A training quality monitoring system for civil aviation based on a digital twin model

Abstract

The application relates to the technical field of civil aviation training, in particular to a training quality monitoring system for civil aviation based on a digital twin model, which comprises a data acquisition unit, a state analysis unit used for determining a training state according to a posture change value and a trajectory discreteness of a training task, a selection analysis unit used for determining a to-be-detected point selection mode according to the training state, an image division unit used for determining a sub-image division mode according to a mutation pixel point distribution coefficient, a quality confirmation unit used for determining an image quality value confirmation mode according to an analysis condition, and a processing analysis unit used for determining a processing mode according to the image quality value. The application can guarantee the image quality of a flight training simulation scene, thereby guaranteeing the training quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of civil aviation training, and particularly relates to a civil aviation training quality monitoring system based on a digital twin model. BACKGROUND

[0002] Traditional civil aviation training is mainly through actual operation and written materials, which has many limitations, such as high cost, high safety risk, etc. However, when using a digital twin model to simulate an actual scene for training of flight personnel, the image quality of the simulated scene cannot be guaranteed, which makes it difficult to meet user requirements for training results. Therefore, how to improve the image quality of the simulated scene simulated by the digital twin model and thus ensure the training quality is a technical problem that needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN106530894B discloses a flight training device virtual flat display method and system, which includes: constructing a virtual scene according to pre-acquired map data; processing the image of the virtual scene through a digital image processing method to obtain saliency information of the scene image; constructing an augmented reality model through a pattern recognition algorithm in combination with the saliency information, and superimposing the augmented reality model into the virtual scene; and embedding a flat display picture in the virtual scene into an existing visual system to display in a video scene. It can be seen that the above technical solution has the following technical problems: it cannot guarantee the image quality of the flight training simulation scene, it cannot detect the image quality of the virtual scene according to different ways according to the actual scene, and it cannot guarantee the training quality. SUMMARY

[0004] Therefore, the present application provides a civil aviation training quality monitoring system based on a digital twin model to overcome the problems in the prior art that the image quality of the flight training simulation scene cannot be guaranteed, the image quality of the virtual scene cannot be detected according to different ways according to the actual scene, and the training quality cannot be guaranteed.

[0005] To achieve the above-mentioned purpose, the present application provides a civil aviation training quality monitoring system based on a digital twin model, which comprises:

[0006] A data acquisition unit is configured to acquire training information.

[0007] A state analysis unit is connected to the data acquisition unit and configured to determine a training state according to a posture change value and a trajectory dispersion degree of a training task.

[0008] An analysis unit is selected, which is connected to the data acquisition unit and the state analysis unit respectively, and is used to determine the selection method of the test point according to the training state. The selection method of the test point is to determine the selection optimization method or uniformly select the test point according to the set conditions. The selection optimization method is to select the test point according to the combined paragraph or the fuel consumption slope.

[0009] An image segmentation unit, connected to the selection and analysis unit, is used to determine the sub-image segmentation method based on the distribution coefficient of abrupt pixel points. The sub-image segmentation method is either uniform segmentation according to a set angle or segmentation based on pixel combinations.

[0010] A quality confirmation unit, which is connected to the selection and analysis unit and the image segmentation unit respectively, is used to determine the image quality value confirmation method according to the analysis conditions. The image quality value confirmation method is to determine the region based on the scene feature value corresponding to the sub-image or to determine the whole based on the Gaussian feature value.

[0011] The processing and analysis unit, which is connected to the quality confirmation unit, is used to determine the processing method based on the image quality value. The processing method is to send a quality qualified reminder to the user or to determine the quality adjustment method based on the trajectory deviation value. The quality adjustment method is to adjust the data transmission delay or compression reference value.

[0012] Furthermore, the state analysis unit determines the training state based on the pose change values ​​and trajectory dispersion of the training task. The training state includes:

[0013] The first training state is when the attitude change value is greater than or equal to the preset attitude change value or the trajectory dispersion is less than the preset trajectory dispersion.

[0014] The second training state is characterized by an attitude change value less than a preset attitude change value and a trajectory dispersion greater than or equal to a preset trajectory dispersion.

[0015] Furthermore, the selection analysis unit responds to the training state to determine the method for selecting the detection point;

[0016] The first training state of the analysis unit response is selected, and the selection method of the test point is determined by the set conditions to optimize the selection method.

[0017] The second training state of the analysis unit response is selected, and the detection points are selected uniformly.

[0018] When selecting test points uniformly, the training time is divided into several equal parts, and the training time points corresponding to each division are used as test points.

[0019] Furthermore, the selection analysis unit responds to the set conditions to determine the selection optimization method;

[0020] The selection criteria for the analysis unit response are: the node shear float value is greater than or equal to the preset node shear float value and the node distribution coefficient is greater than or equal to the preset node distribution coefficient; the selection optimization method is to select the detection point based on the combined paragraph.

[0021] The selected conditions for the analysis unit response are that the node shear fluctuation value is less than the preset node shear fluctuation value or the node distribution coefficient is less than the preset node distribution coefficient, and the selected optimization method is to select the detection point based on the fuel consumption slope.

[0022] When selecting the optimization method based on the fuel consumption slope to select the test points, a fuel consumption change curve with training time as the horizontal axis and fuel consumption as the vertical axis is obtained. The curve is divided into several curve segments with the change points as the dividing points. The fuel consumption slope corresponding to each curve segment is detected. The curve segment with a fuel consumption slope less than the preset fuel consumption slope is recorded as a feature segment. A preset number of training time points in each feature segment are selected as test points.

[0023] Furthermore, the selection and analysis unit selects the points to be detected based on the combined paragraphs, including:

[0024] Similarity analysis is performed on each trajectory point in the predicted trajectory. When performing similarity analysis on a trajectory point in the predicted trajectory image, the trajectory point is recorded as the target trajectory point, and the trajectory points that do not include the target trajectory point and are not included in the similarity combination are recorded as reference trajectory points.

[0025] The set of all reference trajectory points preceding the first reference trajectory point that is located after the target trajectory point in the moving sequence and whose shear difference value with the target trajectory point is greater than the preset shear difference value, together with the target trajectory point, is denoted as a similar combination;

[0026] Perform similarity analysis on trajectory points that are not recorded as similar combinations until all reference trajectory points are recorded as similar combinations.

[0027] Each similar combination whose combined reference value is less than the preset combined reference value is recorded as a combined segment, and the number of points to be detected corresponding to the combined segment is determined according to the combined reference value.

[0028] The number of test points corresponding to a single combined paragraph is positively correlated with the combined reference value.

[0029] Furthermore, the image segmentation unit determines the sub-image segmentation method based on the distribution coefficient of abrupt pixel points;

[0030] If the distribution coefficient of the mutation pixel is greater than or equal to the preset mutation similarity distribution coefficient, the sub-image is divided into uniform segments according to the set angle.

[0031] If the distribution coefficient of the mutation pixels is less than the preset mutation similarity distribution coefficient, the sub-image segmentation method is to segment based on pixel combinations;

[0032] The set angle is negatively correlated with the distribution coefficient of abrupt pixel points.

[0033] Furthermore, the image segmentation unit performs segmentation based on pixel combinations, wherein,

[0034] The virtual image is evenly divided into several viewpoint images. Combined detection is performed on each viewpoint image. When combined detection is performed on a single viewpoint image, the viewpoint image is recorded as the target viewpoint image. Viewpoint images that do not include the target viewpoint image and are not included in the pixel combination are recorded as the reference viewpoint image.

[0035] The number of mutation pixels corresponding to the target view image is accumulated one by one with the number of mutation pixels corresponding to each reference view image located after the target view image in the clockwise direction until the total number of mutation pixels is greater than the preset total number of mutation pixels. The reference view images before the reference view image corresponding to the last accumulated number of mutation pixels and the target view image are recorded as a pixel combination.

[0036] Perform combination detection on viewpoint images that have not recorded pixel combinations until pixel combinations are recorded for all viewpoint images;

[0037] The pixels are divided using the starting and ending edges corresponding to each pixel combination as dividing lines.

[0038] Furthermore, the quality verification unit responds to the analysis conditions to determine the image quality value verification method;

[0039] The analysis conditions for the quality confirmation unit response are that the comprehensive change value is greater than or equal to the preset comprehensive change value or the proportion of abruptly changed pixels is greater than or equal to the preset proportion of abruptly changed pixels. The image quality value confirmation method is to determine the region based on the scene feature value corresponding to the sub-image.

[0040] When the image quality value is determined by region determination, the virtual image is segmented according to the sub-image division method to obtain several sub-images. The scene feature values ​​corresponding to each sub-image are input into the regression model to determine the image quality value.

[0041] The analysis conditions for the quality confirmation unit response are that the comprehensive change value is less than the preset comprehensive change value and the proportion of abruptly changed pixels is less than the preset proportion of abruptly changed pixels. The image quality value confirmation method is to make an overall judgment based on Gaussian feature values.

[0042] When the image quality value is determined as a whole, the Gaussian feature value corresponding to the virtual image is input into the regression model to determine the image quality value.

[0043] Furthermore, the processing and analysis unit determines the processing method based on the image quality value;

[0044] If the image quality value is greater than or equal to the preset image quality value, the processing method is to send a quality qualified reminder to the user.

[0045] If the image quality value is less than the preset image quality value, the processing method is to determine the quality adjustment method based on the trajectory deviation value.

[0046] Furthermore, the processing and analysis unit determines the quality adjustment method based on the trajectory deviation value;

[0047] If the trajectory deviation value is greater than or equal to the preset trajectory deviation value, the quality adjustment method is to reduce the data transmission delay.

[0048] If the trajectory deviation value is less than the preset trajectory deviation value, the quality adjustment method is to increase the compression reference value.

[0049] The decrease in data transmission delay is positively correlated with the positional deviation, while the increase in the compression reference value is negatively correlated with the image similarity.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows: the technical solution of the present invention determines the training state based on the attitude change value and trajectory dispersion, effectively reflects the training difficulty of flight training through the attitude change value and trajectory dispersion, and then adaptively selects different detection point selection methods according to the training state. This makes the selection of detection point selection methods more in line with the actual application scenario, avoiding the problem in the prior art that detection is only based on the student's training process and cannot select different methods to detect the image quality of the virtual scene according to the actual scenario. It also avoids the problem of low image quality detection efficiency and insufficient accuracy caused by the inability to select suitable detection points when the training difficulty is high. It can ensure the accuracy and effectiveness of image quality detection results, thereby ensuring the training quality.

[0051] Furthermore, this invention effectively reflects the changes in the estimated movement trajectory by setting conditions, and then adaptively selects different selection optimization methods according to the set conditions. This makes the selection of optimization methods more in line with the actual application scenario, avoiding the problem of poor detection efficiency caused by the inability to select a suitable detection point when the changes in the estimated trajectory are highly complex. This improves the efficiency and accuracy of image quality detection.

[0052] Furthermore, this invention effectively reflects the distribution of features in a virtual image through the abrupt pixel distribution coefficient. Based on this coefficient, different sub-image partitioning methods are adaptively selected, making the choice of sub-image partitioning methods more consistent with actual application scenarios. This avoids the problem of poor image quality assessment accuracy due to poor image quality value confirmation accuracy, thereby ensuring the quality of flight training.

[0053] Furthermore, this invention effectively reflects the flight changes corresponding to the detection point and the characteristics of the simulated image corresponding to the detection point by analyzing the conditions. Then, different image quality value confirmation methods are adaptively selected according to the analysis conditions, making the selection of image quality value confirmation methods more in line with actual application scenarios. This avoids the problem of poor accuracy in image quality evaluation caused by a single image quality value confirmation method, and ensures that the accuracy of image quality value judgment can be improved when the flight changes are complex or the simulated image has many features, thereby ensuring the quality of flight training.

[0054] Furthermore, in this invention, the image quality value effectively reflects the image quality of the simulated scene, thereby adaptively selecting different processing methods. This makes the selection of processing methods more in line with the actual application scenario, avoiding the problem of poor training quality when the image quality of the simulated scene is poor. In turn, the image quality of the simulated scene is improved according to the quality adjustment method, thereby ensuring the quality of flight training.

[0055] Furthermore, this invention effectively reflects the degree of deviation between the flight trajectory and the predicted trajectory during the learning and training process through the trajectory deviation value. Different quality adjustment methods are then adaptively selected based on the trajectory deviation value. When the trajectory deviation value is large, the data transmission delay is reduced to ensure the real-time performance and accuracy of the flight trajectory data, thereby improving image quality. When the trajectory deviation value is small, the compression reference value is increased to help retain more detailed information, thus maintaining the accuracy of the digital twin model and further improving image quality. Attached Figure Description

[0056] Fig. 1 This is a unit connection diagram of the civil aviation training quality monitoring system based on a digital twin model according to the present invention.

[0057] Fig. 2 This is a flowchart illustrating the method for determining the selection of detection points based on the training state in this invention.

[0058] Fig. 3 This is a flowchart illustrating how the present invention determines the selection of an optimization method based on set conditions;

[0059] Fig. 4 This is a flowchart illustrating the method for determining image quality values ​​based on analysis conditions according to the present invention. Detailed Implementation

[0060] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0061] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0062] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0063] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0064] Please see Figs. 1 to 4 As shown, this invention provides a civil aviation training quality monitoring system based on a digital twin model, comprising:

[0065] The data acquisition unit is used to collect training information;

[0066] A state analysis unit, which is connected to the data acquisition unit, is used to determine the training state based on the attitude change value and trajectory dispersion of the training task.

[0067] An analysis unit is selected, which is connected to the data acquisition unit and the state analysis unit respectively, and is used to determine the selection method of the test point according to the training state. The selection method of the test point is to determine the selection optimization method or uniformly select the test point according to the set conditions. The selection optimization method is to select the test point according to the combined paragraph or the fuel consumption slope.

[0068] An image segmentation unit, connected to the selection and analysis unit, is used to determine the sub-image segmentation method based on the distribution coefficient of abrupt pixel points. The sub-image segmentation method is either uniform segmentation according to a set angle or segmentation based on pixel combinations.

[0069] A quality confirmation unit, which is connected to the selection and analysis unit and the image segmentation unit respectively, is used to determine the image quality value confirmation method according to the analysis conditions. The image quality value confirmation method is to determine the region based on the scene feature value corresponding to the sub-image or to determine the whole based on the Gaussian feature value.

[0070] The processing and analysis unit, which is connected to the quality confirmation unit, is used to determine the processing method based on the image quality value. The processing method is to send a quality qualified reminder to the user or to determine the quality adjustment method based on the trajectory deviation value. The quality adjustment method is to adjust the data transmission delay or compression reference value.

[0071] The application scenario of this invention is image quality detection in simulated scenarios during civil aviation training using a digital twin model. The simulated scenario is obtained by inputting flight training data into the digital twin model. The simulated scenario includes simulated images corresponding to several training time points, where the training time points are set by the user. A method for setting training time points is provided, recording each training time point in 1-second intervals from earliest to latest, i.e., generating a simulated image every 1 second. The training information includes several flight training data related to the training tasks, including but not limited to flight training under complex weather conditions, night flight training, navigation training, and mountain flight training. The flight training data includes, but is not limited to, aircraft position, speed, acceleration, direction of movement, light intensity, air pressure, and scene images. The scene images are captured images of the aircraft and the environment. The flight training data is input into the digital twin model to obtain the estimated movement trajectory. The estimated movement trajectory corresponds to the predicted aircraft position for each training time point. This is easily understood by those skilled in the art and will not be elaborated further.

[0072] This invention includes several historical records. Each historical record contains at least one civil aviation training session, including attitude change values, trajectory dispersion, node shear fluctuation values, node distribution coefficients, shear difference values, and combined reference values. Each historical record also has a corresponding pass / fail marker, which indicates whether the civil aviation training effect meets the user's needs. The pass / fail marker can be recorded manually.

[0073] Specifically, the state analysis unit determines the training state based on the pose change values ​​and trajectory dispersion of the training task. The training state includes:

[0074] The first training state is when the attitude change value is greater than or equal to the preset attitude change value or the trajectory dispersion is less than the preset trajectory dispersion.

[0075] The second training state is characterized by an attitude change value less than a preset attitude change value and a trajectory dispersion greater than or equal to a preset trajectory dispersion.

[0076] Specifically, the standard deviations of the rotational angular rates on the pitch, roll, and yaw axes of the aircraft are measured respectively, and the maximum value among the standard deviations is recorded as the attitude change value. The rotational angular rates corresponding to the pitch, roll, and yaw axes are measured by gyroscopes, which is easy for those in the field of computing to understand and will not be elaborated on in detail.

[0077] The trajectory dispersion is confirmed by detecting the predicted aircraft position at each training time point and recording the standard deviation of the shortest distance from each predicted aircraft position to the reference baseline as the trajectory dispersion. The reference baseline is the line connecting the predicted aircraft position at the start training time point and the predicted aircraft position at the end training time point. The start training time point and the end training time point are the first and last training time points in the training time in the order from early to late, respectively.

[0078] The user can determine the preset attitude change value and preset trajectory dispersion value according to the actual application scenario. The higher the user's demand for image quality detection effect, the smaller the preset attitude change value and preset trajectory dispersion value should be. The system provides a preset attitude change value and preset trajectory dispersion value. The system detects the historical records of the points to be detected evenly selected, and records the average of the attitude change values ​​that meet the user's requirements as the preset attitude change value, and records the average of the trajectory dispersion values ​​that meet the user's requirements as the preset trajectory dispersion value.

[0079] Specifically, the selection analysis unit responds to the training state to determine the method for selecting the detection point;

[0080] The first training state of the analysis unit response is selected, and the selection method of the test point is determined by the set conditions to optimize the selection method.

[0081] The second training state of the analysis unit response is selected, and the detection points are selected uniformly.

[0082] When selecting test points uniformly, the training time is divided into several equal parts, and the training time points corresponding to each division are used as test points.

[0083] When dividing the training time into several equal parts, one method of division is provided, in which the training time is divided into φ equal parts, and the number of division points obtained is (φ-1). The value of φ can be determined by the user according to the actual needs. The greater the detection accuracy of the image quality, the larger the value of φ. One value of φ is provided, which is 30.

[0084] Specifically, the selection analysis unit response setting conditions are used to determine the selection optimization method;

[0085] The selection criteria for the analysis unit response are: the node shear float value is greater than or equal to the preset node shear float value and the node distribution coefficient is greater than or equal to the preset node distribution coefficient; the selection optimization method is to select the detection point based on the combined paragraph.

[0086] The selected conditions for the analysis unit response are that the node shear fluctuation value is less than the preset node shear fluctuation value or the node distribution coefficient is less than the preset node distribution coefficient, and the selected optimization method is to select the detection point based on the fuel consumption slope.

[0087] When selecting the optimization method based on the fuel consumption slope to select the test points, a fuel consumption change curve with training time as the horizontal axis and fuel consumption as the vertical axis is obtained. The curve is divided into several curve segments with the change points as the dividing points. The fuel consumption slope corresponding to each curve segment is detected. The curve segments with a fuel consumption slope less than the preset fuel consumption slope are recorded as feature segments. A preset number of training time points in each feature segment are selected as test points.

[0088] The setting conditions include a first setting condition and a second setting condition. The first setting condition is that the node shear float value is greater than or equal to the preset node shear float value and the node distribution coefficient is greater than or equal to the preset node distribution coefficient. The second setting condition is that the node shear float value is less than the preset node shear float value or the node distribution coefficient is less than the preset node distribution coefficient.

[0089] After obtaining the estimated movement trajectory, a three-dimensional Cartesian coordinate system is established with the predicted aircraft position corresponding to the initial training time point as the origin. The predicted aircraft positions in the estimated movement trajectories corresponding to each training time point (excluding the initial training time point) are recorded as trajectory points. The vector angle corresponding to each trajectory point is detected. The method for confirming the vector angle corresponding to a single trajectory point is as follows: for a trajectory point, this trajectory point is recorded as the target trajectory point. The vector angle corresponding to the trajectory points adjacent to the target trajectory point and located before the target trajectory point in the movement sequence is determined. Vector corresponding to the target trajectory point The angle ε between them is denoted as the vector angle. The vector corresponding to each trajectory point is tangent to the movement trajectory, and the direction of the vector is the same as the movement direction. Trajectory points with vector angles greater than a preset vector angle are recorded as trajectory nodes. The node shear float value is the standard deviation of the vector angle corresponding to each trajectory node. The value of the preset vector angle can be determined by the user according to the actual application scenario. One preset vector angle value is provided, which is 60°. The movement order is the order in which the trajectory points are reached.

[0090] The node distribution coefficient is the average of the reference minimum distances corresponding to each trajectory node. For a single trajectory node, the trajectory node is recorded as the target trajectory node, and other trajectory nodes outside the target trajectory node are recorded as reference trajectory nodes. The shortest distance from the target trajectory node to each reference trajectory node is detected, and the minimum value among the shortest distances is recorded as the reference minimum distance.

[0091] The user can determine the values ​​of the preset node shear fluctuation value and the preset node distribution coefficient according to the actual application scenario. The larger the values ​​of the preset node shear fluctuation value and the preset node distribution coefficient, the higher the user's demand for selecting the detection point based on the fuel consumption slope. The system provides a set of preset node shear fluctuation values ​​and preset node distribution coefficient values, detects the historical records of selecting the detection point based on the fuel consumption slope, and records the average value of the node shear fluctuation value corresponding to the historical records that meet the user's needs as the preset node shear fluctuation value, and records the average value of the node distribution coefficient corresponding to the historical records that meet the user's needs as the preset node distribution coefficient.

[0092] The process of designating curve segments with fuel consumption slope deviations less than preset fuel consumption slope deviations as characteristic segments includes: performing feature analysis on each curve segment; when performing feature analysis on a curve segment, designating that curve segment as the target curve segment; and combining other curve segments other than the target curve segment as reference curve segments; designating the reference curve segments with fuel consumption slope deviations less than preset fuel consumption slope deviations from the target curve segment, along with the target curve segment, as a characteristic segment; and continuing to perform feature analysis on curve segments not yet designated as characteristic segments until all curve segments are designated as characteristic segments.

[0093] Fuel consumption is the amount of fuel consumed by the aircraft at each training time point. For a curve segment, the fuel consumption slope corresponding to the curve segment = (fuel consumption at the end point of the segment - fuel consumption at the beginning point of the segment) / (training time at the end point of the segment - training time at the beginning point of the segment). The beginning point and the end point of the segment are the first and last training time points of the curve segment, respectively.

[0094] For any two curve segments, the larger of the corresponding fuel consumption slopes is denoted as k1, and the smaller as k2. The fuel consumption slope deviation is calculated as |k1-k2| / k1. Users can determine the preset values ​​for the fuel consumption slope deviation and the preset quantity based on their actual application scenarios. The greater the user's need for similarity among the trajectory points within the combined segment, the smaller the preset fuel consumption slope deviation and the larger the preset quantity. One preset value for the fuel consumption slope deviation and the preset quantity is provided: the fuel consumption slope deviation is 20%, and the preset quantity is 10% of the number of training time points contained in a single feature segment.

[0095] Specifically, the selection and analysis unit selects the points to be detected based on the combined paragraphs, including:

[0096] Similarity analysis is performed on each trajectory point in the predicted trajectory. When performing similarity analysis on a trajectory point in the predicted trajectory image, the trajectory point is recorded as the target trajectory point, and the trajectory points that do not include the target trajectory point and are not included in the similarity combination are recorded as reference trajectory points.

[0097] The set of all reference trajectory points preceding the first reference trajectory point that is located after the target trajectory point in the moving sequence and whose shear difference value with the target trajectory point is greater than the preset shear difference value, together with the target trajectory point, is denoted as a similar combination;

[0098] Perform similarity analysis on trajectory points that are not recorded as similar combinations until all reference trajectory points are recorded as similar combinations.

[0099] Each similar combination whose combined reference value is less than the preset combined reference value is recorded as a combined segment, and the number of points to be detected corresponding to the combined segment is determined according to the combined reference value.

[0100] The number of test points corresponding to a single combined paragraph is positively correlated with the combined reference value.

[0101] Specifically, each similar combination whose combined reference value is less than the preset combined reference value is recorded as a combined paragraph. This includes: performing combined analysis on each similar combination; when performing combined analysis on a similar combination, recording the similar combination as the target similar combination; recording other similar combinations besides the target similar combination as reference similar combinations; recording the reference similar combination whose combined reference value with the target similar combination is less than the preset combined reference value and the target similar combination as a combined paragraph, and continuing to perform combined analysis on similar combinations not recorded as combined paragraphs until each similar combination is recorded as a combined paragraph.

[0102] The shear difference value is the absolute value of the difference between the vector angles corresponding to two trajectory points. The combined reference value is the absolute value of the difference between the combined mean values ​​corresponding to two combined segments. The combined mean value is the average value of the vector angles of each trajectory point within a combined segment.

[0103] Users can determine the preset shear difference value and preset combination reference value according to the actual application scenario. The greater the user's need for similarity among trajectory points within the combined segment, the smaller the preset shear difference value and preset combination reference value will be. A preset shear difference value and preset combination reference value are provided. The historical records of the points to be detected are selected based on the combined segment. The average value of the shear difference value corresponding to the historical records that meet the user's needs is recorded as the preset shear difference value, and the average value of the combination reference value corresponding to the historical records that meet the user's needs is recorded as the preset combination reference value.

[0104] After determining the number of test points corresponding to a combined paragraph based on the combined reference value, if the number of test points corresponding to a combined paragraph is 'a', then randomly select 'a' trajectory points in the combined paragraph and use the training time points corresponding to each trajectory point as test points.

[0105] Specifically, the image segmentation unit determines the sub-image segmentation method based on the distribution coefficient of abrupt pixel points;

[0106] If the distribution coefficient of the mutation pixel is greater than or equal to the preset mutation similarity distribution coefficient, the sub-image is divided into uniform segments according to the set angle.

[0107] If the distribution coefficient of the mutation pixels is less than the preset mutation similarity distribution coefficient, the sub-image segmentation method is to segment based on pixel combinations;

[0108] The set angle is negatively correlated with the distribution coefficient of abrupt pixel points.

[0109] Each detection point corresponds to a virtual image. In this invention, the virtual image is a panoramic image. The panoramic image is mapped into a spherical pattern in the actual viewing scene through spherical projection technology. The sphere is centered on the viewer's position. This is something that is easy for those skilled in the art to understand, and will not be elaborated on in detail.

[0110] The angle is set to the angle of the sector projected onto the equatorial plane of a single sub-image. The equatorial plane is a circle with the equator as its outline and the center of the sphere as its center.

[0111] When dividing the sub-image into uniform segments according to the set angle, the equator of the sphere is divided into equal parts according to the number of division points. The line connecting the spheres at each division point is used as the division line to divide the virtual image. The number of division points = 360° / set angle.

[0112] For a single pixel, if the difference between the minimum pixel value of this pixel and the minimum pixel value of its adjacent pixels is greater than a preset difference, then this pixel is a mutated pixel. The user can set the preset difference value according to the actual situation. The higher the user's requirement for accuracy in confirming the proportion of mutated pixels, the smaller the preset difference value will be. One preset difference value is provided, which is the minimum difference between the minimum pixel value of each mutated pixel and the minimum pixel value of its adjacent pixels in the historical records.

[0113] The mutation pixel distribution coefficient is the average of the reference distances corresponding to each mutation pixel. For a single mutation pixel, this mutation pixel is recorded as the target mutation pixel, and the other mutation pixels besides the target mutation pixel are recorded as reference mutation pixels. The shortest distance from the target mutation pixel to each reference mutation pixel is detected, and the average of the shortest distances is recorded as the reference distance. The value of the preset mutation similarity distribution coefficient can be determined by the user according to the actual application scenario. The greater the user's requirement for the accuracy of image quality values, the smaller the value of the preset mutation similarity distribution coefficient should be. A preset mutation similarity distribution coefficient value is provided. Historical records that are uniformly segmented according to a set angle are detected, and the average of the mutation similarity distribution coefficients corresponding to the historical records that meet the user's requirements is recorded as the preset mutation similarity distribution coefficient.

[0114] Specifically, the image segmentation unit performs segmentation based on pixel combinations, wherein,

[0115] The virtual image is evenly divided into several viewpoint images. Combined detection is performed on each viewpoint image. When combined detection is performed on a single viewpoint image, the viewpoint image is recorded as the target viewpoint image. Viewpoint images that do not include the target viewpoint image and are not included in the pixel combination are recorded as the reference viewpoint image.

[0116] The number of mutation pixels corresponding to the target view image is accumulated one by one with the number of mutation pixels corresponding to each reference view image located after the target view image in the clockwise direction until the total number of mutation pixels is greater than the preset total number of mutation pixels. The reference view images before the reference view image corresponding to the last accumulated number of mutation pixels and the target view image are recorded as a pixel combination.

[0117] Perform combination detection on viewpoint images that have not recorded pixel combinations until pixel combinations are recorded for all viewpoint images;

[0118] The pixels are divided using the starting and ending edges corresponding to each pixel combination as dividing lines.

[0119] The process of uniformly dividing the virtual image into several viewpoint images includes: dividing the equator into n equal parts along the sphere's equator, and using the line connecting the points of division as the dividing line to divide the virtual image. The value of n can be determined by the user according to actual needs. One possible value of n is 360.

[0120] The number of mutated pixels is the total number of mutated pixels in a single viewpoint image; the total number of mutated pixels is the sum of the number of mutated pixels corresponding to each viewpoint image.

[0121] The user can determine the value of the preset total number of mutation pixels according to the actual application scenario. The greater the user's requirement for the precision of the image quality value, the smaller the value of the preset total number of mutation pixels. A preset value of the total number of mutation pixels is provided. The historical records of segmentation based on pixel combinations are detected, and the average value of the total number of mutation pixels corresponding to the historical records that meet the user's requirements is recorded as the preset total number of mutation pixels.

[0122] For a single pixel combination, the first dividing line of the first viewpoint image in the clockwise direction is the starting edge corresponding to the pixel combination, and the second dividing line of the last viewpoint image in the clockwise direction is the ending edge corresponding to the pixel combination.

[0123] Specifically, the quality verification unit responds to analysis conditions to determine the image quality value verification method;

[0124] The analysis conditions for the quality confirmation unit response are that the comprehensive change value is greater than or equal to the preset comprehensive change value or the proportion of abruptly changed pixels is greater than or equal to the preset proportion of abruptly changed pixels. The image quality value confirmation method is to determine the region based on the scene feature value corresponding to the sub-image.

[0125] When the image quality value is determined by region determination, the virtual image is segmented according to the sub-image division method to obtain several sub-images. The scene feature values ​​corresponding to each sub-image are input into the regression model to determine the image quality value.

[0126] The analysis conditions for the quality confirmation unit response are that the comprehensive change value is less than the preset comprehensive change value and the proportion of abruptly changed pixels is less than the preset proportion of abruptly changed pixels. The image quality value confirmation method is to make an overall judgment based on Gaussian feature values.

[0127] When the image quality value is determined as a whole, the Gaussian feature value corresponding to the virtual image is input into the regression model to determine the image quality value.

[0128] The analysis conditions include a first analysis condition and a second analysis condition. The first analysis condition is that the comprehensive change value is greater than or equal to the preset comprehensive change value or the proportion of mutated pixels is greater than or equal to the preset proportion of mutated pixels. The second analysis condition is that the comprehensive change value is less than the preset comprehensive change value and the proportion of mutated pixels is less than the preset proportion of mutated pixels.

[0129] Each detection point corresponds to a virtual image. The method for confirming the comprehensive change value and the proportion of abruptly changed pixels is as follows: For a single detection point, the detection point is recorded as the target detection point. The comprehensive change value corresponding to the target detection point = the fuel consumption slope corresponding to the target detection point + the vector angle corresponding to the target detection point. The virtual image corresponding to the target detection point is recorded as the target image. The proportion of abruptly changed pixels = the number of abruptly changed pixels in the target image / the total number of pixels in the target image.

[0130] Users can determine the preset comprehensive change value and preset mutation pixel percentage based on the actual application scenario. The greater the user's requirement for the accuracy of the image quality value, the smaller the preset comprehensive change value and preset mutation pixel percentage should be. A preset comprehensive change value and preset mutation pixel percentage are provided. The image quality value is confirmed by the overall judgment historical records. The average of the comprehensive change values ​​corresponding to the historical records that meet the user's needs is recorded as the preset comprehensive change value, and the average of the mutation pixel percentages corresponding to the historical records that meet the user's needs is recorded as the preset mutation pixel percentage.

[0131] The method for confirming scene feature values ​​is as follows: For a single sub-image, the sub-image is separated into three independent color channels: R, G, and B. The mean value of the first pixel corresponding to each channel is calculated. The mean value of the first pixel corresponding to a single channel is the average value of the pixel values ​​of each pixel in the single channel. For a single channel, the pixel with the value of the first pixel is recorded as the center pixel, and the pixels adjacent to the center pixel are recorded as neighboring pixels. If the pixel value corresponding to the neighboring pixel is greater than or equal to the mean value of the first pixel, it is recorded as 1. If the pixel value corresponding to the neighboring pixel is less than the mean value of the first pixel, it is recorded as 0. The binary number obtained by concatenating the comparison results of each channel in RGB order is recorded as the scene feature value.

[0132] The method for confirming Gaussian eigenvalues ​​is as follows: Gaussian smoothing is performed on the virtual image, and then the first derivative images of the smoothed image in the x and y directions are obtained through a difference filter. This is a content that is easy for those skilled in the art to understand, and will not be elaborated in detail. For a single first derivative image, the pixel point with a pixel value of the second pixel mean is recorded as the center pixel point, and the pixels adjacent to the center pixel point are recorded as neighboring points. If the pixel value corresponding to the neighboring point is greater than or equal to the second pixel mean, it is recorded as 1; if the pixel value corresponding to the neighboring point is less than the second pixel mean, it is recorded as 0. The binary number obtained by concatenating the comparison results of each first derivative image is recorded as the Gaussian eigenvalue.

[0133] The scene feature values ​​or Gaussian feature values ​​are input into the regression model for training and prediction to obtain the image quality value. This is something that is easy for those skilled in the art to understand, and will not be elaborated on in detail.

[0134] Specifically, the processing and analysis unit determines the processing method based on the image quality value;

[0135] If the image quality value is greater than or equal to the preset image quality value, the processing method is to send a quality qualified reminder to the user.

[0136] If the image quality value is less than the preset image quality value, the processing method is to determine the quality adjustment method based on the trajectory deviation value.

[0137] The user can determine the preset image quality value based on the actual application scenario. The higher the user's demand for training quality, the larger the preset image quality value will be. The preset image quality value is set by taking the average value of the historical image quality values ​​that can meet the user's needs.

[0138] Specifically, the processing and analysis unit determines the quality adjustment method based on the trajectory deviation value;

[0139] If the trajectory deviation value is greater than or equal to the preset trajectory deviation value, the quality adjustment method is to reduce the data transmission delay.

[0140] If the trajectory deviation value is less than the preset trajectory deviation value, the quality adjustment method is to increase the compression reference value.

[0141] The decrease in data transmission delay is positively correlated with the positional deviation, while the increase in the compression reference value is negatively correlated with the image similarity.

[0142] The process involves recording the aircraft's movement route during training as the actual movement route, detecting the positional deviation at each training time point, and recording the average of these positional deviations as the trajectory deviation value. The positional deviation is the shortest distance between the aircraft's position on the actual movement route and its position on the estimated movement route at a given training time point. The preset trajectory deviation value can be determined by the user based on the actual application scenario. The smaller the preset trajectory deviation value, the greater the user's need to reduce data transmission latency. A preset trajectory deviation value is provided, and historical records of adjustments to the compression reference value are detected. The average of the trajectory deviation values ​​corresponding to the historical records that meet the user's needs is recorded as the preset trajectory deviation value.

[0143] Data transmission latency is the time required to transmit flight training data to the digital twin model. It is understandable that the flight training data needs to be compressed after acquisition to improve data transmission efficiency. However, since the flight data contains scene images, the resolution of the scene images is reduced after compression, resulting in a decrease in data quality. Consequently, the image quality of the simulated image is lower. The compression reference value = compressed data capacity / original data capacity. The compressed data capacity is the memory capacity occupied by the compressed flight training data, in GB. The original data capacity is the memory capacity occupied by the uncompressed flight training data, in GB.

[0144] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0145] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A training quality monitoring system for civil aviation based on a digital twin model, characterized in that, include: The data acquisition unit is used to collect training information; A state analysis unit, which is connected to the data acquisition unit, is used to determine the training state based on the attitude change value and trajectory dispersion of the training task. An analysis unit is selected, which is connected to the data acquisition unit and the state analysis unit respectively, and is used to determine the selection method of the test point according to the training state. The selection method of the test point is to determine the selection optimization method or uniformly select the test point according to the set conditions. The selection optimization method is to select the test point according to the combined paragraph or the fuel consumption slope. An image segmentation unit, connected to the selection and analysis unit, is used to determine the sub-image segmentation method based on the distribution coefficient of abrupt pixel points. The sub-image segmentation method is either uniform segmentation according to a set angle or segmentation based on pixel combinations. A quality confirmation unit, which is connected to the selection and analysis unit and the image segmentation unit respectively, is used to determine the image quality value confirmation method according to the analysis conditions. The image quality value confirmation method is to determine the region based on the scene feature value corresponding to the sub-image or to determine the whole based on the Gaussian feature value. The processing and analysis unit, which is connected to the quality confirmation unit, is used to determine the processing method based on the image quality value. The processing method is to send a quality qualified reminder to the user or to determine the quality adjustment method based on the trajectory deviation value. The quality adjustment method is to adjust the data transmission delay or compression reference value. The selection and analysis unit selects the points to be detected based on the combined paragraphs, including: Similarity analysis is performed on each trajectory point in the predicted trajectory. When performing similarity analysis on a trajectory point in the predicted trajectory image, the trajectory point is recorded as the target trajectory point, and the trajectory points that do not include the target trajectory point and are not included in the similarity combination are recorded as reference trajectory points. The set of all reference trajectory points preceding the first reference trajectory point whose shear difference value with the target trajectory point is greater than the preset shear difference value, and the target trajectory point, is denoted as a similar combination; Perform similarity analysis on trajectory points that are not recorded as similar combinations until all reference trajectory points are recorded as similar combinations. Each similar combination whose combined reference value is less than the preset combined reference value is recorded as a combined segment, and the number of points to be detected corresponding to the combined segment is determined according to the combined reference value. The number of test points corresponding to a single combined paragraph is positively correlated with the combined reference value; The setting conditions include a first setting condition and a second setting condition. The first setting condition is that the node shear float value is greater than or equal to the preset node shear float value and the node distribution coefficient is greater than or equal to the preset node distribution coefficient. The second setting condition is that the node shear float value is less than the preset node shear float value or the node distribution coefficient is less than the preset node distribution coefficient. The analysis conditions include a first analysis condition and a second analysis condition. The first analysis condition is that the comprehensive change value is greater than or equal to the preset comprehensive change value or the proportion of mutated pixels is greater than or equal to the preset proportion of mutated pixels. The second analysis condition is that the comprehensive change value is less than the preset comprehensive change value and the proportion of mutated pixels is less than the preset proportion of mutated pixels. For a single point to be tested, this point is recorded as the target point to be tested. The comprehensive change value corresponding to the target point to be tested is equal to the fuel consumption slope corresponding to the target point to be tested plus the vector angle corresponding to the target point to be tested.

2. The civil aviation training quality monitoring system based on a digital twin model according to claim 1, characterized in that, The state analysis unit determines the training state based on the attitude change values ​​and trajectory dispersion of the training task. The training state includes: The first training state is when the attitude change value is greater than or equal to the preset attitude change value or the trajectory dispersion is less than the preset trajectory dispersion. The second training state is characterized by an attitude change value less than a preset attitude change value and a trajectory dispersion greater than or equal to a preset trajectory dispersion.

3. The civil aviation training quality monitoring system based on a digital twin model according to claim 2, characterized in that, The selection analysis unit responds to the training state to determine the method for selecting the points to be detected. The first training state of the analysis unit response is selected, and the selection method of the test point is determined by the set conditions to optimize the selection method. The second training state of the analysis unit response is selected, and the detection points are selected uniformly. When selecting test points uniformly, the training time is divided into several equal parts, and the training time points corresponding to each division are used as test points.

4. The civil aviation training quality monitoring system based on a digital twin model according to claim 3, characterized in that, The selection analysis unit response setting conditions are used to determine the selection optimization method; The selection criteria for the analysis unit response are: the node shear float value is greater than or equal to the preset node shear float value and the node distribution coefficient is greater than or equal to the preset node distribution coefficient; the selection optimization method is to select the detection point based on the combined paragraph. The selected conditions for the analysis unit response are that the node shear fluctuation value is less than the preset node shear fluctuation value or the node distribution coefficient is less than the preset node distribution coefficient, and the selected optimization method is to select the detection point based on the fuel consumption slope. When selecting the optimization method based on the fuel consumption slope to select the test points, a fuel consumption change curve with training time as the horizontal axis and fuel consumption as the vertical axis is obtained. The curve is divided into several curve segments with the change points as the dividing points. The fuel consumption slope corresponding to each curve segment is detected. The curve segments with a fuel consumption slope less than the preset fuel consumption slope are recorded as feature segments. A preset number of training time points in each feature segment are selected as test points.

5. The civil aviation training quality monitoring system based on a digital twin model according to claim 1, characterized in that, The image segmentation unit determines the sub-image segmentation method based on the distribution coefficient of abrupt pixel points; If the distribution coefficient of the mutation pixel is greater than or equal to the preset mutation similarity distribution coefficient, the sub-image is divided into uniform segments according to the set angle. If the distribution coefficient of the mutated pixels is less than the preset mutation similarity distribution coefficient, the sub-image segmentation method is to segment based on pixel combinations; The set angle is negatively correlated with the distribution coefficient of abrupt pixel points.

6. The civil aviation training quality monitoring system based on a digital twin model according to claim 5, characterized in that, The image segmentation unit segments the image based on pixel combinations, wherein... The virtual image is evenly divided into several viewpoint images. Combined detection is performed on each viewpoint image. When combined detection is performed on a single viewpoint image, the viewpoint image is recorded as the target viewpoint image. Viewpoint images that do not include the target viewpoint image and are not included in the pixel combination are recorded as the reference viewpoint image. The number of mutated pixels corresponding to the target view image is accumulated one by one with the number of mutated pixels corresponding to each reference view image located after the target view image in the clockwise direction, until the total number of mutated pixels is greater than the preset total number of mutated pixels. The reference view images before the reference view image corresponding to the last accumulated number of mutated pixels and the target view image are recorded as a pixel combination. Perform combination detection on viewpoint images that have not recorded pixel combinations until pixel combinations are recorded for all viewpoint images; The pixels are divided using the starting and ending edges corresponding to each pixel combination as dividing lines.

7. The civil aviation training quality monitoring system based on a digital twin model according to claim 6, characterized in that, The quality verification unit responds to analysis conditions to determine the image quality value verification method; The analysis conditions for the quality confirmation unit response are that the comprehensive change value is greater than or equal to the preset comprehensive change value or the proportion of abruptly changed pixels is greater than or equal to the preset proportion of abruptly changed pixels. The image quality value confirmation method is to determine the region based on the scene feature value corresponding to the sub-image. When the image quality value is determined by region determination, the virtual image is segmented according to the sub-image division method to obtain several sub-images. The scene feature values ​​corresponding to each sub-image are input into the regression model to determine the image quality value. The analysis conditions for the quality confirmation unit response are that the comprehensive change value is less than the preset comprehensive change value and the proportion of abruptly changed pixels is less than the preset proportion of abruptly changed pixels. The image quality value confirmation method is to make an overall judgment based on Gaussian feature values. When the image quality value is determined as a whole, the Gaussian feature value corresponding to the virtual image is input into the regression model to determine the image quality value.

8. The civil aviation training quality monitoring system based on a digital twin model according to claim 7, characterized in that, The processing and analysis unit determines the processing method based on the image quality value; If the image quality value is greater than or equal to the preset image quality value, the processing method is to send a quality qualified reminder to the user. If the image quality value is less than the preset image quality value, the processing method is to determine the quality adjustment method based on the trajectory deviation value.

9. The civil aviation training quality monitoring system based on a digital twin model according to claim 8, characterized in that, The processing and analysis unit determines the quality adjustment method based on the trajectory deviation value; If the trajectory deviation value is greater than or equal to the preset trajectory deviation value, the quality adjustment method is to reduce the data transmission delay. If the trajectory deviation value is less than the preset trajectory deviation value, the quality adjustment method is to increase the compression reference value. The decrease in data transmission delay is positively correlated with the positional deviation, while the increase in the compression reference value is negatively correlated with the image similarity.

Images