Aircraft trajectory data correction method, device, equipment and storage medium

Abstract

The present invention relates to the technical field of data processing, and provides a method, device, equipment and storage medium for correcting aircraft trajectory data. The method performs abnormality detection on prior data of aircraft trajectory and a posteriori data of time series prediction before filtering the trajectory data, and uses a pre-trained intelligent correction model to correct the detected abnormal data, thereby reducing the error accumulation caused by the increase of the time series of the trajectory data, ensuring the accuracy of the prior data and the posteriori data, and then pairs the corrected trajectory data with the detected non-abnormal data and performs filtering processing, thereby improving the accuracy of the filtering processing, and based on iterative processing of the prior data and the posteriori data of the trajectory data until there is no abnormal data, reducing the error accumulation in the filtering processing process, thereby improving the accuracy of the aircraft trajectory data.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an aircraft trajectory data correction method, device, equipment and storage medium. Background Art

[0002] In aircraft control systems, trajectory data is the basis for achieving aircraft control. Accurate trajectory data is crucial for determining the position, speed, and attitude of the aircraft. For aircraft that need to perform complex tasks, high-precision trajectory data is a prerequisite for completing the task.

[0003] The existing method of processing aircraft trajectory data mainly uses data filtering technology, that is, using the prior data and a posteriori data of the aircraft trajectory to filter the trajectory data, which can effectively eliminate the errors caused by interference during the flight of the aircraft and obtain more accurate trajectory data. The filtering algorithm used for data filtering needs to ensure that the prior data and a posteriori data of the aircraft trajectory have high accuracy, thereby ensuring the accuracy of the filtering algorithm. Otherwise, the output of the filtered trajectory data will have large errors, and the errors will accumulate as the time series increases, seriously affecting the accuracy of the aircraft trajectory data. Summary of the invention

[0004] The present invention provides an aircraft trajectory data correction method, device, equipment and storage medium, which are used to solve the defect that the filtering processing of aircraft trajectory data in the prior art will lead to error accumulation and affect the accuracy of aircraft trajectory data.

[0005] The present invention provides a method for correcting aircraft trajectory data, comprising the following steps:

[0006] Acquire first trajectory data of the aircraft including a first timestamp; the first trajectory data includes position data, velocity data, and acceleration data;

[0007] Predicting second trajectory data of a time stamp next to the first time stamp based on the first trajectory data;

[0008] Performing anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data;

[0009] Inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data;

[0010] Pairing the third trajectory data with the non-abnormal data according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data to obtain fourth trajectory data;

[0011] The fourth trajectory data is filtered to obtain fifth trajectory data, and the fifth trajectory data is used as the first trajectory data, and the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data is returned and executed until there is no abnormal data in the first trajectory data and the second trajectory data, and the corrected target trajectory data is obtained.

[0012] According to the aircraft trajectory data correction method provided by the present invention, the predicting of second trajectory data of the next timestamp of the first timestamp based on the first trajectory data includes:

[0013] Calculating, based on the first trajectory data, an acceleration mean of acceleration data of two adjacent timestamps in the first timestamp;

[0014] According to a first speed corresponding to a target timestamp in the speed data and an average of accelerations corresponding to the target timestamp, a second speed of a next timestamp of the target timestamp is predicted, and according to a first position corresponding to the target timestamp in the position data and an average of accelerations corresponding to the target timestamp, a second position of a next timestamp of the target timestamp is predicted, so as to obtain second trajectory data of a next timestamp of the first timestamp; the target timestamp is any one of the first timestamps.

[0015] According to the aircraft trajectory data correction method provided by the present invention, the first trajectory data and the second trajectory data are subjected to abnormality detection to obtain abnormal data and non-abnormal data, including:

[0016] Based on the first timestamp and the second timestamp of the second trajectory data, performing mean calculation on the first trajectory data and the second trajectory data at the same timestamp to obtain a trajectory mean corresponding to each timestamp; the trajectory mean includes a position mean and a speed mean;

[0017] Based on the same timestamp, performing a difference calculation on the first trajectory data and the trajectory mean to obtain a first trajectory difference, and performing a difference calculation on the second trajectory data and the trajectory mean to obtain a second trajectory difference; the first trajectory difference includes a first speed difference and a first position difference, and the second trajectory difference includes a second speed difference and a second position difference;

[0018] Abnormal data and non-abnormal data in the first trajectory data and the second trajectory data are determined according to the first trajectory difference and the second trajectory difference.

[0019] According to the aircraft trajectory data correction method provided by the present invention, determining abnormal data and non-abnormal data in the first trajectory data and the second trajectory data according to the first trajectory difference and the second trajectory difference includes:

[0020] Based on the first timestamp and the second timestamp of the second trajectory data, performing a difference calculation on the first trajectory data and the second trajectory data at the same timestamp to obtain a third trajectory difference;

[0021] Calculating a normalized mean of the third trajectory difference;

[0022] Calculating uncertainty entropy of the first trajectory data and the second trajectory data according to the first trajectory difference, the second trajectory difference, the number of the first timestamps, and the normalized mean;

[0023] A numerical interval is calculated based on the uncertainty entropy, and the first trajectory difference value and the second trajectory difference value are compared with the numerical interval to determine abnormal data and non-abnormal data in the first trajectory data and the second trajectory data.

[0024] According to the aircraft trajectory data correction method provided by the present invention, the third trajectory data and the non-abnormal data are paired according to the third timestamp of the third trajectory data and the fourth timestamp of the non-abnormal data to obtain the fourth trajectory data, including:

[0025] Determine, according to the first target timestamp in the third timestamp of the third trajectory data, a second target timestamp in the fourth timestamp of the non-abnormal data that is the same as the first target timestamp; the first target timestamp is any one of the third timestamps of the third trajectory data;

[0026] Second target trajectory data corresponding to a next timestamp of the second target timestamp is acquired from the non-abnormal data, and the second target trajectory data is paired with the first target trajectory data corresponding to the first target timestamp to obtain fourth trajectory data.

[0027] According to the aircraft trajectory data correction method provided by the present invention, before inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain the third trajectory data, the method further includes:

[0028] Acquire labeling information of the abnormal data; the labeling information is used to characterize the correction of the abnormal data;

[0029] Inputting the abnormal data into the intelligent correction model for forward reasoning to obtain a forward reasoning result;

[0030] A target difference between the forward reasoning result and the annotation information is calculated, and the target difference is input into the intelligent correction model for reverse reasoning to update the model parameters of the intelligent correction model.

[0031] According to the aircraft trajectory data correction method provided by the present invention, the intelligent correction model is constructed based on a multi-layer convolutional neural network, and the multi-layer convolutional neural network at least includes an input layer, a convolutional pooling layer, a hidden layer, an attention mechanism layer and an output layer connected in sequence, wherein the convolutional pooling layer includes multiple layers.

[0032] The present invention also provides an aircraft trajectory data correction device, comprising the following modules:

[0033] A data acquisition module, used to acquire first trajectory data of the aircraft including a first timestamp; the first trajectory data includes position data, velocity data and acceleration data;

[0034] A time series prediction module, configured to predict second trajectory data of a next time stamp of the first time stamp based on the first trajectory data;

[0035] An anomaly detection module, used to perform anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data;

[0036] An intelligent correction module, used for inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data;

[0037] a data filtering module, configured to pair the third trajectory data with the non-abnormal data according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data to obtain fourth trajectory data; filter the fourth trajectory data to obtain fifth trajectory data, use the fifth trajectory data as the first trajectory data, return to and execute the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data, until there is no abnormal data in the first trajectory data and the second trajectory data, and obtain the corrected target trajectory data.

[0038] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the aircraft trajectory data correction method described in any one of the above methods is implemented.

[0039] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the aircraft trajectory data correction method as described in any one of the above is implemented.

[0040] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the aircraft trajectory data correction method described in any one of the above is implemented.

[0041] The aircraft trajectory data correction method, device, equipment and storage medium provided by the present invention ensure the accuracy of the prior data and the posterior data by performing abnormality detection on the prior data and the posterior data of the aircraft trajectory data, and correcting the detected abnormal data based on an intelligent correction model, and then pairing the corrected trajectory data with the detected non-abnormal data and performing filtering processing, thereby improving the accuracy of the filtering processing, and based on the iterative processing of the prior data and the posterior data of the trajectory data until there is no abnormal data, the error accumulation in the filtering processing process is reduced, thereby improving the accuracy of the aircraft trajectory data. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 It is a flow chart of the aircraft trajectory data correction method provided by the present invention.

[0044] Figure 2 It is one of the structural schematic diagrams of the aircraft trajectory data processing system provided by the present invention.

[0045] Figure 3 This is the second structural schematic diagram of the aircraft trajectory data processing system provided by the present invention.

[0046] Figure 4 It is a structural schematic diagram of the aircraft trajectory data correction device provided by the present invention.

[0047] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Figure 1 FIG. 1 is a flow chart of the method for correcting the trajectory data of an aircraft provided by the present invention. Figure 1 As shown, the method comprises the following steps:

[0050] Step 100, obtaining first trajectory data of the aircraft including a first timestamp; the first trajectory data includes position data, velocity data and acceleration data;

[0051] Step 200, predicting second trajectory data of a time stamp next to the first time stamp based on the first trajectory data;

[0052] Step 300, performing anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data;

[0053] Step 400, inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data;

[0054] Step 500, pairing the third trajectory data with the non-abnormal data according to the third timestamp of the third trajectory data and the fourth timestamp of the non-abnormal data to obtain fourth trajectory data;

[0055] Step 600: Filter the fourth trajectory data to obtain fifth trajectory data, and use the fifth trajectory data as the first trajectory data. Return to and execute the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data until there is no abnormal data in the first trajectory data and the second trajectory data, and obtain the corrected target trajectory data.

[0056] First, first trajectory data of the aircraft is obtained. The first trajectory data contains a timestamp, which is recorded as a first timestamp. The first trajectory data is trajectory data marked by the first timestamp. The first trajectory data includes position data, velocity data, and acceleration data.

[0057] Optionally, the first trajectory data is obtained by measuring the position, velocity and acceleration of the aircraft at a timestamp, and the first timestamp includes a position timestamp, a velocity timestamp and an acceleration timestamp. That is, the first trajectory data includes the position data, velocity data and acceleration data marked by the first timestamp.

[0058] Predicting second trajectory data of a next timestamp of the first timestamp based on the first trajectory data, optionally, the first timestamp includes multiple timestamps, that is, the first trajectory data includes trajectory data at multiple timestamps, when predicting the second trajectory data, predicting is performed for each timestamp in the first trajectory data respectively, and the predicted trajectory data of the next timestamp of each timestamp in the first trajectory data is obtained, thereby obtaining the second trajectory data.

[0059] Exemplarily, the first timestamp is composed of [timestamp 1, timestamp 2, ..., timestamp n], and the corresponding first trajectory data is [trajectory data 1, trajectory data 2, ..., trajectory data n]. Accordingly, based on trajectory data 1 at timestamp 1, the trajectory data of the next timestamp is predicted, that is, the trajectory data of timestamp 2 is predicted. By analogy, the second trajectory data is obtained in the form of [trajectory data 2, trajectory data 3, ..., trajectory data n+1], and the corresponding timestamps are [timestamp 2, timestamp 3, ..., timestamp n+1]. Among them, the first trajectory data includes position data, velocity data and acceleration data, and the trajectory data n is [position n, velocity n, acceleration n]. Similarly, the second trajectory data also includes position data, velocity data and acceleration data. Based on this, when predicting the second trajectory data, the position of the next timestamp is predicted according to the acceleration and position at the first timestamp, and the velocity of the next timestamp is predicted according to the acceleration and velocity at the first timestamp.

[0060] Optionally, the first trajectory data is a priori data of the aircraft trajectory, and the second trajectory data is a posteriori data of the aircraft trajectory, the first trajectory data is actual measurement data, and the second trajectory data is predicted data based on the actual measurement data, and the second timestamp overlaps with the first timestamp.

[0061] Furthermore, anomaly detection is performed on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data. Optionally, the first trajectory data may include abnormal data and non-abnormal data, and the second trajectory data may also include abnormal data and non-abnormal data. Anomaly detection on the first trajectory data and the second trajectory data may be performed by comparing data with the same timestamp in the first trajectory data and the second trajectory data to determine abnormal data and non-abnormal data.

[0062] The abnormal data is input into the intelligent correction model for correction to obtain third trajectory data, and the third trajectory data is paired with the non-abnormal data to obtain paired fourth trajectory data. Optionally, the pairing of the third trajectory data and the non-abnormal data is implemented based on timestamps, and the third trajectory data and the non-abnormal data are paired according to the third timestamp of the third trajectory data and the fourth timestamp of the non-abnormal data to obtain the fourth trajectory data.

[0063] In one embodiment, the third trajectory data and the non-abnormal data are paired according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data. Specifically, a time sequence position corresponding to the third trajectory data in the non-abnormal data is determined according to a sequence relationship between the third timestamp and the fourth timestamp, and the third trajectory data and the non-abnormal data are paired according to the time sequence position to obtain fourth trajectory data.

[0064] The fourth trajectory data is filtered to obtain fifth trajectory data, and the fifth trajectory data is processed in the same manner as the first trajectory data to implement iterative processing of the first trajectory data. Specifically, the fifth trajectory data is used as the first trajectory data, and accordingly, the timestamp of the fifth trajectory data is the first timestamp, and the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data is re-executed until there is no abnormal data in the first trajectory data and the second trajectory data, and the corrected target trajectory data is obtained.

[0065] Optionally, the filtering processing of the fourth trajectory data includes but is not limited to Kalman filtering, extended Kalman filtering (EKF), unscented Kalman filtering (UKF), particle swarm optimization extended Kalman filtering and interactive multiple model algorithm (IMM). Among them, Kalman filtering is an optimal estimation method based on minimum mean square error, which is suitable for state estimation of linear dynamic systems. It estimates state values ​​in real time through state prediction and observation update to generate three-dimensional trajectories. Extended Kalman filtering EKF can transform nonlinear problems into linear problems through linearization technology, and then apply the processing framework of Kalman filtering to process them. Unscented Kalman filtering (UKF) uses unscented transformation to deal with nonlinear problems, avoiding error accumulation in the linearization process. It has higher accuracy and stability when dealing with nonlinear problems and is suitable for complex aircraft trajectory tracking scenarios. Particle swarm optimization extended Kalman filtering aims at the problem that the initial parameters of extended Kalman filtering are difficult to determine when solving the motion state estimation of nonlinear models. The particle swarm optimization algorithm is introduced to adjust the parameters of the filter online. The combination of particle swarm optimization and extended Kalman filtering can obtain smaller filtering errors and improve the accuracy of state estimation. The interactive multi-model algorithm IMM solves the problem of low filtering accuracy caused by incomplete description of the motion state by a single model by establishing a set of motion state models of maneuvering targets. It can adjust the process noise covariance matrix in real time to improve filtering accuracy.

[0066] Different filtering algorithms have different advantages and can be applied to different scenarios. In practical applications, we can select a suitable filtering algorithm for filtering based on the type of aircraft, flight environment, and performance requirements of the mission to be performed, combined with factors such as the real-time performance and computational complexity of the filtering algorithm, to ensure the safe and stable operation of the aircraft.

[0067] In this embodiment, anomaly detection is performed on the prior data and a posteriori data of the aircraft's trajectory data, and the detected abnormal data is corrected based on an intelligent correction model to ensure the accuracy of the prior data and the posteriori data. The corrected trajectory data is then paired with the detected non-abnormal data and filtered, thereby improving the accuracy of the filtering process. In addition, based on the iterative processing of the prior data and the posteriori data of the trajectory data until no abnormal data exists, the error accumulation in the filtering process is reduced, thereby improving the accuracy of the aircraft trajectory data.

[0068] Optionally, when predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data, specifically the position and speed at the same timestamp in the first trajectory data are predicted respectively, and based on the relative reliability of the accelerometer, the prediction of the speed and position is achieved based on the acceleration.

[0069] Specifically, step 200 may also include:

[0070] Step 210: Calculate the acceleration mean of the acceleration data of two adjacent timestamps in the first timestamp based on the first trajectory data;

[0071] Step 220: predicting a second speed of a next timestamp of the target timestamp according to a first speed corresponding to the target timestamp in the speed data and an average value of acceleration corresponding to the target timestamp, and predicting a second position of a next timestamp of the target timestamp according to a first position corresponding to the target timestamp in the position data and an average value of acceleration corresponding to the target timestamp, to obtain second trajectory data of a next timestamp of the first timestamp; the target timestamp is any one of the first timestamps.

[0072] Based on the acceleration data in the first trajectory data, first calculate the average acceleration of any two adjacent timestamps in the first timestamp, and then, based on the speed data in the first trajectory data, predict the second speed of the next timestamp of the target timestamp according to the first speed corresponding to the target timestamp in the speed data and the average acceleration corresponding to the target timestamp. Similarly, based on the position data in the first trajectory data, predict the second position of the next timestamp of the target timestamp according to the first position corresponding to the target timestamp in the position data and the average acceleration corresponding to the target timestamp, and obtain the predicted trajectory of the next timestamp of the target timestamp. Among them, the target timestamp is any one of the first timestamps, and the predicted trajectory of the next timestamp of each target timestamp is the second trajectory data corresponding to the second timestamp.

[0073] In one embodiment, the acceleration component of the aircraft accelerometer in the spatial rectangular coordinate system at each timestamp is obtained from the first trajectory data, and the speed of the next timestamp is predicted based on the speed data of each timestamp in the first timestamp, until the speed of the next timestamp of the last timestamp in the first timestamp is predicted. Based on the reliability of the accelerometer and the short time interval between timestamps, the average of the acceleration of two adjacent timestamps is calculated for speed and position prediction. Assuming that the trajectory speed of the current timestamp in the spatial rectangular coordinate system is [Vx, Vy, Vz], the trajectory speed of the next timestamp is predicted to be [Vxa, Vya, Vza], and the acceleration data of the current timestamp is [a x , a y , a z ], the acceleration data of the next timestamp is [a xa , a ya , a za ], the interval between adjacent timestamps is t, then the speed of the next timestamp of the current timestamp is predicted according to the following formula 1:

[0074] Vxa=Vx+(a x +a xa )×t / 2; Vya=Vy+(a y +a ya )×t / 2; Vza=Vz+(a z +a za )×t / 2. (1)

[0075] Accordingly, the trajectory position is predicted in a manner substantially the same as the prediction of the trajectory speed. Specifically, based on the first trajectory data, according to the position data of each timestamp in the first timestamp, the trajectory position of the next timestamp is predicted until the trajectory position of the next timestamp of the last timestamp in the first timestamp is predicted. Based on the calculated average acceleration value at each timestamp, if the position data of the current timestamp in the spatial rectangular coordinate system is [Sx, Sy, Sz], the trajectory position of the next timestamp is predicted to be [Sxa, Sya, Sza], and the acceleration data of the current timestamp is [a x , a y , a z ], the acceleration data of the next timestamp is [a xa , a ya , a za ], the interval between adjacent timestamps is t, then the position of the next timestamp of the current timestamp is predicted according to the following formula 2:

[0076] Sxa=Sx+(a x +a xa )×t / 2; Sya=Sy+(a y +aya )×t / 2; Sza=Sz+(a z +a za )×t / 2. (2)

[0077] Optionally, the abnormality detection on the first trajectory data and the second trajectory data is achieved by calculating the mean of the first trajectory data and the second trajectory data, and then comparing the first trajectory data and the second trajectory data with the calculated mean respectively.

[0078] Based on this, step 300 may also include:

[0079] Step 310: Based on the first timestamp and the second timestamp of the second trajectory data, the first trajectory data and the second trajectory data are averaged at the same timestamp to obtain a trajectory average corresponding to each timestamp; the trajectory average includes a position average and a speed average;

[0080] Step 320: Based on the same timestamp, performing a difference calculation on the first trajectory data and the trajectory mean to obtain a first trajectory difference, and performing a difference calculation on the second trajectory data and the trajectory mean to obtain a second trajectory difference; the first trajectory difference includes a first speed difference and a first position difference, and the second trajectory difference includes a second speed difference and a second position difference;

[0081] Step 330 : determining abnormal data and non-abnormal data in the first trajectory data and the second trajectory data according to the first trajectory difference and the second trajectory difference.

[0082] Based on the first timestamp of the first trajectory data and the second timestamp of the second trajectory data, average calculation is performed on the first trajectory data and the second trajectory data at the same timestamp to obtain trajectory averages corresponding to each timestamp, where the trajectory averages at least include a position average and a speed average.

[0083] Then, based on the same timestamp, the first trajectory data and the trajectory mean are subjected to difference calculation to obtain a first trajectory difference, and the second trajectory data and the trajectory mean are subjected to difference calculation to obtain a second trajectory difference, wherein the first trajectory difference includes a first position difference and a first speed difference, and the second trajectory difference includes a second position difference and a second speed difference. According to the first trajectory difference and the second trajectory difference, abnormal data and non-abnormal data in the first trajectory data and the second trajectory data are determined.

[0084] Optionally, based on any timestamp, the position data at the timestamp in the first trajectory data and the position mean at the timestamp in the trajectory mean are calculated to obtain a first position difference, and the speed data at the timestamp in the first trajectory data and the speed mean at the timestamp in the trajectory mean are calculated to obtain a first speed difference. Correspondingly, based on the timestamp, the position data at the timestamp in the second trajectory data and the position mean at the timestamp in the trajectory mean are calculated to obtain a second position difference, and the speed data at the timestamp in the second trajectory data and the speed mean at the timestamp in the trajectory mean are calculated to obtain a second speed difference.

[0085] In one embodiment, based on the target timestamp in the first timestamp and the second timestamp, the average of the first speed and the second speed is calculated to obtain the speed average, and the average of the first position and the second position is calculated to obtain the position average; wherein the first speed is the speed data corresponding to the target timestamp in the first trajectory data, the second speed is the position data corresponding to the target timestamp in the second trajectory data, the first position is the position data corresponding to the target timestamp in the first trajectory data, and the second position is the position data corresponding to the target timestamp in the second trajectory data; the target timestamp is any one of the first timestamp and the second timestamp.

[0086] Further, based on the target timestamp, the difference between the first speed and the speed mean, and the difference between the first position and the position mean are calculated to obtain a first speed difference corresponding to the first speed and a first position difference corresponding to the first position; accordingly, based on the target timestamp, the difference between the second speed and the speed mean, and the difference between the second position and the position mean are calculated to obtain a second speed difference corresponding to the second speed and a second position difference corresponding to the second position.

[0087] According to the calculated first trajectory difference and second trajectory difference, abnormal data and non-abnormal data are determined, specifically, the first trajectory difference and the second trajectory difference are compared with a preset threshold value, and the abnormal data and non-abnormal data are determined according to the comparison. Exemplarily, according to the comparison result, in the first trajectory data, the first trajectory difference greater than the preset threshold value is abnormal data, in the second trajectory data, the second trajectory difference greater than the preset threshold value is abnormal data, and in the first trajectory data and the second trajectory data, except for the abnormal data is non-abnormal data.

[0088] In one embodiment, when the first trajectory difference is greater than a preset threshold, at least one of the first position difference and the first speed difference in the first trajectory difference is greater than the preset threshold. Correspondingly, when the second trajectory difference is greater than the preset threshold, at least one of the second position difference and the second speed difference in the second trajectory difference is greater than the preset threshold. In another embodiment, the preset threshold includes a first threshold corresponding to the position difference and a second threshold corresponding to the speed difference. When the first trajectory difference is compared with the preset threshold, if the first position difference is greater than the first threshold, and / or the first speed difference is greater than the second threshold, then the first trajectory difference is greater than the preset threshold; correspondingly, when the second trajectory difference is compared with the preset threshold, if the second position difference is greater than the first threshold, and / or the second speed difference is greater than the second threshold, then the second trajectory difference is greater than the preset threshold. The first threshold and the second threshold may be the same or different, and there is no specific limitation on this.

[0089] Optionally, when determining abnormal data and non-abnormal data in the first trajectory data and the second trajectory data according to the first trajectory difference and the second trajectory difference, specifically, uncertainty entropies of the first trajectory data and the second trajectory data are calculated respectively according to the first trajectory difference and the second trajectory difference, and abnormal data and non-abnormal data are identified according to the uncertainty entropy.

[0090] Specifically, step 330 may also include:

[0091] Step 331, based on the first timestamp and the second timestamp of the second trajectory data, performing a difference calculation on the first trajectory data and the second trajectory data at the same timestamp to obtain a third trajectory difference;

[0092] Step 332, calculating the normalized mean of the third trajectory difference;

[0093] Step 333, calculating the uncertainty entropy of the first trajectory data and the second trajectory data according to the first trajectory difference, the second trajectory difference, the number of the first timestamps and the normalized mean;

[0094] Step 334 : Calculate a numerical interval based on the uncertainty entropy, and compare the first trajectory difference and the second trajectory difference with the numerical interval to determine abnormal data and non-abnormal data in the first trajectory data and the second trajectory data.

[0095] Based on the first timestamp and the second timestamp, the first trajectory data and the second trajectory data are difference-calculated at the same timestamp to obtain a third trajectory difference, and a normalized mean of the third trajectory difference is calculated. Optionally, the normalized mean of the third trajectory difference represents a normalized mean of an error between an actual measurement value and a predicted value of the trajectory data of the aircraft, and in the absence of an abnormality or fault, the theoretical value of the normalized mean is 0.

[0096] The uncertainty entropy of the first trajectory data and the second trajectory data is calculated according to the calculated first trajectory difference, the second trajectory difference, the number of first timestamps and the normalized mean. Optionally, the first trajectory data and the second trajectory data correspond to one uncertainty entropy, or the uncertainty entropy is calculated for the first trajectory data and the second trajectory data respectively, and the obtained uncertainty entropy includes a first entropy value corresponding to the first trajectory data and a second entropy value corresponding to the second trajectory data.

[0097] Furthermore, a numerical interval is calculated based on the calculated uncertainty entropy, and the first trajectory difference and the second trajectory difference are compared with the numerical interval, so as to determine abnormal data and non-abnormal data in the first trajectory data and the second trajectory data.

[0098] In one embodiment, the uncertainty entropy The calculation method is:

[0099] ; (3)

[0100] In formula 3, The timestamp number representing the first timestamp or the second timestamp, represents the trajectory difference at the i-th timestamp in the first trajectory difference or the second trajectory difference, The normalized mean of the third trajectory difference between the first trajectory data and the second trajectory data.

[0101] Further, according to the calculated uncertainty entropy, the corresponding numerical interval is calculated, and the numerical interval is [-2En, 2En]. When the uncertainty entropy is calculated based on the first trajectory data and the second trajectory data, the first trajectory data and the second trajectory data correspond to different numerical intervals. The first trajectory difference of the first trajectory data is compared with the corresponding numerical interval. The first trajectory difference within the numerical interval is non-abnormal data, and the first trajectory difference outside the numerical interval is abnormal data. Correspondingly, the second trajectory difference of the second trajectory data is compared with the corresponding numerical interval. The second trajectory difference within the numerical interval is non-abnormal data, and the second trajectory difference outside the numerical interval is abnormal data.

[0102] Furthermore, the pairing of the third trajectory data and the non-abnormal data is implemented based on the timestamp, and step 500 may further include:

[0103] Step 510: determining a second target timestamp in the fourth timestamp of the non-abnormal data that is the same as the first target timestamp according to the first target timestamp in the third timestamp of the third trajectory data; the first target timestamp is any one of the third timestamps of the third trajectory data;

[0104] Step 520: Acquire second target trajectory data corresponding to a timestamp next to the second target timestamp from the non-abnormal data, and pair the second target trajectory data with the first target trajectory data corresponding to the first target timestamp to obtain fourth trajectory data.

[0105] According to the first target timestamp in the third timestamp of the third trajectory data, a second target timestamp in the fourth timestamp of the non-abnormal data which is the same as the first target timestamp is determined, and the first target timestamp is any one of the third timestamps of the third trajectory data.

[0106] The second target trajectory data corresponding to the next timestamp under the second target timestamp is obtained from the non-abnormal data, and the second target trajectory data is paired with the first target trajectory data corresponding to the first target timestamp in the third trajectory data to obtain the fourth trajectory data. That is, the third trajectory data is paired with the data of the next timestamp of the same timestamp in the non-abnormal data to obtain the paired fourth trajectory data.

[0107] Optionally, an intelligent correction model for correcting abnormal data is specifically constructed based on a multi-layer convolutional neural network, which includes at least an input layer, a convolutional pooling layer, a hidden layer, an attention mechanism layer and an output layer connected in sequence, wherein the convolutional pooling layer includes multiple layers. The data vectors of each dimension of the abnormal data in the spatial rectangular coordinate system are used as the input of the input layer, and then after being processed by multiple convolutional pooling layers, the hidden layer is connected to the attention mechanism layer for feature extraction and correction, and the corrected data is output through the output layer. Correction of abnormal data based on the intelligent correction model can eliminate errors in aircraft trajectory data, thereby improving the accuracy of prior data and posterior data of aircraft trajectory.

[0108] Optionally, the attention mechanism layer is constructed using a spatial attention mechanism, or the attention mechanism layer is constructed using a multidimensional attention mechanism, which at least includes a spatial attention mechanism, and the multidimensional attention mechanism may also include at least one of a temporal attention mechanism, a local attention mechanism, and a global attention mechanism.

[0109] The aircraft trajectory data correction method provided in the embodiment of the present invention further includes pre-training of the intelligent correction model. Based on this, before step 400, the following steps may also be included:

[0110] Step 401, obtaining labeling information of the abnormal data; the labeling information is used to characterize the correction of the abnormal data;

[0111] Step 402, inputting the abnormal data into the intelligent correction model for forward reasoning to obtain a forward reasoning result;

[0112] Step 403, calculating a target difference between the forward reasoning result and the annotation information, and inputting the target difference into the intelligent correction model for reverse reasoning to update the model parameters of the intelligent correction model.

[0113] Obtain labeling information for abnormal data, which can be obtained through manual labeling, and is used to characterize the correction of abnormal data. The abnormal information is input into the intelligent correction model for forward reasoning to obtain forward reasoning results, which characterize the correction results of the intelligent model for the abnormal data.

[0114] Furthermore, a target difference between the forward reasoning result and the acquired annotation information is calculated, and the target difference is input into the intelligent correction model for layer-by-layer reverse reasoning, thereby adjusting and updating the model parameters of the intelligent correction model.

[0115] In one embodiment, firstly, based on iterative processing of trajectory data, the intelligent correction model is iteratively trained according to the annotation information. After the intelligent correction model training is completed, the trained intelligent correction model is used to correct the detected abnormal data during the iterative processing of trajectory data.

[0116] Specifically, multiple batches of trajectory data of the aircraft are obtained as the first trajectory data, and the first trajectory data includes position data, velocity data and acceleration data marked by timestamps in a spatial rectangular coordinate system. In each round of iterative processing, the position and velocity of the next timestamp of the first trajectory data are predicted to obtain the second trajectory data, and the first trajectory data and the second trajectory data are detected for abnormality to obtain abnormal data and non-abnormal data, and then the annotation information of the abnormal data is obtained, and the intelligent correction model is trained based on the annotation information to update the model parameters. The abnormal data and non-abnormal data with the annotation information are filtered, and the trajectory data after the filtering is used as the first trajectory data, and the step of predicting the position and velocity of the next timestamp of the first trajectory data to obtain the second trajectory data is re-executed until no abnormal data is detected from the first trajectory data and the second trajectory data, then the training of the intelligent correction model is completed, and the detected abnormal data is subsequently corrected using the intelligent correction model.

[0117] Among them, in each round of iterative processing of trajectory data, the training of the intelligent correction model is to input the abnormal data into the intelligent correction model for forward reasoning to obtain the forward reasoning result, and then calculate the difference between the forward reasoning result and the annotation information, and input the difference into the intelligent correction model for layer-by-layer reverse reasoning to update the model parameters.

[0118] After the intelligent correction model training is completed, the intelligent correction model is used to correct the abnormal data. Specifically, the first trajectory data of the aircraft is obtained. The first trajectory data includes position data, velocity data and acceleration data marked with timestamps. The position and velocity of the next timestamp of the first trajectory data are predicted based on the first trajectory data to obtain the second trajectory data.

[0119] Anomaly detection is performed on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data, and the abnormal data is input into the trained intelligent correction model for correction to obtain the third trajectory data. The third trajectory data is paired with the non-abnormal data to obtain the fourth trajectory data, and the fourth trajectory data is filtered to obtain the fifth trajectory data. The fifth trajectory data is used as the first trajectory data to re-predict the trajectory, thereby realizing iterative processing of the first trajectory data. In each round of iterative processing, trajectory prediction, anomaly detection, abnormal data correction based on the intelligent correction model, and data filtering are included. Until no abnormal data is detected in the first trajectory data and the second trajectory data, the corrected target trajectory data is obtained.

[0120] In one embodiment, the aircraft trajectory data correction method provided by the embodiment of the present invention is applied to an aircraft trajectory data processing system, referring to Figure 2 The system structure diagram shown in FIG. 1 includes a data acquisition module, a time series prediction module, an anomaly detection module, a labeling module, a pre-training module, an intelligent correction module, a data filtering module, and an output module. The system's processing of trajectory data mainly includes a model pre-training stage based on manual labeling and a model application stage based on an intelligent correction model.

[0121] During the model training phase, the data acquisition module is used to obtain the trajectory data of the aircraft with timestamps, and input the acquired trajectory data into the timing prediction module; the timing prediction module is used to predict the trajectory data of the next timestamp based on the timestamp of the trajectory data, and input the input trajectory data and the predicted trajectory data into the anomaly detection module for processing; the anomaly detection module performs anomaly detection on the input trajectory data to determine abnormal data and non-abnormal data.

[0122] On the one hand, the anomaly detection module is also used to input the abnormal data into the labeling module for labeling; the labeling module is used to obtain the labeling information of the abnormal data, and input the labeling information and the abnormal data into the pre-training module; the pre-training module inputs the abnormal data into the intelligent correction module, trains the intelligent correction model in the intelligent correction module, and the intelligent correction model in the intelligent correction module performs forward reasoning on the abnormal data, obtains the forward reasoning result and returns it to the pre-training module; the pre-training module calculates the difference between the forward reasoning result and the labeling information, and inputs the calculated difference into the intelligent correction module; the intelligent correction model in the intelligent correction module performs reverse reasoning based on the input difference, thereby updating the model parameters.

[0123] On the other hand, the anomaly detection module inputs non-abnormal data into the data filtering module, and the labeling module inputs abnormal data with labeling information into the data filtering module. The data filtering module performs pairing and filtering on the input trajectory data, and inputs the filtered trajectory data into the time series prediction module to continue trajectory prediction, implement iterative processing of trajectory data, and iterative training of the intelligent correction model. The training of the intelligent correction model in the intelligent correction module is completed until the anomaly detection module no longer outputs abnormal data.

[0124] In some embodiments, both the intelligent correction module and the pre-training module are provided with an intelligent correction model. In the model pre-training stage, the pre-training module performs model training based on abnormal data and annotation information, and then inputs the updated model parameters into the intelligent correction module for storage, and updates the model parameters of the intelligent correction model in the intelligent correction module. The specific model training process includes forward reasoning of abnormal data, calculating the difference between the forward reasoning result and the annotation information, and reverse reasoning based on the difference, which will not be repeated here.

[0125] After completing the training of the intelligent correction model, the intelligent correction module is used to replace the annotation module to correct the abnormal data in the subsequent trajectory data. Figure 3The schematic diagram of the system structure shown in the figure shows that in the model application stage, the data acquisition module is used to obtain the trajectory data to be processed by the aircraft and input it into the time series prediction module. The trajectory data includes the position data, velocity data and acceleration data marked by the timestamp. The time series prediction module is used to perform time series-based trajectory prediction on the input trajectory data, specifically, predict the trajectory of the next timestamp of its timestamp according to the input trajectory data, obtain the predicted trajectory data, and input the input trajectory data and the predicted trajectory data into the abnormality detection module for abnormality detection. The abnormality detection module is used to perform abnormality detection on the input trajectory data, so as to determine the abnormal data and non-abnormal data, and then input the abnormal data into the intelligent correction module for correction, and input the non-abnormal data into the data filtering module. The intelligent correction module uses the pre-trained intelligent correction model to correct the input abnormal data, and inputs the corrected trajectory data into the data filtering module. The data filtering module pairs the input non-abnormal data and the corrected trajectory data according to the timestamp, and filters the paired trajectory data, and inputs the filtered data into the time series prediction module, continues to perform trajectory prediction, and completes an iterative processing of the trajectory data. Until there is no abnormal data in the detection result of the abnormal detection module, the iterative processing of the trajectory data is completed to obtain the corrected target trajectory data.

[0126] Among them, the prediction of trajectory data by the timing prediction module, the detection process of abnormal data by the anomaly detection module, and the pairing process of the corrected abnormal data and non-abnormal data by the data filtering module can be referenced with the descriptions in the various embodiments of the aircraft trajectory data correction method provided by the present invention above, and will not be repeated here.

[0127] In this embodiment, in order to accurately mark and correct abnormal data in the trajectory data of the aircraft under actual circumstances, the trajectory data of the aircraft is used as input, trajectory prediction and abnormality detection processing are performed in sequence, abnormal data are detected from the prior data and a posteriori data of the aircraft trajectory, and based on a pre-trained intelligent correction model, the abnormal data is corrected to reduce the accumulation of time series errors in the trajectory data and improve the accuracy of the prior data and a posteriori data. Finally, the corrected trajectory data is filtered to obtain accurate trajectory data, thereby accurately marking and correcting the trajectory data with errors, effectively improving the accuracy of the trajectory data.

[0128] The marking of the trajectory data specifically refers to marking whether the trajectory data is abnormal, and the correction of the trajectory data specifically refers to correcting the trajectory data marked as abnormal.

[0129] Furthermore, for the intelligent correction model, a combination of manual labeling and intelligent correction is adopted. First, in the context of errors in the prior data and a posteriori data of the aircraft trajectory, in order to accurately label and correct the aircraft trajectory data in actual situations, abnormal data is detected by an abnormal detection algorithm based on uncertainty entropy, and then the labeling information of the abnormal data is obtained by manual labeling for model training, and the correction knowledge of manual labeling is converted into a multi-layer convolutional neural network model of intelligent correction. After completing the training of the intelligent correction model, before filtering the trajectory data, the detected abnormal data is automatically corrected using the pre-trained intelligent correction model, avoiding the influence of erroneous data on the filtering algorithm, so that more accurate flight trajectory data can be obtained after data filtering, improving the control accuracy and operation safety of the aircraft.

[0130] In the process of predicting trajectory data, the method of time series data prediction is adopted, and the measurement value of the accelerometer of the aircraft (that is, the acceleration data in the first trajectory data) is used as a trusted device. The average acceleration of each timestamp and the next timestamp in the prior data is used as the average acceleration of the time series prediction of the timestamp, which is used to predict the position and speed of the next timestamp. This can effectively avoid the acceleration error when the aircraft makes large maneuvers, thereby helping to reduce the error of the prior data of the time series prediction.

[0131] In the process of detecting abnormal data, the mean and difference of the trajectory data of each dimension of the spatial rectangular coordinate system of each time stamp and the trajectory data of the time series prediction are calculated respectively, and then based on the same timestamp, the difference between the mean at the timestamp and the prior trajectory data, and the difference between the mean at the timestamp and the trajectory data of the time series prediction are calculated respectively. Finally, based on the uncertainty entropy of the cloud model theory, uncertainty analysis is performed on all the differences obtained to obtain the entropy value, and the entropy value of the distribution of low-credibility data is quantified to obtain accurately quantified abnormal data. This method not only improves the detection accuracy of abnormal data, but also greatly reduces the detection time and improves the detection efficiency compared to the manual abnormality detection method.

[0132] In the process of correcting abnormal data, in the model training stage, manual labeling is used to manually correct the abnormal data, and the intelligent correction model is trained based on the manual labeling information. Through iterative processing, the intelligent correction model built based on the multi-layer convolutional neural network can fully learn the correction knowledge of abnormal data by manual labeling, so as to ensure that the intelligent correction model can correctly correct the detected abnormal data in the model application stage, thereby improving the correction efficiency and accuracy.

[0133] The aircraft trajectory data correction device provided by the present invention is described below. The aircraft trajectory data correction device described below and the aircraft trajectory data correction method described above can be referenced to each other.

[0134] Reference Figure 4 The aircraft trajectory data correction device provided by the embodiment of the present invention includes:

[0135] The data acquisition module 10 is used to acquire first trajectory data of the aircraft including a first timestamp; the first trajectory data includes position data, velocity data and acceleration data;

[0136] A time series prediction module 20, configured to predict second trajectory data of a time stamp next to the first time stamp based on the first trajectory data;

[0137] An anomaly detection module 30, used to perform anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data;

[0138] An intelligent correction module 40, used for inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data;

[0139] The data filtering module 50 is configured to pair the third trajectory data with the non-abnormal data according to the third timestamp of the third trajectory data and the fourth timestamp of the non-abnormal data to obtain fourth trajectory data; filter the fourth trajectory data to obtain fifth trajectory data, use the fifth trajectory data as the first trajectory data, return to and execute the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data, until there is no abnormal data in the first trajectory data and the second trajectory data, and obtain the corrected target trajectory data.

[0140] In one embodiment, the timing prediction module 20 is further used for:

[0141] Calculating, based on the first trajectory data, an acceleration mean of acceleration data of two adjacent timestamps in the first timestamp;

[0142] According to a first speed corresponding to a target timestamp in the speed data and an average of accelerations corresponding to the target timestamp, a second speed of a next timestamp of the target timestamp is predicted, and according to a first position corresponding to the target timestamp in the position data and an average of accelerations corresponding to the target timestamp, a second position of a next timestamp of the target timestamp is predicted, so as to obtain second trajectory data of a next timestamp of the first timestamp; the target timestamp is any one of the first timestamps.

[0143] In one embodiment, the anomaly detection module 30 is further used to:

[0144] Based on the first timestamp and the second timestamp of the second trajectory data, performing mean calculation on the first trajectory data and the second trajectory data at the same timestamp to obtain a trajectory mean corresponding to each timestamp; the trajectory mean includes a position mean and a speed mean;

[0145] Based on the same timestamp, performing a difference calculation on the first trajectory data and the trajectory mean to obtain a first trajectory difference, and performing a difference calculation on the second trajectory data and the trajectory mean to obtain a second trajectory difference; the first trajectory difference includes a first speed difference and a first position difference, and the second trajectory difference includes a second speed difference and a second position difference;

[0146] Abnormal data and non-abnormal data in the first trajectory data and the second trajectory data are determined according to the first trajectory difference and the second trajectory difference.

[0147] In one embodiment, the anomaly detection module 30 is further used to:

[0148] Based on the first timestamp and the second timestamp of the second trajectory data, performing a difference calculation on the first trajectory data and the second trajectory data at the same timestamp to obtain a third trajectory difference;

[0149] Calculating a normalized mean of the third trajectory difference;

[0150] Calculating uncertainty entropy of the first trajectory data and the second trajectory data according to the first trajectory difference, the second trajectory difference, the number of the first timestamps, and the normalized mean;

[0151] A numerical interval is calculated based on the uncertainty entropy, and the first trajectory difference value and the second trajectory difference value are compared with the numerical interval to determine abnormal data and non-abnormal data in the first trajectory data and the second trajectory data.

[0152] In one embodiment, the data filtering module 50 is further used for:

[0153] Determine, according to the first target timestamp in the third timestamp of the third trajectory data, a second target timestamp in the fourth timestamp of the non-abnormal data that is the same as the first target timestamp; the first target timestamp is any one of the third timestamps of the third trajectory data;

[0154] Second target trajectory data corresponding to a next timestamp of the second target timestamp is acquired from the non-abnormal data, and the second target trajectory data is paired with the first target trajectory data corresponding to the first target timestamp to obtain fourth trajectory data.

[0155] In one embodiment, the aircraft trajectory data correction device further includes a pre-training module for:

[0156] Acquire labeling information of the abnormal data; the labeling information is used to characterize the correction of the abnormal data;

[0157] Inputting the abnormal data into the intelligent correction model for forward reasoning to obtain a forward reasoning result;

[0158] A target difference between the forward reasoning result and the annotation information is calculated, and the target difference is input into the intelligent correction model for reverse reasoning to update the model parameters of the intelligent correction model.

[0159] In one embodiment, the intelligent correction model is constructed based on a multi-layer convolutional neural network, which includes at least an input layer, a convolutional pooling layer, a hidden layer, an attention mechanism layer and an output layer connected in sequence, wherein the convolutional pooling layer includes multiple layers.

[0160] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communications interface 520 and the memory 530 communicate with each other via the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the steps of the aircraft trajectory data correction method, for example including:

[0161] Acquire first trajectory data of the aircraft including a first timestamp; the first trajectory data includes position data, velocity data, and acceleration data;

[0162] Predicting second trajectory data of a time stamp next to the first time stamp based on the first trajectory data;

[0163] Performing anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data;

[0164] Inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data;

[0165] Pairing the third trajectory data with the non-abnormal data according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data to obtain fourth trajectory data;

[0166] The fourth trajectory data is filtered to obtain fifth trajectory data, and the fifth trajectory data is used as the first trajectory data, and the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data is returned and executed until there is no abnormal data in the first trajectory data and the second trajectory data, and the corrected target trajectory data is obtained.

[0167] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0168] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can perform the steps of the aircraft trajectory data correction method provided by the above methods, for example, including:

[0169] Acquire first trajectory data of the aircraft including a first timestamp; the first trajectory data includes position data, velocity data, and acceleration data;

[0170] Predicting second trajectory data of a time stamp next to the first time stamp based on the first trajectory data;

[0171] Performing anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data;

[0172] Inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data;

[0173] Pairing the third trajectory data with the non-abnormal data according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data to obtain fourth trajectory data;

[0174] The fourth trajectory data is filtered to obtain fifth trajectory data, and the fifth trajectory data is used as the first trajectory data, and the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data is returned and executed until there is no abnormal data in the first trajectory data and the second trajectory data, and the corrected target trajectory data is obtained.

[0175] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the aircraft trajectory data correction method provided by the above methods, for example, including:

[0176] Acquire first trajectory data of the aircraft including a first timestamp; the first trajectory data includes position data, velocity data, and acceleration data;

[0177] Predicting second trajectory data of a time stamp next to the first time stamp based on the first trajectory data;

[0178] Performing anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data;

[0179] Inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data;

[0180] Pairing the third trajectory data with the non-abnormal data according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data to obtain fourth trajectory data;

[0181] The fourth trajectory data is filtered to obtain fifth trajectory data, and the fifth trajectory data is used as the first trajectory data, and the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data is returned and executed until there is no abnormal data in the first trajectory data and the second trajectory data, and the corrected target trajectory data is obtained.

[0182] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0183] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0184] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for correcting aircraft trajectory data, characterized in that: include: Acquire first trajectory data of the aircraft including a first timestamp; The first trajectory data includes position data, velocity data and acceleration data; Predicting second trajectory data of a time stamp next to the first time stamp based on the first trajectory data; Performing anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data; Inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data; Pairing the third trajectory data with the non-abnormal data according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data to obtain fourth trajectory data; The fourth trajectory data is filtered to obtain fifth trajectory data, and the fifth trajectory data is used as the first trajectory data, and the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data is returned and executed until there is no abnormal data in the first trajectory data and the second trajectory data, and the corrected target trajectory data is obtained.

2. The method for correcting aircraft trajectory data according to claim 1, characterized in that: The predicting, based on the first trajectory data, second trajectory data of a next timestamp of the first timestamp includes: Calculating, based on the first trajectory data, an acceleration mean of acceleration data of two adjacent timestamps in the first timestamp; According to a first speed corresponding to a target timestamp in the speed data and an average of accelerations corresponding to the target timestamp, a second speed of a next timestamp of the target timestamp is predicted, and according to a first position corresponding to the target timestamp in the position data and an average of accelerations corresponding to the target timestamp, a second position of a next timestamp of the target timestamp is predicted, so as to obtain second trajectory data of a next timestamp of the first timestamp; the target timestamp is any one of the first timestamps.

3. The method for correcting aircraft trajectory data according to claim 1, characterized in that: The performing anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data includes: Based on the first timestamp and the second timestamp of the second trajectory data, performing mean calculation on the first trajectory data and the second trajectory data at the same timestamp to obtain a trajectory mean corresponding to each timestamp; the trajectory mean includes a position mean and a speed mean; Based on the same timestamp, performing a difference calculation on the first trajectory data and the trajectory mean to obtain a first trajectory difference, and performing a difference calculation on the second trajectory data and the trajectory mean to obtain a second trajectory difference; the first trajectory difference includes a first speed difference and a first position difference, and the second trajectory difference includes a second speed difference and a second position difference; Abnormal data and non-abnormal data in the first trajectory data and the second trajectory data are determined according to the first trajectory difference and the second trajectory difference.

4. The method for correcting aircraft trajectory data according to claim 3, characterized in that: The determining, according to the first trajectory difference and the second trajectory difference, abnormal data and non-abnormal data in the first trajectory data and the second trajectory data comprises: Based on the first timestamp and the second timestamp of the second trajectory data, performing a difference calculation on the first trajectory data and the second trajectory data at the same timestamp to obtain a third trajectory difference; Calculating a normalized mean of the third trajectory difference; Calculating uncertainty entropy of the first trajectory data and the second trajectory data according to the first trajectory difference, the second trajectory difference, the number of the first timestamps, and the normalized mean; A numerical interval is calculated based on the uncertainty entropy, and the first trajectory difference value and the second trajectory difference value are compared with the numerical interval to determine abnormal data and non-abnormal data in the first trajectory data and the second trajectory data.

5. The method for correcting aircraft trajectory data according to claim 1, characterized in that: The step of pairing the third trajectory data with the non-abnormal data according to the third timestamp of the third trajectory data and the fourth timestamp of the non-abnormal data to obtain fourth trajectory data includes: Determine, according to the first target timestamp in the third timestamp of the third trajectory data, a second target timestamp in the fourth timestamp of the non-abnormal data that is the same as the first target timestamp; the first target timestamp is any one of the third timestamps of the third trajectory data; Second target trajectory data corresponding to a next timestamp of the second target timestamp is acquired from the non-abnormal data, and the second target trajectory data is paired with the first target trajectory data corresponding to the first target timestamp to obtain fourth trajectory data.

6. The method for correcting aircraft trajectory data according to claim 1, characterized in that: Before inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain the third trajectory data, the method further includes: Acquire labeling information of the abnormal data; the labeling information is used to characterize the correction of the abnormal data; Inputting the abnormal data into the intelligent correction model for forward reasoning to obtain a forward reasoning result; A target difference between the forward reasoning result and the annotation information is calculated, and the target difference is input into the intelligent correction model for reverse reasoning to update the model parameters of the intelligent correction model.

7. The method for correcting aircraft trajectory data according to claim 1, characterized in that: The intelligent correction model is constructed based on a multi-layer convolutional neural network, which includes at least an input layer, a convolutional pooling layer, a hidden layer, an attention mechanism layer and an output layer connected in sequence, wherein the convolutional pooling layer includes multiple layers.

8. An aircraft trajectory data correction device, characterized in that: include: A data acquisition module, used for acquiring first trajectory data of the aircraft including a first timestamp; The first trajectory data includes position data, velocity data and acceleration data; A time series prediction module, configured to predict second trajectory data of a time stamp next to the first time stamp based on the first trajectory data; An anomaly detection module, used to perform anomaly detection on the first trajectory data and the second trajectory data to obtain abnormal data and non-abnormal data; An intelligent correction module, used for inputting the abnormal data into a pre-trained intelligent correction model for correction to obtain third trajectory data; a data filtering module, configured to pair the third trajectory data with the non-abnormal data according to a third timestamp of the third trajectory data and a fourth timestamp of the non-abnormal data to obtain fourth trajectory data; filter the fourth trajectory data to obtain fifth trajectory data, use the fifth trajectory data as the first trajectory data, return to and execute the step of predicting the second trajectory data of the next timestamp of the first timestamp based on the first trajectory data, until there is no abnormal data in the first trajectory data and the second trajectory data, and obtain the corrected target trajectory data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the aircraft trajectory data correction method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the aircraft trajectory data correction method according to any one of claims 1 to 7 is implemented.

Images