CN113393032B - Track circulation prediction method based on resampling - Google Patents
Track circulation prediction method based on resampling Download PDFInfo
- Publication number
- CN113393032B CN113393032B CN202110658739.1A CN202110658739A CN113393032B CN 113393032 B CN113393032 B CN 113393032B CN 202110658739 A CN202110658739 A CN 202110658739A CN 113393032 B CN113393032 B CN 113393032B
- Authority
- CN
- China
- Prior art keywords
- track
- data
- batch
- resampling
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Strategic Management (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Radar Systems Or Details Thereof (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a track circulation prediction method based on resampling, which mainly solves the problems of short prediction time length and overlarge prediction error caused by the change of a target in a future motion state in track prediction in the prior art. The implementation scheme is as follows: simulating a historical track and a future track of the maneuvering target; the target historical track data is subjected to pretreatment of filtering, resampling and normalization in sequence; constructing a neural network model consisting of a Bi-LSTM layer, a Dropout layer, a Dense layer and an activation layer, and training the neural network model by utilizing the preprocessed track data; generating partial historical track data by using a circulation strategy, and calculating the partial historical track data by using trained neural network model parameters; and carrying out smooth filtering on the calculation result to obtain a final predicted track. The method has the advantages of smaller prediction error and longer prediction time, can still obtain more accurate prediction tracks when the motion state of the target changes in the future, and can be used for target tracking.
Description
Technical Field
The invention belongs to the technical field of communication, and particularly relates to a cyclic prediction method of a track, which can be used for target tracking.
Background
The target track prediction technology is used for accurately predicting future track state information of a target, and is one of key technologies in the field of target tracking.
With the trend of complicating aviation flight environment, the sensor can not continuously detect target track information due to the influence of uncertain factors such as systematic errors, bad weather, abnormal performance of the sensor, and the like, so that the problem of affecting flight safety can be caused. Therefore, the method needs to have certain track prediction capability, provides more complete data information for subsequent target tracking, and ensures flight safety.
At present, the flight path prediction algorithm is mainly divided into a dynamics model based on flight performance parameters, an optimal estimation model based on parameters and a machine learning method based on historical data.
In the aspect of a dynamics model, the method for predicting the paper track of the strength provides a track prediction method based on a large circular track and an equiangular track in the application analysis of the paper track prediction method in the course flight, and information such as the circular track and the equiangular track is introduced into the model. Liao Chaowei an aerodynamic-based track prediction method is provided in the aircraft runway sliding track prediction method, and the method is used for carrying out stress analysis on a target and establishing a sliding dynamics model. Porretta in paper Performance Evaluation of a Novel 4D Trajectory Prediction Model for Civil Aircraft presents an aircraft performance model that considers wind speed, lateral braking force of the aircraft, and speed estimation in combination. However, model parameters such as weather forecast, scene control intention, flight plan information and the like required by the methods are difficult to obtain in practical application, so that under the condition that the model parameters are missing, the target cannot be accurately modeled, and the prediction result is inaccurate.
In terms of parameter optimal estimation, the most classical algorithm is to predict the trajectory of an aircraft by performing a target state estimation based on kalman filtering. Gong Shuli an interactive multimode combined unscented Kalman filtering algorithm is proposed in the paper for airport surface moving object tracking based on an IMM algorithm, and the algorithm models the motion process of an airplane and performs predictive calculation of a track. Shang Xinmin A collision-free 4D flight path prediction algorithm based on the hybrid system theory is provided in paper, and the algorithm builds a parameter evolution model of an aircraft dynamics model aiming at the motion characteristics of the aircraft in different flight sections, builds a state transition model to model the switching among different flight sections, and completes the flight path prediction of the aircraft by adjusting corresponding parameters. However, the execution efficiency of the algorithm is low, and when the motion state of the target is not known, the target cannot be accurately modeled, so that the predicted track error is larger. In the aspect of a machine learning model, ma Yong provides a data mining-based precise track prediction method in paper four-dimensional track precise prediction method research, the method comprises the steps of firstly clustering historical tracks, then solving a dense track of each cluster, and combining a hidden Markov model to realize map matching of an air transportation network so as to finish precise track prediction. Such algorithms are more accurate than classical prediction methods, but still suffer from shorter prediction duration.
In conclusion, the existing target track prediction technology has the defects of large error, single prediction model and shorter prediction duration, so that the accuracy of the predicted track is poor, and the flight safety is affected.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides a cycle prediction method based on resampling, so as to reduce the prediction error of a maneuvering target track, increase the prediction duration and improve the track prediction accuracy.
The technical scheme of the invention is that target historical track data is preprocessed; constructing and training a neural network model; generating partial historical track data by using a circulation strategy, and calculating the partial historical track data by using trained network model parameters; and carrying out smooth filtering on the calculation result to obtain a final predicted track.
According to the thought, the track circulation prediction method based on resampling is characterized by comprising the following steps of:
(1) Filtering and normalizing historical track data of the maneuvering target;
(2) Setting a resampling period T' to be 10 times of a system sampling period T, sampling normalized track data into 10 resampling tracks as a network data set, and determining the dimension of an input sample and the dimension of a label sample;
(3) Constructing a neural network which sequentially consists of a Bi-LSTM layer, a Dropout layer, a Dense layer and an activation layer which are two-way long and short time memory units, wherein the neural network consists of four layers of structures;
(4) Setting the maximum iteration times as N and batch size batch_size, sending a network data set into a built network, performing iterative training on parameters of the network by using a batch gradient descent method, and obtaining a trained network model when the iteration times reach N;
(5) Predicting a future track for a period of time:
(5a) Processing the resampled 10 track data by adopting a cyclic strategy to generate 10 batches of data for predicting future tracks;
(5b) Sequentially calculating 10 batches of data processed by the circulation strategy by calling trained neural network parameters to obtain 10 predicted tracks, and combining the 10 tracks into one predicted track according to the sequence of original time slots before resampling;
(6) And smoothing filtering the predicted track by adopting a smooth method to obtain a final predicted result.
Compared with the prior art, the invention has the following advantages:
1) The invention uses the Bi-LSTM layer of the neural network to bidirectionally read the historical track of the target and learn the change characteristics of the historical track, and can still accurately predict the future track when the motion state of the target changes in the future.
2) According to the invention, the resampling strategy is used for resampling the historical track under different time slots, so that the change information of the historical track can be reflected by the data set more uniformly, the prediction error of the neural network prediction track is reduced, and the prediction accuracy is further improved.
3) According to the invention, a circulation strategy is adopted for resampled track data to generate a plurality of batches of data, and the batches of data are predicted in sequence, so that compared with the traditional single prediction method, the data information of the known tracks is utilized to a greater extent, the predicted track with longer prediction duration can be obtained, a more accurate flight strategy is provided for a target, and the flight safety of the target is ensured.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
fig. 2 is a diagram of a neural network model used in the present invention.
Detailed Description
Embodiments and effects of the present invention are described in further detail below with reference to the accompanying drawings.
Referring to fig. 1, the track circulation prediction method based on resampling comprises the following implementation steps:
and step 1, generating a track data set.
Setting the duration of a system to 3000s, setting the sampling period T of the system to 1s, setting the variance of the process evolution noise to 0.01, setting the initial speed of a target x-axis to 200m/s, setting the initial speed of a target y-axis to 200m/s, setting the initial speed of a z-axis to 0m/s, setting the initial coordinate of the x-axis to 15, setting the initial coordinate of the y-axis to 15 and the initial coordinate of the z-axis to 15000m;
the target moves periodically with 600s as a period, and the maneuvering state parameters in a single period are shown in a table 1:
TABLE 1 partial period target maneuver information
Table 1 shows the change of the maneuvering state of the target in the first period (0 to 600 s), the uniform motion of 300s, the left turning motion of 150s and the right turning motion of 150s in a single period. The entire trajectory of the target lasts for 3000s for five cycles. And let the first 2400s of the target track be the known historical track and the last 600s be the unknown future track.
Observing a target through a radar sensor observation platform, wherein the observation sampling frequency is set to be 1s, the radar delay is set to be 1s, observation noise is added based on the angle and the radius of the radar, and the noise size is set to be [0.001 degrees, 100m ];
the motion states of the above objects are fitted using an interactive multi-model algorithm, generating the known track data for the front 2400s and the unknown track data for the rear 600s of the object.
And 2, carrying out Kalman filtering processing on the known track information of 2400s before the target.
(2.1) calculating the state prediction value of the next time k of the targetAnd error covariance P k|k-1 ：
Where k is the discrete time period, F is the target state transition matrix, T is the matrix transpose,for the current time state value, P k-1|k-1 The error covariance of the current moment is represented by Q, and the process noise covariance matrix is represented by Q;
(2.2) calculating the track State update valueAnd error covariance update value P k|k ：
P k|k ＝[I-[P k|k-1 H T (HP k|k-1 H T +R) -1 ]H]P k|k-1 ，
Wherein H is the observation matrix of the track, R is the measurement noise covariance matrix, Z k The radar measurement data at the moment k is obtained, and I is a unit array;
(2.3) repeating (2.1) and (2.2) a total of 2400 times to obtain a filtered track.
Step 3, normalizing the filtered flight path to obtainNormalized track data X s ：
Wherein x is i Is the ith time slot data of the filtered track.
Step 4, normalizing the track data X s The samples are taken into 10 resampled tracks.
The specific implementation of the steps is as follows:
taking the 1 st time slot of the 1 st resampling track as a starting point, and resampling with a resampling interval T' of 10 s;
taking the 2 nd time slot of the original track as a starting point, and resampling with a resampling interval T' of 10 s;
similarly, the 10 th resampling track selects the 10 th time slot of the original track as a starting point, and resampling is carried out at a resampling interval T' of 10 s;
each resampling track comprises 300 sampling points, and the corresponding relation between the time slots of the resampling 10 tracks and the time slots of the original tracks is obtained, as shown in table 2:
table 2 resampling of the track to original track time slot correspondence table
And 5, determining the dimension of the network input sample and the dimension of the label sample.
The specific implementation of the steps is as follows:
determination of network input sample dimensions: firstly, determining that the length W of a sliding window of an input sample is 120, and the number of features is 1; sequentially carrying out sliding segmentation on the resampled 10 flight path data according to W, and splicing segmentation results to obtain network input samples with sample batch of 610 and dimension of 610 x 120 x 1;
determination of label sample dimensions: setting the sample batch and the feature number of the label samples to be the same as those of the input samples, and setting the time step number of each batch of samples to be 60, so as to obtain the label samples with the dimension of 610 x 60 x 1;
resampling track time slots corresponding to input samples and label samples are as in table 3:
table 3 input sample/tag sample and track time slot correspondence table
And 6, constructing a neural network model.
Referring to fig. 2, the neural network model constructed in this step is composed of four layers of structures, i.e., a Bi-LSTM layer, a Dropout layer, a Dense layer, and an activation layer, from top to bottom, and functions and parameters of each layer are as follows:
the Bi-LSTM layer is used for extracting the change characteristics of the historical track data set, and the hidden node number units is 200;
dropout layer for preventing network from over fitting during training, and discarding rate dropout_ratio is 0.2;
the Dense layer is used for fitting a label sample Y_train during network training, and the hidden node number units is 60;
and the activation layer is used for enhancing the adaptability of the network model to nonlinear data, and the activation function is a linear activation function.
And 7, training a neural network model.
The prior method for training the neural network comprises a batch gradient descent method, a random gradient descent method and a mini gradient descent method, wherein the batch gradient descent method is adopted in the step, but is not limited to the implementation method, and the implementation method is as follows:
(7.1) setting the batch size of data as 64, dividing the track data set into a plurality of small batch data according to the batch size of the data, and sequentially sending the small batch data into a neural network for single training;
(7.2) setting a network optimization algorithm as an adaptive moment estimation algorithm Adam, and optimizing network parameters by calculating and correcting a first moment and a second moment of each training gradient;
and (7.3) setting the maximum iteration number of the network as 100, and repeating the steps (7.1) and (7.2) to reach the maximum iteration number to obtain a trained network model.
Step 8, generating 10 batches of data for predicting future tracks.
The specific implementation of the steps is as follows:
(8.1) a duration T predicted from the track requirements pre Calculating a predicted lot N of 10 lots of data according to the resampling period T' and the system sampling period T:
N＝T pre /T/T'，
wherein T in this example pre 600s, 1s for T and 10s for T', the predicted batch N was found to be 60 from the formula;
(8.2) selecting 60 batches of test data for 10 resampled tracks using a cyclic strategy to generate 10 batches of data:
selecting the track data of the 62 th to 181 th time slots of the 1 st resampling track for the 1 st batch of test data of the 1 st batch of data;
selecting the track data of the 63 th to 182 th time slots of the 1 st resampling track for the test data of the 2 nd batch of the 1 st batch of data;
similarly, selecting track data of 121-240 time slots of the 1 st resampling track for the 60 th batch of test data of the 1 st batch of data;
selecting the track data of 62 th to 181 th time slots of the 2 nd resampling track for the 1 st batch of test data of the 2 nd batch of data;
selecting the track data of the 63 th to 182 th time slots of the 2 nd resampling track for the 2 nd batch of test data of the 2 nd batch of data;
and by analogy, selecting track data of 121-240 time slots of the 10 th resampled track for the 60 th batch of test data of the 10 th batch.
The correspondence between 60 batches of test data and 10 resampled track time slots is shown in table 4:
table 4 network input resampled track time slot correspondence table
And 9, sequentially calculating 10 batches of data processed by the circulation strategy by calling trained neural network parameters to obtain 10 predicted tracks.
(9.1) the 60 test data of the 1 st lot obtained in (8.2) were sequentially calculated:
invoking the trained network parameters to calculate the 1 st batch of test data of the 1 st batch of data to obtain a flight path of 182-241 time slots;
invoking the trained network parameters to calculate the 2 nd batch of test data of the 1 st batch of data to obtain a flight path of 183-242 time slots;
and the like, calling the trained network parameters to calculate the 60 th batch of test data of the 1 st batch of data to obtain a track of 241 to 300 time slots,
the 60 batches of predicted results obtained above have a corresponding relationship with the resampled track time slots, as shown in table 5:
table 5 60 lot prediction results and re-sampled track time slot correspondence table
And (9.2) sequentially selecting the last time slot point of 60 batches of predicted results, splicing the last time slot points into a predicted track containing 60 time slot points, namely selecting the last time slot point 241 of the predicted results in the 1 st predicted batch, selecting the last time slot point 242 of the predicted results in the 2 nd predicted batch, and so on, selecting the last time slot point 300 of the predicted results in the 60 th predicted batch, and sequentially splicing the data into track data containing 60 time slot points, namely the predicted track of the 1 st batch of data.
(9.3) predicting the 2 nd to 9 th batches of data by adopting the methods (9.1) and (9.2) to obtain a predicted track of the 2 nd to 9 th batches of data,
and (9.4) combining the 10 predicted tracks obtained in the steps (9.1) - (9.3) into one predicted track according to the original time slot sequence before resampling.
And step 10, smoothing filter processing is carried out on the predicted track by adopting a smooth method, and a final predicted result is obtained.
The formula for smoothing filtering is as follows:
yy(n)＝(y(1)+y(2)+y(3)+...+y(n))/n，
wherein y (n) represents the value before smoothing the nth element, yy (n) represents the value after smoothing the nth element, and the smoothed prediction track is the final prediction result.
The effect of the present invention can be further illustrated by the following simulation experiment.
Simulation conditions:
the neural network model takes a Keras framework of Python3.6 as a simulation platform;
the track generation part and the Kalman filtering part take MATLAB 2018a as a simulation platform;
second, simulation content:
under the above conditions, the present invention and the existing two track prediction algorithms are adopted to respectively predict the future 600s track of the target for a long time, and the prediction results are obtained as shown in table 6:
table 6 comparison of different track prediction algorithm results
As can be seen from Table 6, compared with the existing algorithm, the resampling-based track cycle prediction method provided by the invention has lower error in track prediction, longer prediction duration, and can provide a more accurate flight strategy for the target, thereby ensuring the flight safety of the target.
The above description is only one specific example of the invention and does not constitute any limitation of the invention, it will be apparent to those skilled in the art that various modifications and changes in form and details may be made without departing from the principle and construction idea of the invention, but these modifications and changes based on the idea of the invention remain within the scope of the claims of the invention.
Claims (10)
1. The track circulation prediction method based on resampling is characterized by comprising the following steps of:
(1) Filtering and normalizing historical track data of the maneuvering target;
(2) Setting a resampling period T' to be 10 times of a system sampling period T, sampling normalized track data into 10 resampling tracks as a network data set, and determining the dimension of an input sample and the dimension of a label sample;
(3) Constructing a neural network which sequentially consists of a Bi-LSTM layer, a Dropout layer, a Dense layer and an activation layer which are two-way long and short time memory units, wherein the neural network consists of four layers of structures;
(4) Setting the maximum iteration times as N and batch size batch_size, sending a network data set into a built network, performing iterative training on parameters of the network by using a batch gradient descent method, and obtaining a trained network model when the iteration times reach N;
(5) Predicting a future track for a period of time:
(5a) Processing the resampled 10 track data by adopting a cyclic strategy to generate 10 batches of data for predicting future tracks;
(5b) Sequentially calculating 10 batches of data processed by the circulation strategy by calling trained neural network parameters to obtain 10 predicted tracks, and combining the 10 predicted tracks into one predicted track according to the time slot sequence of the original track before resampling;
(6) And smoothing filtering the predicted track by adopting a smooth method to obtain a final predicted result.
2. The method of claim 1, wherein (1) filtering the track data is performed by performing state prediction and state update on the historical track data using a kalman filter algorithm to eliminate system measurement noise, which is implemented as follows:
(1a) Calculating state predictive value of next moment k of targetAnd error covariance P k|k-1 ：
Wherein k is a discrete time period, F is a target state transition matrix, and Q is system process noise;
(1b) Calculating track state update valuesAnd error covariance update value P k|k ：
P k|k ＝[I-G k H k ]P k|k-1
Wherein H is the observation matrix of the track, R k The method is characterized in that the method is a noise covariance matrix, T is a matrix transposition, and Z is radar measurement data;
(1c) And (3) performing iterative computation on the steps (1 a) and (1 b), wherein the iterative times are the time slot points of the historical tracks, and obtaining the filtered target tracks.
3. The method of claim 1 wherein the track data is normalized in (1) as follows:
wherein x is i For the ith time slot data of the filtered track, X scaled Is normalized data.
4. The method of claim 1, wherein (2) sampling the normalized track data into 10 resampled tracks is accomplished by:
the 1 st resampling track selects the 1 st time slot of the original track as a starting point, and resamples with a sampling interval of T';
the 2 nd resampling track selects the 2 nd time slot of the original track as a starting point, and resampling with a sampling interval of T' is carried out;
similarly, the 10 th resampling track selects the 10 th time slot of the original track as a starting point, resampling is carried out at a sampling interval of T', and the original track is changed into 10 resampling tracks.
5. The method of claim 1, wherein determining network input sample dimensions and label sample dimensions in (2) is accomplished by:
for the determination of the dimension of the network input samples, firstly, determining that the length W of a sliding window of the input samples is 120, the number of features is 1, then sequentially carrying out sliding segmentation on 10 resampled track data according to W, and splicing segmentation results to obtain network input samples with the sample batch of 610 and the dimension of 610 x 120 x 1;
for determining the dimension of the label sample, the sample batch and the feature number of the label sample are the same as those of the input sample, and the time step number of each batch of sample is set to be 60, so that the label sample with the dimension of 610 x 60 x 1 can be obtained.
6. The method of claim 1, wherein the neural network constructed in (3) functions and parameters as follows:
the Bi-LSTM layer is used for extracting the change characteristics of the historical track data set, and the hidden node number units is 200;
dropout layer for preventing network from over fitting during training, and discarding rate dropout_ratio is 0.2;
the Dense layer is used for fitting a label sample Y_train during network training, and the hidden node number units is 60;
and the activation layer is used for enhancing the adaptability of the network model to nonlinear data, and the activation function is a linear activation function.
7. The method of claim 1, wherein (4) iteratively training parameters of the neural network using a small batch gradient descent method is performed as follows:
(4a) Setting the batch size of data as 64, dividing a track data set into a plurality of small batch data according to the batch size of the data, and sequentially sending the small batch data into a neural network for single training;
(4b) Setting a network optimization algorithm as a self-adaptive moment estimation algorithm Adam, and optimizing network parameters by calculating and correcting a first moment and a second moment of each training gradient;
(4c) Setting the maximum iteration number of the network as 100, and repeating the steps (4 a) and (4 b) 100 times to obtain a trained network model.
8. The method of claim 1, wherein (5 a) generating 10 batches of data for predicting future tracks is accomplished by:
(5a1) Duration T predicted from track needs pre Calculating a predicted lot N for each of the 10 lots of data, a resampling period T' and a system sampling period T:
N＝T pre /T/T'；
(5a2) And selecting N batches of network inputs by adopting a cyclic strategy on the 10 resampling tracks to generate 10 batches of data:
selecting the track data of the 62 th to 181 th time slots of the 1 st resampling track for the 1 st batch of test data of the 1 st batch of data;
selecting the track data of the 63 th to 182 th time slots of the 1 st resampling track for the test data of the 2 nd batch of the 1 st batch of data;
similarly, the test data of the N batch of the 1 st batch of data is selected, and track data of 121-240 time slots of the 1 st resampling track are selected;
selecting the track data of 62 th to 181 th time slots of the 2 nd resampling track for the 1 st batch of test data of the 2 nd batch of data;
selecting the track data of the 63 th to 182 th time slots of the 2 nd resampling track for the 2 nd batch of test data of the 2 nd batch of data;
similarly, the test data of the 10 th batch and the N th batch are selected, and track data of 121-240 time slots of the 10 th resampled track are selected.
9. The method of claim 1, wherein (5 b) sequentially calculating 10 batches of data by invoking trained neural network parameters is performed as follows:
(5b1) Invoking the trained network parameters to sequentially calculate N batches of test data of the 1 st batch of data obtained in the step (5 a) to obtain N batches of prediction results;
(5b2) Sequentially selecting the last time slot point of N batches of predicted results, and splicing the last time slot point into a predicted track containing N time slot points, namely, the predicted track of the 1 st batch of data;
(5b3) And (3) predicting the 2 nd to 9 th batches of data by adopting the methods (5 b 1) and (5 b 2) to obtain the predicted tracks of the 2 nd to 9 th batches of data, and combining 10 predicted tracks into one predicted track according to the original time slot sequence before resampling.
10. The method of claim 1 wherein (6) smoothing the predicted track using a smooth method is formulated as follows:
yy(n)＝(y(1)+y(2)+y(3)+...+y(n))/n
wherein y (n) represents the value before smoothing the nth element, yy (n) represents the value after smoothing the nth element, and the smoothed prediction track is the final prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110658739.1A CN113393032B (en) | 2021-06-15 | 2021-06-15 | Track circulation prediction method based on resampling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110658739.1A CN113393032B (en) | 2021-06-15 | 2021-06-15 | Track circulation prediction method based on resampling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113393032A CN113393032A (en) | 2021-09-14 |
| CN113393032B true CN113393032B (en) | 2023-09-12 |
Family
ID=77620808
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110658739.1A Active CN113393032B (en) | 2021-06-15 | 2021-06-15 | Track circulation prediction method based on resampling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113393032B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114048889B (en) * | 2021-10-08 | 2022-09-06 | 天津大学 | Aircraft trajectory prediction method based on long-term and short-term memory network |
| CN114462293B (en) * | 2021-10-27 | 2023-10-13 | 中国空气动力研究与发展中心计算空气动力研究所 | Hypersonic speed target medium-long term track prediction method |
| CN114999233B (en) * | 2022-08-05 | 2022-11-01 | 中国航天科工集团八五一一研究所 | Target intention judgment method based on track association |
| CN116976956A (en) * | 2023-09-22 | 2023-10-31 | 通用技术集团机床工程研究院有限公司 | CRM system business opportunity deal prediction method, device, equipment and storage medium |
| CN117648873B (en) * | 2024-01-30 | 2024-06-04 | 国网天津市电力公司电力科学研究院 | Ground subsidence prediction method, training method, device and equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003130948A (en) * | 2001-10-25 | 2003-05-08 | Mitsubishi Electric Corp | Target tracking apparatus |
| CN108254741A (en) * | 2018-01-16 | 2018-07-06 | 中国人民解放军海军航空大学 | Targetpath Forecasting Methodology based on Recognition with Recurrent Neural Network |
| CN109508812A (en) * | 2018-10-09 | 2019-03-22 | 南京航空航天大学 | A kind of aircraft Trajectory Prediction method based on profound memory network |
| CA3067573A1 (en) * | 2019-01-14 | 2020-07-14 | Harbin Engineering University | Target tracking systems and methods for uuv |
-
2021
- 2021-06-15 CN CN202110658739.1A patent/CN113393032B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003130948A (en) * | 2001-10-25 | 2003-05-08 | Mitsubishi Electric Corp | Target tracking apparatus |
| CN108254741A (en) * | 2018-01-16 | 2018-07-06 | 中国人民解放军海军航空大学 | Targetpath Forecasting Methodology based on Recognition with Recurrent Neural Network |
| CN109508812A (en) * | 2018-10-09 | 2019-03-22 | 南京航空航天大学 | A kind of aircraft Trajectory Prediction method based on profound memory network |
| CA3067573A1 (en) * | 2019-01-14 | 2020-07-14 | Harbin Engineering University | Target tracking systems and methods for uuv |
Non-Patent Citations (1)
| Title |
|---|
| Emergency-Oriented Spatiotemporal Trajectory Pattern Recognition by Intelligent Sensor Devices;Ping Zhang;IEEE;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113393032A (en) | 2021-09-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113393032B (en) | Track circulation prediction method based on resampling | |
| CN109034376B (en) | L STM-based unmanned aerial vehicle flight state prediction method and system | |
| CN109508812B (en) | Aircraft track prediction method based on deep memory network | |
| CN109146162B (en) | A kind of probability wind speed forecasting method based on integrated Recognition with Recurrent Neural Network | |
| CN114241282A (en) | Knowledge distillation-based edge equipment scene identification method and device | |
| CN113110592A (en) | Unmanned aerial vehicle obstacle avoidance and path planning method | |
| CN108596327B (en) | Seismic velocity spectrum artificial intelligence picking method based on deep learning | |
| CN111897353B (en) | Aircraft maneuvering trajectory prediction method based on GRU | |
| CN113408392B (en) | Flight path completion method based on Kalman filtering and neural network | |
| CN112906858A (en) | Real-time prediction method for ship motion trail | |
| CN111476302A (en) | fast-RCNN target object detection method based on deep reinforcement learning | |
| CN110728698B (en) | Multi-target tracking system based on composite cyclic neural network system | |
| CN110068330B (en) | Autonomous positioning method of robot based on ARMA model | |
| CN111461292A (en) | Real-time trajectory prediction method for unmanned aerial vehicle | |
| CN113341919A (en) | Computing system fault prediction method based on time sequence data length optimization | |
| CN111144462B (en) | Unknown individual identification method and device for radar signals | |
| CN114548591A (en) | Time sequence data prediction method and system based on hybrid deep learning model and Stacking | |
| CN112348223A (en) | Missile flight trajectory prediction method based on deep learning | |
| CN113064422B (en) | Autonomous underwater vehicle path planning method based on double neural network reinforcement learning | |
| Chen et al. | Application of Data‐Driven Iterative Learning Algorithm in Transmission Line Defect Detection | |
| CN114972429B (en) | Target tracking method and system for cloud edge cooperative self-adaptive reasoning path planning | |
| CN115056829A (en) | Train motion state estimation method for multi-vehicle type continuous learning | |
| CN115512214A (en) | Indoor visual navigation method based on causal attention | |
| CN115034432A (en) | Wind speed prediction method for wind generating set of wind power plant | |
| CN111160464B (en) | Industrial high-order dynamic process soft measurement method based on multi-hidden-layer weighted dynamic model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |