computer screen according to the person's line of sight and the person's virtual contact distance, and generate the person's visible screen range; based on the screen anti-peeping requirements and the person's visible screen range, the brightness of the computer screen is dynamically adjusted to perform computer screen anti-peeping control.

Images