Tower control visual efficiency optimization method and device based on improved gaze point rendering

Abstract

The application discloses a tower control visual field efficiency optimization method based on improved gaze point rendering, relates to the technical field of air traffic control simulation, and comprises the following steps: acquiring objects in a current scene, control task data, interactive data of a controller and a simulated tower system, and a current gaze point rendering picture; performing feature extraction on the objects in the current scene and the control task data; inputting the extracted feature data and the interactive data into a preset Bayesian information gain framework to output an attention distribution of a current controller; calculating parameters of a Gabor kernel based on the Gabor kernel function according to the attention distribution of the current controller; generating Gabor noise based on the parameters of the Gabor kernel, superimposing the Gabor noise on the current gaze point rendering picture, and forming a current final rendering image. The application significantly improves the rendering efficiency and enhances the immersion and visual reality of a user.

Description

Technical Field

[0001] This invention relates to the field of air traffic control simulation technology, and in particular to a method and apparatus for optimizing tower control visual efficiency based on improved gaze point rendering. Background Technology

[0002] As air traffic control tasks become increasingly complex, tower control simulation systems play a crucial role in air traffic controller training. These systems simulate the real tower environment, helping controllers train in a safe and controlled setting. To enhance the immersive experience and training effectiveness of the simulation system, the rendering quality of the 3D visual system is paramount. However, with the increasing geometric complexity and texture detail of rendered scenes, the rendering burden and hardware requirements have significantly increased, creating a bottleneck in rendering efficiency.

[0003] Traditional rendering methods often rely on full-resolution rendering, which, while ensuring high-quality scene display, requires significant computational resources, leading to low rendering efficiency and frame rates that cannot meet real-time requirements. This problem is particularly pronounced in high-precision simulation scenarios, especially in tower control simulations that require meticulous rendering of every detail. To improve rendering efficiency and reduce hardware requirements, researchers have tried various optimization methods, such as adjusting region resolution based on scene importance or using simplification techniques to reduce rendering overhead. However, these methods either sacrifice image quality or fail to adequately consider the characteristics of the human visual system and the attention distribution of controllers, resulting in poor visual perception in simulation training and severely impacting training effectiveness.

[0004] In recent years, foveated rendering, as a technique for optimizing rendering quality, has been widely used in virtual reality and computer graphics. Based on the physiological structure of the human eye, this technique simulates the differences in attention paid to different visual regions. By rendering the central visual field (foveation point) with high quality and the peripheral areas with lower quality, it significantly improves rendering efficiency. Although this method has achieved good results in many applications, foveated rendering still faces the following challenges in control tower simulation systems: 1. Because control tower tasks require controllers to continuously focus on multiple scene areas, the distribution of attention is highly dynamic. Therefore, accurately predicting and adapting to changes in controller attention becomes crucial for improving rendering efficiency. 2. Traditional foveated rendering methods use simple radial blurring in the peripheral areas, which may lead to loss of image details, affecting the situational awareness of control trainees and reducing user immersion and training effectiveness. Summary of the Invention

[0005] This invention proposes a method and apparatus for optimizing the visual efficiency of air traffic control based on improved foveated rendering, in order to solve the problems of low rendering efficiency, low immersion, and poor training effect in existing air traffic control simulation systems.

[0006] The present invention achieves the above objectives through the following technical solutions:

[0007] This invention provides a method for optimizing the visual efficiency of control towers based on improved foveated rendering, comprising:

[0008] Acquire objects in the current scene, control task data, interaction data between the controller and the simulated control tower system, and the current gaze point rendering screen;

[0009] Feature extraction is performed on objects and control task data in the current scene;

[0010] The extracted feature data and interaction data are input into a preset Bayesian information gain framework to output the current controller's attention distribution. The Bayesian information gain framework is pre-inputted with a prior distribution based on expert experience. The prior distribution includes the importance ranking of objects in airport simulation scenarios under different control tasks and the conditional probability distribution obtained by modeling the eye-tracking data of experts in the practical teaching process.

[0011] The parameters of the Gabor kernel are calculated based on the Gabor kernel function according to the current controller attention distribution.

[0012] Gabor noise is generated based on the parameters of the Gabor kernel, and the Gabor noise is superimposed on the current foveated rendering image to form the current final rendering image.

[0013] Furthermore, the extracted feature data and interaction data are input into a pre-defined Bayesian information gain framework to output the current controller's attention distribution, including:

[0014] The input is identified as two behavioral models: saccade and gaze, based on the discrete threshold recognition algorithm.

[0015] The two behavioral models are used as likelihood functions in Bayes' theorem. Bayes' theorem is used to update the spatial information, which is the set of all objects in the simulated control tower scenario where the attention allocation of the control trainer exceeds the threshold.

[0016] Calculate the posterior distribution based on the updated spatial information;

[0017] Calculate the uncertainty difference between the prior and posterior distributions, use the uncertainty difference as the updated information gain, output the updated information gain as the current controller's attention distribution, and use the posterior distribution obtained in this calculation to update the prior distribution for subsequent loops.

[0018] Furthermore, the formula for calculating the parameters g(x,y,ξ) of the Gabor kernel is as follows:

[0019]

[0020]

[0021] x and y represent the coordinates of the current pixel in the screen pixels, ξ is the controller's attention distribution on the object where the current pixel is located, and K(x,ξ) is the amplitude of the local noise. F0(x,ξ) represents the width of the Gaussian envelope, and its value is positively correlated with the eccentricity. Let represent the amplitude and direction of the local noise frequency, and τ be a scaling factor. It is the position x in the Laplacian pyramid decomposition L of the fixation point image. The value of the level, This represents perceptual normalization performed using a contrast sensitivity function, whose parameter f L (x) and f t (x) represents the spatial and temporal frequencies at point x in the rendered image at the current gaze point, respectively.

[0022] Furthermore, K(x,ξ) is a random variable that follows a log-normal distribution. The random variable The expression is:

[0023]

[0024]

[0025]

[0026] Where Z is a normal distribution (0,1), Standard deviation Let be the variance, and k be a user-defined parameter in the range (0,1). F is the scaling factor. L (x) and F H (x) represents the lowest and highest frequencies that the user can recognize in the system, respectively.

[0027] Furthermore, the formula for generating Gabor noise N(x,y,ξ) based on the parameters of the Gabor kernel is as follows:

[0028]

[0029]

[0030] (x i ,y i ) represents the random position of the Gabor kernel g, M is the mask related to the attention parameter ξ, and the Gabor kernel g is a pulse. i Let ξ0 represent the pulse, and ξ0 be the preset attention distribution threshold. It is a random sampling binary weight in terms of amplitude.

[0031] Furthermore, the construction of control tasks includes:

[0032] Construct control tasks categorized into emergency response, monitoring and command, and planning and coordination.

[0033] Special aircrew response missions are interspersed within the control mission processes of emergency response, monitoring and command, and planning and coordination missions. These special aircrew response missions include missions involving sudden mechanical failures of aircraft, flight conflict missions, and missions involving unauthorized intrusion into airspace.

[0034] Furthermore, based on the new prior distribution, new Gabor noise is generated. This new Gabor noise is then superimposed on the next frame's gaze point rendering image to generate the corresponding final rendered image. The final rendered image is then used as a real-time output for the user to observe. This process is repeated continuously until the simulation training ends.

[0035] This invention discloses a tower control visual efficiency optimization device based on improved foveated rendering, comprising:

[0036] The acquisition module is used to acquire objects in the current scene, control task data, interaction data between the controller and the simulated control tower system, and the current gaze point rendering screen.

[0037] The extraction module is used to extract features from objects and control task data in the current scene;

[0038] The gain module is used to input the extracted feature data and interaction data into a preset Bayesian information gain framework and output the current controller's attention distribution. The Bayesian information gain framework is pre-inputted with a prior distribution based on expert experience. The prior distribution includes the importance ranking of objects in airport simulation scenarios under different control tasks and the conditional probability distribution obtained by modeling the eye-tracking data of experts in the practical teaching process.

[0039] The calculation module is used to calculate the parameters of the Gabor kernel based on the Gabor kernel function according to the current controller's attention distribution;

[0040] The overlay module is used to generate Gabor noise based on the parameters of the Gabor kernel, and overlay the Gabor noise with the current gaze point rendering image to form the current final rendering image.

[0041] The beneficial effects of this invention are as follows:

[0042] This invention proposes a method and apparatus for optimizing the visual efficiency of air traffic control based on improved gaze-based rendering. By combining user eye-tracking data with a Bayesian information gain framework (BIG framework), it accurately predicts the controller's attention distribution, achieving adaptive adjustment of rendering quality. By using high-quality rendering in the user's gaze area while simultaneously rendering the surrounding areas with low quality, rendering efficiency is significantly improved, and hardware requirements are reduced. Local Gabor noise synthesis technology is used for texture reconstruction, effectively restoring visual details in low-quality rendered areas and enhancing user immersion and visual realism. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0044] Figure 2 This is a schematic diagram of the framework of the user attention distribution prediction model of the present invention;

[0045] Figure 3 This is a schematic diagram of the gaze point rendering module of the present invention;

[0046] Figure 4 This is a schematic diagram of the texture reconstruction module of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0048] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0049] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0050] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0051] A method and apparatus for optimizing the visual efficiency of control tower based on improved foveated rendering, comprising:

[0052] S1: Extract features from the current control task to obtain the current control task feature data. The control task is constructed in real time based on the airport simulation scenario, which is built based on the site information of the actual airport.

[0053] The construction of a simulated control tower scenario requires reference to real airport geographic data, which is digitized using Geographic Information System (GIS) technology. The influence of terrain and buildings is taken into account, and functional areas such as runways and aprons are divided. High-precision models of aircraft, ground facilities, and other objects are created using 3D modeling software. Physical properties are assigned to the objects and behavioral logic code is written. Finally, environmental effects such as weather and lighting are simulated, and data interaction interfaces with eye tracking, control task generation, and rendering systems are developed to achieve seamless integration with the control tower visual efficiency optimization system.

[0054] S2: Obtain the current controller's eye movement data and the current gaze point rendering screen;

[0055] S3: The current controller's eye movement data and the control task feature data are fused together and input into a preset Bayesian information gain framework. The current controller's attention distribution is output. The Bayesian information gain framework is pre-inputted with a prior distribution based on expert experience. The prior distribution based on expert experience includes the importance ranking of objects in airport simulation scenarios under different control tasks and the conditional probability distribution obtained by modeling the eye movement data of experts in the practical teaching process.

[0056] S4: Calculate the parameters of the Gabor kernel based on the Gabor kernel function according to the current controller's attention distribution;

[0057] S5: Generate Gabor noise based on the parameters of the Gabor kernel, and superimpose the Gabor noise with the current gaze point rendering image to form the current final rendering image;

[0058] S6: Based on the new prior distribution, generate new Gabor noise. The new Gabor noise is used to superimpose with the next frame of foveated rendering to generate the corresponding final rendering image, until all foveated rendering images form the corresponding final rendering images.

[0059] like Figure 1 As shown, this invention specifically includes the following: Building an airport simulation scene in a rendering engine based on the actual site; constructing air traffic control tasks such as emergency response, monitoring and command, and planning and coordination; and interspersing special air situation handling (emergency) tasks such as sudden mechanical failures of aircraft, flight conflicts, and illegal intrusions into the air traffic control task process. Predicting attention distribution based on user eye-tracking data and training task feature data. Estimating Gabor noise parameters based on attention distribution prediction. Reconstructing textures using Gabor noise.

[0060] The flowchart for predicting user attention distribution is as follows: Figure 2 As shown, the specific implementation method is as follows:

[0061] To improve the accuracy of user attention distribution prediction, two prior distributions are introduced into the BIG (Bayesian information gain) framework. The first prior distribution is based on the importance of objects in the simulated control tower scenario during the control task. For example, when an emergency occurs during control operations, controllers usually quickly focus on the location of the emergency. Eye movement features are highly correlated with attention distribution, so it is assumed that objects with higher gaze duration, gaze frequency, and revisit frequency tend to receive more attention. The second prior distribution is modeled using eye movement data from experts during practical teaching, and input into the system as a probability distribution. This approach can effectively improve learner performance.

[0062] The BIG (Bayesian information gain) framework receives data from interactions between controllers and the simulated control tower system, as well as extracted feature data, as input. Within this framework, spatial information... For control trainees to exceed a threshold in simulated control tower scenarios where all their attention is allocated. object θ i The set of, where, , θ i It is an abstract representation of an object in a simulated control tower scene, consisting of binary tuples. ,in It is the unique identifier of an object in the scene, ξ iThe dynamic probability that object i is noticed by the controller is represented by X. The output of the Bayesian information gain framework, i.e., the current controller's (i.e., user's) attention distribution, is defined by the quadruple (O, D, T, A), where O is the object in the simulated control tower system, D is the currently executed teaching task, T is the in-game execution time of the current teaching task, and A represents the controller's gaze state. The user-input data Y represents the interaction data between the controller and the simulated control tower system, as well as the extracted feature data. The interaction data includes eye-tracking data, current training task feature data, current task time data, scheduling operations performed by the controller, and the controller's proficiency data. Among them, the training task feature data, similar to the prior distribution, is given by an expert experience model, while the eye-tracking data is the real-time eye-tracking data during system operation. The eye-tracking data includes the following three indicators: gaze duration, gaze count, and revisit count. This paper uses the Discrete Thresholding (I-DT) algorithm to measure the relevant indicators of the eye-tracking data. This algorithm has high robustness to noise and artifacts in the eye-tracking data and can effectively identify gaze and saccade behaviors. The control task characteristic information includes: task ID, whether there is a special situation, special situation ID (if present), current process status within the task, and system data.

[0063] The Bayesian information gain framework, based on a discrete threshold recognition algorithm, is used to identify controller eye-tracking data into two behavioral models: saccades and gazes. The expressions for these two behavioral models are as follows:

[0064] (1)

[0065] (2)

[0066] In the formula, , This represents the similarity between the object at the center of the field of view and objects in the scene under both salivation and staring modes. Its value is the reciprocal of the weighted Euclidean distance of the object on the task feature data dimension. The system converts the similarity value into a likelihood using the softmax function, the formula of which is:

[0067] (3)

[0068] During framework runtime, the following steps are continuously repeated for dynamic updates:

[0069] The user's input data is identified using the corresponding behavioral model;

[0070] Using the two behavioral models as likelihood functions in Bayes' theorem, we apply Bayes' theorem to our beliefs about spatial information. Update;

[0071] posterior distribution Alternative Prior Distribution .

[0072] The framework uses a prior distribution and posterior distribution The difference in uncertainty between them is taken as information gain (IG), and its formula is:

[0073] (4)

[0074] Uncertainty is calculated using Shannon's entropy function, the formula of which is:

[0075] (5)

[0076] in Represents the probability distribution of the parameters. This represents the entropy value of the information space.

[0077] Substituting the priors, the formula for information gain can be expanded as follows:

[0078] (6)

[0079] In actual computation, eye-tracking data is discontinuous due to limitations imposed by the refresh rate of the simulated control tower visual system and the sampling frequency of the eye tracker itself. Therefore, the iteration process of the BIG framework is controlled by a clock signal within the system to ensure the accuracy of attention distribution prediction. Simultaneously, to optimize computational efficiency, the system employs a spatial information sharing mechanism, where multiple viewports share the same spatial information set to characterize the user's overall attention distribution within the current virtual control tower scene. This ultimately outputs the user attention prediction result, providing crucial attention distribution information to the subsequent foveated rendering module. This information guides the module in adaptively adjusting the rendering quality of different regions within the visual scene, thereby optimizing the efficiency of the control tower visual system.

[0080] Figure 3 The complete rendering pipeline of this method is demonstrated. The noise parameter estimation module plays a key role in the tower control visual efficiency optimization system, and its specific implementation is as follows:

[0081] The calculated attention distribution parameter ξ is used for parameter estimation of the Gabor convolution kernel. In this module, the formula for the Gabor kernel is:

[0082] (7)

[0083] In the formula, x and y represent the coordinates of the current pixel in the screen pixels, ξ is the controller's attention distribution on the object where the current pixel is located, K(ξ) is the amplitude of local noise; a is the width of the Gaussian envelope, whose value is positively correlated with the eccentricity; F0(ξ) and The amplitude and direction of local noise frequencies are defined. When the attention level is below a set threshold ξ0, the region is not reconstructed. A higher ξ indicates that the user is allocating more attention to the object, requiring more resources for texture detail reconstruction.

[0084] The amplitude K(x,ξ) can be expressed as:

[0085] (8)

[0086] In the formula, τ is a scaling factor, which allows for parameterized control. It is the position x in the Laplacian pyramid decomposition L of the fixation point image. The value of the level; This represents perceptual normalization performed using the Contrast Sensitivity Function (CSF), whose parameter f L (x) and f t (x) represents the spatial and temporal frequencies at point x in the rendered image at the current gaze point, respectively.

[0087] In this model, the frequency F0(x,ξ) is modeled as a random variable that follows a log-normal distribution. Its specific definition is:

[0088] (9)

[0089] Where Z is a normal distribution (0,1) and the standard deviation is... and variance They are defined as follows:

[0090] (10)

[0091] (11)

[0092] Where k is a user-defined parameter (0,1); F is the scaling factor; L (x) and F H (x) represents the lowest and highest frequencies that the user can recognize in the system, respectively.

[0093] The orientation is obtained by convolving the gaze point image with horizontal and vertical Sobel filters using the Sobel operator.

[0094] In the post-processing stage, the estimated Gabor kernel parameters are used to generate corresponding Gabor noise according to formula (12), and this noise is superimposed on the current gaze point rendering image and output to the screen to form the final rendered image. Based on the new prior distribution, new Gabor noise is generated. This new Gabor noise is used to generate the corresponding final rendered image after being superimposed on the next frame gaze point rendering image, thereby achieving adaptive adjustment of the rendering quality of different regions. While ensuring that the user's perceived image quality remains unchanged, this effectively improves the rendering efficiency of the visual system, reduces hardware costs, and meets the actual needs of tower control simulation.

[0095] Texture reconstruction is crucial for improving the visibility quality of control towers. Its specific implementation is as follows:

[0096] This module defines the noise N used for texture reconstruction as a random impulse process, where the Gabor kernel g is an impulse whose random position follows a Poisson distribution with a mean of λ, denoted as (x i ,y i ).

[0097] (12)

[0098] (13)

[0099] Where (x) i ,y i ) represents the random position of the Gabor nucleus g, where g is a pulse. i Let ξ0 represent the pulse, M be the preset attention distribution threshold, and M be a mask related to the attention parameter ξ, used to skip pixels that do not need to be processed. It is a random sampling binary weight in amplitude, used to control the average brightness of the original image without changing the synthesized noise.

[0100] After generating noise, it is convolved with the foveated rendering image in the rendering pipeline to obtain a foveated rendering image with enhanced peripheral vision, thereby reconstructing the texture details of objects in the peripheral visual region. Throughout the process, the texture reconstruction module can flexibly control the resource allocation required for texture reconstruction according to the attention distribution of objects, avoiding over-reconstruction of low-attention areas, effectively improving rendering efficiency, and enhancing the visual perception quality for users during the simulated control tower process, thus enhancing immersion and realism.

[0101] In one embodiment, the construction of the control task includes:

[0102] Construct control tasks categorized into emergency response, monitoring and command, and planning and coordination.

[0103] Special aircrew response missions are interspersed within the control mission processes of emergency response, monitoring and command, and planning and coordination missions. These special aircrew response missions include missions involving sudden mechanical failures of aircraft, flight conflict missions, and missions involving unauthorized intrusion into airspace.

[0104] In one embodiment, new Gabor noise is generated based on the new prior distribution. This new Gabor noise is then superimposed on the next frame of the gaze point rendering to generate the corresponding final rendered image. The final rendered image is then used as a real-time output for the user to observe. This operation is repeated continuously until the simulation training ends.

[0105] The present invention also provides a tower control visual efficiency optimization device based on improved foveated rendering, comprising:

[0106] The acquisition module is used to acquire objects in the current scene, control task data, interaction data between the controller and the simulated control tower system, and the current gaze point rendering screen.

[0107] The extraction module is used to extract features from objects and control task data in the current scene;

[0108] The gain module is used to input the extracted feature data and interaction data into a preset Bayesian information gain framework and output the current controller's attention distribution. The Bayesian information gain framework is pre-inputted with a prior distribution based on expert experience. The prior distribution includes the importance ranking of objects in airport simulation scenarios under different control tasks and the conditional probability distribution obtained by modeling the eye-tracking data of experts in the practical teaching process.

[0109] The calculation module is used to calculate the parameters of the Gabor kernel based on the Gabor kernel function according to the current controller's attention distribution;

[0110] The overlay module is used to generate Gabor noise based on the parameters of the Gabor kernel, and overlay the Gabor noise with the current gaze point rendering image to form the current final rendering image.

[0111] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing the visual efficiency of control tower based on improved foveated rendering, characterized in that, include: Acquire objects in the current scene, control task data, interaction data between the controller and the simulated control tower system, and the current gaze point rendering screen; Feature extraction is performed on objects and control task data in the current scene; The extracted feature data and interaction data are input into a preset Bayesian information gain framework to output the current controller's attention distribution. The Bayesian information gain framework is pre-inputted with a prior distribution based on expert experience. The prior distribution includes the importance ranking of objects in airport simulation scenarios under different control tasks and the conditional probability distribution obtained by modeling the eye-tracking data of experts in the practical teaching process. The parameters of the Gabor kernel are calculated based on the Gabor kernel function according to the current controller attention distribution. Gabor noise is generated based on the parameters of the Gabor kernel, and the Gabor noise is superimposed on the current foveated rendering image to form the current final rendered image.

2. The method for optimizing the visual efficiency of control tower based on improved foveated rendering according to claim 1, characterized in that, The extracted feature data and interaction data are input into a pre-defined Bayesian information gain framework to output the current controller's attention distribution, including: The input is identified as two behavioral models: saccade and gaze, based on the discrete threshold recognition algorithm. The two behavioral models are used as likelihood functions in Bayes' theorem. Bayes' theorem is used to update the spatial information, which is the set of all objects in the simulated control tower scenario where the attention allocation of the control trainer exceeds the threshold. Calculate the posterior distribution based on the updated spatial information; Calculate the uncertainty difference between the prior and posterior distributions, use the uncertainty difference as the updated information gain, output the updated information gain as the current controller's attention distribution, and use the posterior distribution obtained in this calculation to update the prior distribution for subsequent loops.

3. The method for optimizing the visual efficiency of control tower based on improved foveated rendering according to claim 1, characterized in that, The parameters of the Gabor core The calculation formula is: ， ， x and y represent the coordinates of the current pixel in the screen pixels, respectively. This represents the distribution of the controller's attention on the object containing the current pixel. The amplitude of local noise. denoted as the width of the Gaussian envelope, whose value is positively correlated with the eccentricity. and The amplitude and direction of the local noise frequency. Scaling factor It is the position x in the Laplacian pyramid decomposition L of the fixation point image. The value of the level, This represents perceptual normalization performed using a contrast sensitivity function, whose parameters... and The spatial and temporal frequencies at point x in the image are rendered for the current gaze point, respectively.

4. The method for optimizing the visual efficiency of control tower based on improved foveated rendering according to claim 3, characterized in that, The A random variable that follows a log-normal distribution The random variable The expression is: ， ， ， Where Z is a normal distribution (0,1), Standard deviation Let be the variance, and k be a user-defined parameter in the range (0,1). Scaling factor and These are the lowest and highest frequencies that users can recognize in the system, respectively.

5. The method for optimizing the visual efficiency of control tower based on improved foveated rendering according to claim 3, characterized in that, Gabor noise is generated based on the parameters of the Gabor kernel. The formula is as follows: ， Let M represent the random position of the Gabor kernel g, and M be the attention parameter. The relevant mask, Gabor kernel g as a pulse, g i Indicates a pulse. The preset attention distribution threshold, It is a random sampling binary weight in terms of amplitude.

6. The method for optimizing the visual efficiency of control tower based on improved foveated rendering according to claim 1, characterized in that, The construction of control tasks includes: Construct control tasks categorized into emergency response, monitoring and command, and planning and coordination. Special aircrew response missions are interspersed within the control mission processes of emergency response, monitoring and command, and planning and coordination missions. These special aircrew response missions include missions involving sudden mechanical failures of aircraft, flight conflict missions, and missions involving unauthorized intrusion into airspace.

7. The method for optimizing the visual efficiency of control tower based on improved foveated rendering according to claim 1, characterized in that, Based on the new prior distribution, new Gabor noise is generated. This new Gabor noise is then superimposed on the next frame's gaze point rendering image to generate the corresponding final rendering image. The final rendering image is then used as a real-time output for the user to observe. This process is repeated until the simulation training ends.

8. An apparatus for optimizing the visual efficiency of control tower based on improved foveated rendering as described in any one of claims 1-7, characterized in that, include: The acquisition module is used to acquire objects in the current scene, control task data, interaction data between the controller and the simulated control tower system, and the current gaze point rendering screen. The extraction module is used to extract features from objects and control task data in the current scene; The gain module is used to input the extracted feature data and interaction data into a preset Bayesian information gain framework and output the current controller's attention distribution. The Bayesian information gain framework is pre-inputted with a prior distribution based on expert experience. The prior distribution includes the importance ranking of objects in airport simulation scenarios under different control tasks and the conditional probability distribution obtained by modeling the eye-tracking data of experts in the practical teaching process. The calculation module is used to calculate the parameters of the Gabor kernel based on the Gabor kernel function according to the current controller's attention distribution; The overlay module is used to generate Gabor noise based on the parameters of the Gabor kernel, and overlay the Gabor noise with the current foveated rendering image to form the current final rendering image.

Images