Distributed simulation model data acquisition method and system for aircraft attitude control
By performing image processing and network model optimization on the aircraft operation video data, the problem of complex and inaccurate acquisition of aircraft attitude control simulation data in the prior art is solved, and the rapid and accurate data acquisition of aircraft attitude control is achieved, and distributed simulation is improved Model simulation accuracy and efficiency.
Patent Information
- Application Number
- CN202411230381.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-09-04
AI Technical Summary
In the prior art, the simulation data acquisition method of aircraft attitude control is complex and inaccurate, resulting in low computing efficiency of simulation models and difficult to meet the high accuracy requirements of modern aircraft design.
By acquiring the aircraft's running video data, performing image processing to filter out key image sequences, and introducing network models to improve data accuracy, the loss function is optimized by comprehensive attention weights and regularization terms to extract the aircraft's state data.
It realizes fast and accurate data acquisition of aircraft attitude control, improves the simulation accuracy and efficiency of distributed simulation models, and enhances the reliability of subsequent simulation data.
Smart Images

Figure CN118734460B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of aircraft simulation, and more specifically, to a distributed simulation model data acquisition method and system for aircraft attitude control. Background Art
[0002] Aircraft design is a complex system engineering involving multiple disciplines such as aerodynamics, structure, control, materials, and electronics. With the rapid development of modern aircraft design technology, the coupling relationship between these disciplines has become increasingly close. In order to fully consider the synergy between disciplines, fully apply the multidisciplinary design optimization concept, improve the overall performance of the aircraft, shorten the design cycle and reduce costs, a distributed multidisciplinary integrated simulation system came into being. For such a complex project, how to improve the accuracy and efficiency of simulation model calculations is a problem that must be solved to improve the efficiency of aircraft design.
[0003] The accuracy of simulation model calculations is closely related to the accuracy of simulation data. Therefore, how to obtain accurate simulation data is particularly important.
[0004] In the related technology, there is a patent application document with publication number CN117130288A, entitled Method and Device for Determining Aircraft Attitude Control Simulation Modeling Data and Storage Medium. It discloses a method for obtaining accurate thrust and torque in the simulation test of an aircraft attitude controller, which leads to an imperfect simulation effect. It provides a basis for modeling and simulation calculation for mathematical simulation, semi-physical simulation and Monte Carlo simulation tests of attitude control systems of aircraft using attitude control engines as actuators.
[0005] The above scheme provides a method for obtaining accurate simulation data, but the above method obtains data through data calculation, and its acquisition method is complicated, and the efficiency of obtaining simulation data under big data is obviously low.
[0006] Therefore, it is particularly important to quickly and accurately obtain the simulation data of the aircraft and improve the simulation accuracy of the subsequent distributed simulation model. Summary of the invention
[0007] The purpose of the present invention is to propose a distributed simulation model data acquisition method and system for aircraft attitude control, so as to solve the problem that the simulation data acquisition method in the prior art is complex and inaccurate; to this end, the present invention provides solutions in the following two aspects.
[0008] In a first aspect, the present invention provides a method for acquiring distributed simulation model data for aircraft attitude control, comprising:
[0009] Acquire a grayscale image sequence of the target aircraft operation video data; filter out clear images from the grayscale image sequence to obtain a clear image sequence;
[0010] Calculate the grayscale mean of the difference image of two consecutive frames of images in the clear image sequence, and select two frames of images when the grayscale mean is greater than the first threshold;
[0011] When two frames of images are adjacent in a grayscale image sequence, the first degree of the previous frame of image is the difference between the grayscale mean of the two adjacent frames of image and the first threshold; when two frames of image are not adjacent in the grayscale image sequence, the second degree of the previous frame of image is the difference between the grayscale mean of the two non-adjacent frames of image and the second threshold, the second threshold is greater than the first threshold, the third degree of the blurred image between the two non-adjacent frames of image is the average of the target degree and the second degree, the target degree is the difference between the grayscale mean of the two non-adjacent frames of image and the first threshold; all blurred images between two non-adjacent frames of image and clear image sequence constitute a key image sequence;
[0012] Input the key image sequence into the network model and output the simulation data of the current aircraft; the loss function of the network model for: ; For the Mean square error loss of key images; is the regularization term; is the regularization coefficient, is the total number of key images, For the The key images correspond to the key levels of the key images, and the key levels are the first level, the second level, or the third level.
[0013] The above scheme first obtains a key image sequence by performing image processing on the acquired video data; then the criticality of each key image in the key image sequence is introduced into the loss function of the network model to improve the accuracy of the network model, thereby accurately extracting the state data of the aircraft in the key image sequence, thereby improving the accuracy of obtaining simulation data.
[0014] Optionally, the first threshold is the grayscale mean corresponding to the 20th percentile in the grayscale mean sequence after the grayscale means of each difference image are sorted from small to large, and the second threshold is the grayscale mean corresponding to the 40th percentile in the grayscale mean sequence. Optionally, the first degree is the ratio of the first difference to the grayscale mean, wherein the first difference is the difference between the grayscale mean of two adjacent frames of images and the first threshold; the second degree is the ratio of the second difference to the grayscale mean, wherein the second difference is the difference between the grayscale mean of two non-adjacent frames of images and the second threshold.
[0015] The above scheme can quantify the importance of different images by acquiring the first degree and the second degree of different images, thereby providing a data basis for the subsequent acquisition of aircraft status data.
[0016] Optionally, before obtaining the criticality, the method further includes a step of normalizing the obtained first degree, second degree and third degree.
[0017] The above scheme can unify the importance values of different images to the same scale through normalization processing.
[0018] Optionally, the mean square error loss for: ;
[0019] In the formula, For the The predicted value of key images, G is the total number of key images; For the The labels of the key images; For the The comprehensive attention weight of key images, , It is The temporal attention weight of the c-th channel of the convolved feature map of the key image; It is The convolved feature map of the key image is the spatial attention weight of the cth channel, where C is the total number of channels.
[0020] The above scheme introduces comprehensive attention weights to better capture the key information of aircraft dynamic features and improve the accuracy and efficiency of extracting aircraft dynamic feature data in key frame data under communication delay.
[0021] Optionally, the simulation data includes a pitch angle, a roll angle, a yaw angle, a track angle and a flight path angle of the aircraft.
[0022] Optionally, the specific process of filtering out clear images from the grayscale image sequence and obtaining the clear image sequence is as follows:
[0023] Calculate the clarity of each grayscale image in the grayscale image sequence;
[0024] In response to the clarity being greater than the clarity threshold, the corresponding grayscale image is marked as a clear image, otherwise, the corresponding grayscale image is marked as a blurred image.
[0025] The above scheme can obtain a clearer image by screening each grayscale image.
[0026] Optionally, before obtaining the grayscale image sequence, the method further includes sequentially performing grayscale processing on each frame of the image in the target aircraft operation video data.
[0027] In a second aspect, the present invention provides a distributed simulation model data acquisition system for aircraft attitude control, which comprises:
[0028] processor;
[0029] A memory stores computer instructions for acquiring distributed simulation model data for aircraft attitude control. When the computer instructions are executed by the processor, the system executes the above-mentioned distributed simulation model data acquisition method for aircraft attitude control.
[0030] The beneficial effects of the present invention are:
[0031] The solution of the present invention first obtains a key image sequence by performing image processing on the acquired video data; then accurately extracts the status data of the aircraft in the key image sequence by improving the loss function of the network model, and uses the extracted status data of the aircraft as subsequent simulation data, thereby improving the accuracy of obtaining the simulation data. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:
[0033] Figure 1 The flowchart schematically shows the steps of the distributed simulation model data acquisition method for aircraft attitude control in this embodiment;
[0034] Figure 2 The structural block diagram of the distributed simulation model data acquisition system for aircraft attitude control in this embodiment is schematically shown. DETAILED DESCRIPTION
[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0036] The present invention is aimed at the fact that the attitude control of the aircraft needs to be carried out through simulation experiments, and the mathematical model used in the simulation experiment should be as consistent as possible with the actual flight process. The accuracy of the mathematical model is closely related to the simulation data, and the accuracy of the simulation data is also particularly important.
[0037] Therefore, based on the above content, the present invention provides a distributed simulation model data acquisition method and system for aircraft attitude control, so as to quickly and accurately acquire simulation data of a distributed simulation model for an aircraft.
[0038] Specifically, Figure 1 As shown, the distributed simulation model data acquisition method for aircraft attitude control in this embodiment includes the following steps:
[0039] Step S1, obtaining a grayscale image sequence of target aircraft operation video data.
[0040] In this embodiment, a large amount of video data and VR data containing the aircraft operation are captured through the Internet. These data contain the flight status of the aircraft under different weather and lighting conditions, including flight attitude and flight trajectory, where the flight attitude includes pitch angle, roll angle, and yaw angle, and the flight trajectory includes track angle and flight path angle.
[0041] Among them, a large number of captured target aircraft are of the same or similar models as the aircraft to be simulated.
[0042] After obtaining a large amount of video data of the target aircraft in operation, it is also necessary to perform grayscale processing on each frame image in the video data to obtain a grayscale image sequence.
[0043] Step S2, screening out clear images from the grayscale image sequence to obtain a clear image sequence; calculating the grayscale mean of the difference image of two consecutive frames of images in the clear image sequence, and selecting two frames of images when each grayscale mean is greater than a first threshold.
[0044] In this embodiment, the method for selecting a clear image from a grayscale image sequence is:
[0045] Calculate the clarity of each grayscale image in the grayscale image sequence;
[0046] In response to the clarity being greater than the clarity threshold, the corresponding grayscale image is marked as a clear image, otherwise, the corresponding grayscale image is marked as a blurred image.
[0047] The clarity can be obtained by using the Tenengrad gradient method, the Laplacian gradient method or the variance method. Since the calculation of clarity is a prior art, it will not be described in detail here. Of course, as another embodiment, the clarity of the image can also be evaluated by calculating the Laplacian operator of the image and analyzing its variance.
[0048] Taking the calculation of the Laplace operator of an image and the analysis of its variance as an example, the clarity threshold in this embodiment may select the clarity corresponding to the 25th percentile of the Laplace variance distribution; of course, the clarity threshold may also be set directly based on experience.
[0049] The above percentile is to sort a set of data from small to large and calculate the corresponding cumulative percentile position, which is the position in a set of data; for example, the p% position is called the pth percentile.
[0050] In this embodiment, a frame difference method is used to obtain a difference image between two consecutive frames of images in a clear image sequence.
[0051] It should be noted that the two consecutive frames of images mentioned above are consecutive in a clear image sequence, and may be adjacent or non-adjacent in a grayscale image sequence.
[0052] In this embodiment, a plurality of images requiring attention are selected by comparing whether the grayscale mean of the difference image between two consecutive frames of images in the clear image sequence is greater than a first threshold.
[0053] The first threshold is the grayscale mean corresponding to the 20th percentile in the grayscale mean sequence; of course, as other implementations, it can also be set according to actual conditions.
[0054] Step S3, when two frames of images are adjacent in a grayscale image sequence, the first degree of the previous frame of image is the difference between the grayscale mean of the two adjacent frames of image and a first threshold value; when two frames of image are not adjacent in the grayscale image sequence, the second degree of the previous frame of image is the difference between the grayscale mean of the two non-adjacent frames of image and a second threshold value, the second threshold value is greater than the first threshold value, and the third degree of the blurred image between the two non-adjacent frames of image is the average of the target degree and the second degree; all blurred images between two non-adjacent frames of image and clear image sequence constitute a key image sequence.
[0055] In this embodiment, it is also determined whether the two consecutive frames of images whose grayscale mean values are greater than the first threshold selected in step S2 are adjacent to each other in the grayscale image sequence to determine the final key image sequence.
[0056] Specifically, when two frames of images are non-adjacent in a grayscale image sequence, there is a blurred image between the two non-adjacent frames of images. Since the grayscale mean of the difference image between the two non-adjacent frames of images is greater than the first threshold, the two non-adjacent frames of images are images that need attention. Then, the blurred image between the two non-adjacent frames of images may also be an image that needs attention. Therefore, the blurred images and clear image sequences between all non-adjacent frames of images constitute a key image sequence for extracting simulation data of the aircraft to be simulated.
[0057] Furthermore, before forming the key image sequence, the acquired blurred image is sharpened to obtain a processed clear image. Specifically, the sharpening process can be processed by a Gaussian filter. Of course, the above-mentioned sharpening process can also be sharpened by USM sharpening, high contrast sharpening, Lab sharpening, and using a plug-in. Since the above-mentioned sharpening process methods are all existing technologies, they will not be introduced in detail here.
[0058] Furthermore, in this embodiment, it is also necessary to calculate the importance of each image in the key image sequence. Specifically, when two frames of images are adjacent in the grayscale image sequence, the first degree of the previous frame of image is the difference change between the grayscale mean of the two adjacent frames of image and the first threshold; when the two frames of image are not adjacent in the grayscale image sequence, the second degree of the previous frame of image is the difference change between the grayscale mean of the two non-adjacent frames of image and the second threshold, and the third degree of the blurred image between the two non-adjacent frames of image is the average of the target degree and the second degree, wherein the target degree is the difference change between the grayscale mean of the two non-adjacent frames of image and the first threshold, wherein the first threshold is less than the second threshold.
[0059] The first threshold is the grayscale mean corresponding to the 20th percentile in the grayscale mean sequence after the grayscale means of each difference image are sorted from small to large, and the second threshold is the grayscale mean corresponding to the 40th percentile in the grayscale mean sequence. It should be noted that, since the grayscale mean of the difference image between the two frames is necessarily large when the two frames are not adjacent in the grayscale image sequence, it is necessary to appropriately increase the value of the second threshold so as to accurately screen out the images that need attention.
[0060] Of course, as other implementation methods, the first threshold in this embodiment can also be set to the grayscale mean corresponding to the 30th percentile, the grayscale mean corresponding to the 50th percentile, or the grayscale mean corresponding to the 60th percentile; similarly, the second threshold can also be set to the grayscale mean corresponding to the 50th percentile, the grayscale mean corresponding to the 70th percentile, or the grayscale mean corresponding to the 80th percentile; of course, the values of the first threshold and the second threshold can also be directly obtained based on experience.
[0061] Specifically, the first degree is: = ; is the first degree of the previous frame image in two adjacent frame images in the grayscale image sequence; is the grayscale mean of the difference image between two frames of images; is the first threshold.
[0062] The second degree is: ; is the second degree of the previous frame image in two non-adjacent frame images in the grayscale image sequence; is the grayscale mean of the difference image between two non-adjacent frames; is the second threshold.
[0063] The third degree of blur between two frames for: , is the second degree of the previous frame image in two non-adjacent frame images in the grayscale image sequence, is the grayscale mean of the difference image between two non-adjacent frames of images, is the first threshold.
[0064] When calculating the third degree of the blurred image, the importance of the blurred image is determined by using the grayscale mean of the difference image of two non-adjacent frames. This is because the grayscale value in the blurred image is difficult to determine. Therefore, the importance of the blurred image can be estimated by estimating the clear images before and after the blurred image, and the importance of the blurred image with the characteristics of the aircraft movement is quantified, that is, the third degree. It should be noted that if there are multiple blurred images between the two frames, the importance of the multiple blurred images is set to the same, that is, all are the third degree.
[0065] Step S4, input the key image sequence into the network model and output the simulation data of the current aircraft.
[0066] In this embodiment, after the key image sequence and the importance of each key image in the key image sequence are determined, the key image sequence is input into a pre-constructed network model to obtain simulation data.
[0067] The simulation data is the status data of the aircraft in each image, including pitch angle, roll angle, yaw angle, track angle, and flight path angle.
[0068] The network model in this embodiment can adopt a three-dimensional convolutional neural network, such as a 3D-CNN model, PointNet, 3DU-Net, etc.
[0069] Taking the 3D-CNN model as an example, the 3D-CNN model mainly includes an input layer, a hard-line layer (or H1 layer), a convolutional layer, a pooling layer, a fully connected layer, and an activation function.
[0070] Among them, before using the network model to obtain the simulation data of the aircraft currently to be simulated, the network model needs to be trained. The specific training process is as follows:
[0071] First, a training data set is obtained; the training data set includes a data set and a label, wherein the data set is a historical key image sequence of video data of various types of aircraft captured through the Internet; the label is an array of five rows and multiple columns of pitch angles, roll angles, yaw angles, track angles, and flight path angles corresponding to all images in the data set.
[0072] In this embodiment, before obtaining the historical key image sequence, it is also necessary to crop, scale, and grayscale the clear images in the video data of various types of aircraft, and calculate the historical importance of each historical key image according to steps S2-S3, where the historical importance is the first historical degree, the second historical degree, or the third historical degree, and the first historical degree, the second historical degree, and the third historical degree are calculated in the same way as the first degree, the second degree, and the third degree in step S3, respectively.
[0073] Secondly, the training data set is input into the network model for training. During the training process, the loss between the output prediction value and the label is calculated; the neural network model parameters are adjusted using the gradient descent method to minimize the prediction error;
[0074] Then, the parameters of the network model are iteratively adjusted until the loss is less than a certain value or the set number of training times is reached, and finally a trained network model is obtained.
[0075] Among them, during the training process of the network model, the loss function for: ; For the Mean square error loss of key images; is the regularization term, which is the sum of the squares of the differences between the gradient parameters corresponding to all key images; is the regularization coefficient, which is used to control the strength of regularization. is the total number of nodes in the network model, For the The criticality of a key image is the first degree, the second degree or the third degree. The above criticality is determined based on the key image corresponding to the node in each network model. For example, when the key image is a blurred image, the criticality at this time corresponds to the third degree of the blurred image.
[0076] Furthermore, before calculating the network model loss, it is also necessary to normalize the criticality of all key images (the criticality is the first degree, the second degree, or the third degree) to obtain the normalized criticality. The normalization method may be maximum and minimum value normalization.
[0077] In one embodiment, the mean square error loss for:
[0078] ;
[0079] In the formula, For the The predicted value of key images, G is the total number of key images; For the The labels of the key images.
[0080] In another embodiment, the mean square error loss for:
[0081] ;
[0082] In the formula, For the The comprehensive attention weight of the key images; For the The predicted value of key images, G is the total number of key images; For the The labels of the key images.
[0083] in, ; It is The time attention weight of the convolved feature map of the key image in the cth channel; It is The convolved feature map of the key image is the spatial attention weight of the cth channel, where C is the total number of channels.
[0084] The feature map mentioned above is the current key image sequence that passes through the convolution layer and outputs multiple three-dimensional feature maps.
[0085] The comprehensive attention weight mentioned above is obtained by combining the temporal attention weight and the spatial attention weight. The comprehensive attention weight can better capture the key information of the dynamic characteristics of the aircraft, improve the accuracy and efficiency of extracting the dynamic feature data of the aircraft in the key frame data under communication delay, and indirectly improve the accuracy of subsequent calculations in the distributed simulation engine.
[0086] It should be noted that the above network model can be independent of the distributed simulation model, or can be directly set in the distributed simulation model.
[0087] When simulating the attitude of the aircraft, the obtained simulation data and the corresponding weather, lighting, simulation time step and total time are input into the distributed simulation engine through the TCP protocol to complete the precise calculation of the dynamic model and output the change of the attitude angle and flight trajectory of the aircraft over time. Since the specific simulation process is an existing technology, it will not be introduced here.
[0088] In order to obtain simulation data for a distributed simulation model, the solution of the present invention is implemented in two steps, namely, the first step is to obtain a key image sequence by performing image processing on the acquired video data; the second step is to accurately extract the status data of the aircraft in the key image sequence by improving the loss function of the network model, and use the extracted status data of the aircraft as subsequent simulation data, thereby improving the accuracy of obtaining the simulation data.
[0089] The present invention also provides a distributed simulation model data acquisition system for aircraft attitude control. Figure 2 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the distributed simulation model data acquisition method for aircraft attitude control according to the present invention is implemented.
[0090] The system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and thus will not be described in detail here.
[0091] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.
[0092] In the description of this specification, "plurality" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.
[0093] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.
Claims
1. A distributed simulation model data acquisition method for aircraft attitude control, characterized in that: include: Acquire a grayscale image sequence of the target aircraft operation video data; filter out clear images from the grayscale image sequence to obtain a clear image sequence; Calculate the grayscale mean of the difference image of two consecutive frames of images in the clear image sequence, and select two frames of images when the grayscale mean is greater than the first threshold; When two frames of images are adjacent in a grayscale image sequence, the first degree of the previous frame of image is the difference between the grayscale mean of the two adjacent frames of image and the first threshold; when two frames of image are not adjacent in the grayscale image sequence, the second degree of the previous frame of image is the difference between the grayscale mean of the two non-adjacent frames of image and the second threshold, the second threshold is greater than the first threshold, the third degree of the blurred image between the two non-adjacent frames of image is the average of the target degree and the second degree, the target degree is the difference between the grayscale mean of the two non-adjacent frames of image and the first threshold; all blurred images between two non-adjacent frames of image and clear image sequence constitute a key image sequence; Input the key image sequence into the network model and output the simulation data of the current aircraft; the loss function of the network model for: ; For the Mean square error loss of key images; is the regularization term; is the regularization coefficient, is the total number of key images, For the A key image corresponds to a key level of the key image, and the key level is a first level, a second level, or a third level.
2. The distributed simulation model data acquisition method for aircraft attitude control according to claim 1, characterized in that: The first threshold is the grayscale mean corresponding to the 20th percentile in the grayscale mean sequence after the grayscale means of each difference image are sorted from small to large, and the second threshold is the grayscale mean corresponding to the 40th percentile in the grayscale mean sequence.
3. The distributed simulation model data acquisition method for aircraft attitude control according to claim 2 is characterized in that: The first degree is the ratio of the first difference to the grayscale mean, wherein the first difference is the difference between the grayscale mean of two adjacent frames of images and a first threshold; the second degree is the ratio of the second difference to the grayscale mean, wherein the second difference is the difference between the grayscale mean of two non-adjacent frames of images and a second threshold.
4. The distributed simulation model data acquisition method for aircraft attitude control according to claim 3 is characterized in that: Before obtaining the criticality, the method further includes a step of normalizing the obtained first degree, second degree and third degree.
5. The distributed simulation model data acquisition method for aircraft attitude control according to claim 1, characterized in that: The mean square error loss for: ; In the formula, For the The predicted value of key images, G is the total number of key images; For the The labels of the key images, For the The comprehensive attention weight of the key images, , It is The temporal attention weight of the c-th channel of the convolved feature map of the key image; It is The convolved feature map of the key image is the spatial attention weight of the cth channel, where C is the total number of channels.
6. The distributed simulation model data acquisition method for aircraft attitude control according to claim 1, characterized in that: The simulation data includes the pitch angle, roll angle, yaw angle, track angle and flight path angle of the aircraft.
7. The distributed simulation model data acquisition method for aircraft attitude control according to claim 1, characterized in that: The specific process of filtering out clear images from the grayscale image sequence and obtaining a clear image sequence is as follows: Calculate the clarity of each grayscale image in the grayscale image sequence; In response to the clarity being greater than the clarity threshold, the corresponding grayscale image is marked as a clear image, otherwise, the corresponding grayscale image is marked as a blurred image.
8. The distributed simulation model data acquisition method for aircraft attitude control according to claim 7, characterized in that: Before obtaining the grayscale image sequence, the method also includes sequentially performing grayscale processing on each frame of the image in the target aircraft operation video data.
9. A distributed simulation model data acquisition system for aircraft attitude control, characterized in that: include: processor; A memory storing computer instructions for acquiring distributed simulation model data for aircraft attitude control. When the computer instructions are executed by the processor, the system executes the method for acquiring distributed simulation model data for aircraft attitude control according to any one of claims 1 to 8.
