CN112465113A - Generating type interrupted track continuing correlation method - Google Patents

Generating type interrupted track continuing correlation method Download PDF

Info

Publication number
CN112465113A
CN112465113A CN202011331943.4A CN202011331943A CN112465113A CN 112465113 A CN112465113 A CN 112465113A CN 202011331943 A CN202011331943 A CN 202011331943A CN 112465113 A CN112465113 A CN 112465113A
Authority
CN
China
Prior art keywords
loss
track
discriminator
network
generator
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.)
Pending
Application number
CN202011331943.4A
Other languages
Chinese (zh)
Inventor
熊伟
徐平亮
崔亚奇
刘克
宋伟健
龚诚
郝延彪
顾祥岐
熊振宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Naval Aeronautical University
Original Assignee
Naval Aeronautical University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Naval Aeronautical University filed Critical Naval Aeronautical University
Priority to CN202011331943.4A priority Critical patent/CN112465113A/en
Publication of CN112465113A publication Critical patent/CN112465113A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/66—Radar-tracking systems; Analogous systems
    • G01S13/72—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • 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
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Abstract

The invention discloses a generating type interrupted track connection correlation method, which mainly solves the problems that the existing connection correlation model needs a large amount of prior information in practical engineering application, has long judgment time and is difficult to directly apply. The method firstly designs a generation countermeasure network model for interrupting track connection association, comprises a generator and a discriminator and trains the model by utilizing mass track data in a training data set. After the training is finished, only the generator needs to be extracted, the generator can be used for finishing the continuous association task of the interrupted flight path, and a continuous association flight path situation map is generated. By comparing with the track situation map in the test data set, parameter tuning of the model can be performed to obtain an optimal model. The method can automatically train and generate the interrupted track continuing correlation model based on the measured data, has the advantages of high correlation speed, good practical effect and the like, and can be directly used for solving the problem of the interrupted track continuing correlation of the same information source in the actual engineering.

Description

Generating type interrupted track continuing correlation method
Technical Field
The invention belongs to the field of radar data processing, relates to generation and reasoning of an interrupted track continuation correlation model, and is suitable for a radar target detection and tracking system and an information processing system.
Background
In the process of tracking a target by a radar, the problem of target track interruption often occurs under the influence of various factors such as target stop and the like, target maneuvering, platform maneuvering, long sampling interval, low detection probability, ground object shielding, electromagnetic interference and the like, namely that the current track of the target suddenly disappears temporarily, and a new track is tracked again in an adjacent area after a period of time. The track interruption can cause the radar to successively generate a plurality of tracks with different lot numbers on the same target, although the number of the real-time tracks is not obviously influenced, the proportion of the newly generated target tracks with unknown attributes is increased, and then the target attribute analysis and judgment task of information personnel is aggravated, and the working efficiency of the information personnel is influenced. Particularly, under the dense target environment, a large number of flight paths are temporarily suspended and restarted, so that local situation confusion is easily caused, and situation analysis and judgment of information personnel are seriously interfered. In order to make the situation clear and accurate and reduce unnecessary workload of information staff, it is necessary to perform continuous association on interrupted tracks so as to realize batch number unification among different track sections of the same target. The existing method for associating the interrupted track with the continuous flight path is obtained by adopting various field theoretical achievements such as fuzzy, evidence, probability statistics, probability estimation and the like and combining self heuristic consideration and deducing under the assumption of different target motion models and measurement noise of researchers, has the problems of unreasonable prior assumption, inapplicability of models, incapability of determining thresholds and the like in practical application, and needs to manually modify and debug the adopted models and threshold parameters by using measured data. In addition, a large amount of complex operations need to be carried out on the original track data in the association process, the time for interrupting track association is long, and real-time processing is difficult. Under the condition of lacking standardization and automation debugging steps and methods, the debugging of the interrupted track connection association model consumes a great deal of time and energy, and the final connection effect after debugging is difficult to achieve the optimal effect, so that the existing model method cannot directly solve the problem of interrupted track connection association in actual engineering, and has a large difference from the actual requirement.
With the increasing development of artificial intelligence, deep learning technology plays an extremely important role in various fields. Such as image retrieval, radar target detection, medical image processing, automated driving, etc. Particularly in the field of computer vision, the generation of the countermeasure network based on the deep learning technology can learn the characteristics of the image through self-countermeasure training, and finally can generate a false and true picture. The invention provides a generation type interruption track connection association method inspired by generation of a countermeasure network by deep learning, aiming at solving the problems that the existing interruption track connection association model method is narrow in application range, needs to manually use measured data to repeatedly modify and debug, and is difficult to ensure the final effect in practical application by utilizing the strong self-learning capacity of deep learning and enabling the network to learn track interruption characteristics and carry out connection association by using any method through countermeasures of a generator and a discriminator in the generation countermeasure network.
Disclosure of Invention
The invention relates to a generating type interrupted track continuing correlation method, which specifically comprises the following technical measures: firstly, a track conversion module is constructed, track vectors are normalized and are converted into a track situation map in a blank map in a one-to-one correspondence mode with quantization grids. And then, in order to enable the generation countermeasure network to learn the probability distribution of the track situation map in the countermeasure process, a countermeasure loss function is designed. The loss function includes two parts: the loss is discriminated and the loss is generated. And judging loss and selecting mean square error loss, comparing the difference between the judgment result of the discriminator and the real label to generate loss selection absolute value loss, and comparing the difference between the generated track situation map and the real track situation map. And then constructing a deep learning generation countermeasure network model, wherein a generation countermeasure network comprises a generator and a discriminator, the discriminator is used for judging whether the track situation graph after continuous correlation is true or false and providing a gradient updating direction for the generator, and the generator is used for generating the continuous correlation track situation graph which is false or true as far as possible so as to achieve the purpose of deceiving the discriminator. The generation of the antagonistic network parameter updates adopts gradient back propagation and random gradient descent algorithms. And training the generated interrupted track continuous association model by using the training data set, and only extracting the generator after the training is finished, so that the generator can complete the continuous association task of the interrupted track, and a continuous association track situation map is generated. And then testing the generator by using the test data set, directly storing the interrupted track continuous association generator for the interrupted track continuous association of the same information source if the effect of the continuous associated track situation diagram generated by current training meets the requirement, and considering that the generated countermeasure network structure is changed or the super parameters such as different learning rates are selected if the effect is not satisfied, and then re-performing the training test of the network.
The invention provides a generating type interrupted flight path continuing association method, which can automatically learn hidden variables in flight path parameters based on measured data, train and generate an interrupted flight path continuing association model, namely a generator, completely avoids a large amount of debugging operations such as manual selection of a motion model, setting of target motion parameters, acquisition of target motion prior information and the like, has the advantages of high association speed, good practical effect, manpower and material resource saving and the like, and the generator after training can be directly used for solving the problem of the same information source interrupted flight path continuing association in actual engineering.
Drawings
FIG. 1 is a diagram of a generator network architecture;
FIG. 2 is a diagram of a discriminator network architecture;
fig. 3 is a flow chart of a generative interrupt track connection association network training.
Detailed Description
The invention provides a generating type interrupted track continuing correlation method, which comprises the following steps:
step 1: converting the original track vector into a standard and dimensionless track vector through normalization, then selecting the grid size of the track situation map, corresponding the grid quantized to the normalized track vector one by one, wherein the normalized track vector corresponding to the quantized grid is the track situation map;
step 1.1: firstly, traversing all track points, finding a maximum value point and a minimum value point of a track coordinate in a current scene, and then subtracting the minimum value point from each track point and dividing the minimum value point by the maximum value point to obtain a standard dimensionless normalized track vector;
step 1.2: setting the size of blank map as M, M is the length of map and is unit as meter, dividing unit length by M to make grid quantization, i.e. using
Figure BDA0002796066860000031
Expressing the normalized track length represented by each pixel in the quantized grid, and corresponding the normalized track coordinates to the quantized grid coordinates one by one to obtain a track situation map;
step 2: designing a loss function, wherein the loss function for generating the confrontation network comprises two parts, namely discrimination loss and generation loss, and the two kinds of loss are alternately propagated in opposite directions to enable the network to achieve the target of confrontation learning;
step 2.1: judging loss and selecting mean square error loss, comparing the difference between the judgment result of the discriminator and the real label, wherein the label 1 represents that the track situation map is true, the label 0 represents that the track situation map is false, when the discriminator is trained, the label is 1 when the track situation map is interrupted and the real track situation map is matched, the label is 0 when the track situation map is generated, when the discriminator is trained, the label is 1 when the track situation map is interrupted and the track situation map is generated, and the judgment loss is,
Figure BDA0002796066860000032
wherein LossDTo discriminate the loss, |DAnd lRRespectively judging results and real labels of the judgers;
step 2.2: the generation loss comprises L1 generation loss and decision loss, L1 generation loss selects absolute value loss for comparing the difference between the generated track situation map and the real track situation map, L1 generation loss is,
LossL1＝|MG-MR|
wherein LossL1GTo generate losses, MGAnd MRRespectively generating a track situation map and a real track situation map;
step 2.3: the total generation penalty is a weighted sum of the L1 generation penalty and the decision penalty, and since training of the generation net is relatively difficult, the L1 generation penalty is given a greater weight, and the total penalty function is,
LossG＝λG×LossL1+λD×LossD
wherein λGAnd λDThe weights for generating loss and discriminating loss are respectively 1 and 10 in the invention;
and step 3: adopting a convolutional neural network and a residual error network to construct a generation countermeasure network, wherein the generation countermeasure network comprises a generator and a discriminator, judging whether a continuous associated track situation graph is true or false by using the discriminator, providing a gradient updating direction for the generator, and generating a continuous associated track situation graph which can be false or true by using the generator so as to achieve the purpose of deceiving the discriminator;
step 3.1: the down-sampling network comprises a convolution layer, a normalization layer and a nonlinear activation layer, aiming at roughly extracting flight path information and interruption information, the residual network consists of N residual blocks, N is a hyper-parameter, each residual block consists of the convolution layer, the normalization layer and the nonlinear activation layer which are connected in series, the up-sampling network comprises a reverse convolution layer, a normalization layer and a nonlinear activation layer which are reverse transformation of down-sampling, the output network compresses the number of channels through the convolution layer, visualizing a track profile;
step 3.2: constructing a discriminator by utilizing P convolution modules, wherein P is a hyper-parameter, each convolution module comprises a convolution layer, a normalization layer and a nonlinear activation layer, the number of output channels of the discriminator is 1, and the discriminator corresponds to the truth of a track situation map;
and 4, step 4: training by utilizing mass track data in a training set to generate a confrontation network, training and updating the network parameters of the discriminator by adopting a random gradient descent error back propagation algorithm, then training and updating the network parameters of the generator, and finally enabling the generator and the discriminator to reach Nash balance to finish training;
step 4.1: generating a continuous track situation map by using a generator, connecting the paired interrupted track situation map and a real track situation map, inputting the connected interrupted track situation map and the real track situation map into a discriminator, setting a label as 1, and outputting a real Loss by the discriminatorD1Connecting the paired interrupted track situation map and the generated continuous track situation map, inputting the connected interrupted track situation map and the generated continuous track situation map into a discriminator, setting a label as 0, and outputting the false Loss by the discriminatorD0The loss of the discriminator is,
LossD＝(LossD1+LossD0)×0.5
updating the network parameters of the discriminator by using the discriminator loss;
step 4.2: training the generator by using the updated arbiter to generate a continuous track situation map, and matching the matched interrupted track situation map with the generated continuous track situation mapInputting the continuous flight situation map after being connected into a discriminator, setting a label as 1, deceiving the discriminator, and outputting real Loss by the discriminatorD1Comparing the generated track situation map with the real track situation map by using the L1 generated Loss function to obtain the L1 generated LossL1GJudging the weighted sum of the loss and the L1 generation loss to be the generation loss, and updating the network parameters of the generator by using the generation loss;
step 4.3: after training in a plurality of periods, the generator and the discriminator reach Nash balance, the generator is taken out at the moment, and the trained generator can be used for carrying out the association task of track interruption and continuation;
and 5: and testing the generator by using the test data set, if the difference between the continuous associated track situation diagram generated by the current training and the real track situation diagram is not large and the association effect meets the requirement, directly storing the interrupted track continuous associated generator for the interrupted track continuous association of the same information source, and if the result is not satisfied, considering that the generated countermeasure network structure is changed or selecting different learning rates and other super parameters, and re-performing the training test of the network.

Claims (6)

1. A generation-type interrupted track continuing associated learning method is characterized by comprising the following steps:
step 1, converting an original track vector into a standard and dimensionless track vector through normalization, then selecting the grid size of a track situation map, corresponding the grid quantized to the normalized track vector one by one, wherein the normalized track vector corresponding to the quantized grid is the track situation map;
step 2, designing a loss function, wherein the loss function for generating the confrontation network comprises two parts, namely discrimination loss and generation loss, and the two losses are alternately propagated in opposite directions to enable the network to achieve the target of confrontation learning;
step 3, constructing a generated countermeasure network by adopting a convolutional neural network and a residual error network, wherein the generated countermeasure network comprises a generator and a discriminator, the discriminator is used for judging whether the track situation graph after continuous correlation is true or false, a gradient updating direction is provided for the generator, and the generator is used for generating a continuous correlation track situation graph which can be false or true, so that the purpose of deceiving the discriminator is achieved;
step 4, training and generating a confrontation network by using mass flight path data in a training set, training and updating the network parameters of the discriminator by adopting a random gradient descent error back propagation algorithm, training and updating the network parameters of the generator, and finally enabling the generator and the discriminator to reach Nash balance to finish training;
and 5, testing the generator by using the test data set, directly storing the interrupted track continuous association generator for the interrupted track continuous association of the same information source if the difference between the continuous association track situation diagram generated by the current training and the real track situation diagram is not large and the association effect meets the requirement, and considering to change the generated confrontation network structure or select different learning rates and carrying out the training test of the network again if the interrupted track continuous association generator is unsatisfied.
2. The method for associating a generated outage trajectory continuation as claimed in claim 1, wherein the step 1 specifically includes the following sub-steps:
step 1.1, firstly traversing all track points, finding a maximum value point and a minimum value point of a track coordinate in a current scene, and then subtracting the minimum value point from each track point and dividing the minimum value point by the maximum value point to obtain a standard dimensionless normalized track vector;
step 1.2, setting the size of the blank map as M multiplied by M, wherein M is the length of the map and the unit is meter, and dividing the unit length 1 by M to carry out grid quantization, namely
Figure FDA0002796066850000011
And expressing the normalized track length represented by each pixel in the quantized grid, and corresponding the normalized track coordinates to the quantized grid coordinates one by one to obtain a track situation map.
3. The method as claimed in claim 1, wherein the step 2 specifically comprises the following sub-steps:
step 2.1, judging loss and selecting mean square error loss, comparing the difference between the judgment result of the discriminator and a real label, wherein the label 1 represents that a track situation diagram is true, the label 0 represents that the track situation diagram is false, when the discriminator is trained, the label for matching the interrupted track situation diagram with the real track situation diagram is 1, the label for matching with the generated track situation diagram is 0, when the generator is trained, the label for matching the interrupted track situation diagram with the generated track situation diagram is 1, judging loss is,
Figure FDA0002796066850000021
wherein LossDTo discriminate the loss, |DAnd lRRespectively judging results and real labels of the judgers;
step 2.2, the generation loss comprises L1 generation loss and decision loss, L1 generation loss selects absolute value loss for comparing the difference between the generated track situation map and the real track situation map, L1 generation loss is,
LossL1G＝|MG-MR|
wherein LossL1GTo generate losses, MGAnd MRRespectively generating a track situation map and a real track situation map;
step 2.3, the total generation loss is a weighted sum of the L1 generation loss and the decision loss, and since training of the generation network is relatively difficult, the L1 generation loss is given a greater weight, and the total loss function is,
LossG＝λG×LossL1G+λD×LossD
wherein LossGFor total loss, λGAnd λDRespectively, are the weights for generating losses and discriminating losses.
4. A method as claimed in claim 3, wherein λ isGAnd λDAre specifically 1 and 10.
5. The method as claimed in claim 1, wherein the step 3 specifically comprises the following sub-steps:
step 3.1, constructing a generator by utilizing an input preprocessing network, a down-sampling network, a residual error network, an up-sampling network and an output network, wherein the down-sampling network, the residual error network and the up-sampling network are of an automatic coding-decoder structure, the input preprocessing network expands the number of picture channels through a convolution layer to be suitable for the processing of the automatic coding-decoder, the down-sampling network comprises a convolution layer, a normalization layer and a nonlinear activation layer and aims at roughly extracting flight path information and interruption information, the residual error network comprises N residual error blocks, N is a hyper-parameter, each residual error block comprises a convolution layer, a normalization layer and a nonlinear activation layer which are connected in series, the up-sampling network comprises an anti-convolution layer, a normalization layer and a nonlinear activation layer and is the down-sampling inverse transformation, the output network compresses the number of the channels through the convolution layer, visualizing a track profile;
and 3.2, constructing a discriminator by utilizing P convolution modules, wherein P is a hyper-parameter, each convolution module comprises a convolution layer, a normalization layer and a nonlinear activation layer, the number of output channels of the discriminator is 1, and the discriminator corresponds to the truth of the track situation map.
6. The method as claimed in claim 1, wherein the step 4 specifically includes the following sub-steps:
step 4.1, generating a continuous track situation map by using a generator, connecting the paired interrupted track situation map and the real track situation map, inputting the connected interrupted track situation map and the real track situation map into a discriminator, setting a label as 1, and outputting real Loss by the discriminatorD1Connecting the paired interrupted track situation map and the generated continuous track situation map, inputting the connected interrupted track situation map and the generated continuous track situation map into a discriminator, setting a label as 0, and outputting the false Loss by the discriminatorD0The loss of the discriminator is,
LossD＝(LossD1+LossD0)×0.5
updating the network parameters of the discriminator by using the discriminator loss;
step 4.2, training the generator is guided by the updated discriminator to generate a continuous track situation map, the paired interrupted track situation map and the generated continuous track situation map are input into the discriminator after being connected, a label is set to be 1, so that the discriminator is deceived to output a real LossD1Comparing the generated track situation map with the real track situation map by using the L1 generated Loss function to obtain the L1 generated LossL1GJudging the weighted sum of the loss and the L1 generation loss to be the generation loss, and updating the network parameters of the generator by using the generation loss;
and 4.3, after training of a plurality of periods, the generator and the discriminator reach Nash balance, the generator is taken out at the moment, and the trained generator can be used for carrying out the association task of the interrupted flight path continuation.
CN202011331943.4A 2020-11-24 2020-11-24 Generating type interrupted track continuing correlation method Pending CN112465113A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011331943.4A CN112465113A (en) 2020-11-24 2020-11-24 Generating type interrupted track continuing correlation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011331943.4A CN112465113A (en) 2020-11-24 2020-11-24 Generating type interrupted track continuing correlation method

Publications (1)

Publication Number Publication Date
CN112465113A true CN112465113A (en) 2021-03-09

Family

ID=74798795

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011331943.4A Pending CN112465113A (en) 2020-11-24 2020-11-24 Generating type interrupted track continuing correlation method

Country Status (1)

Country Link
CN (1) CN112465113A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115048371A (en) * 2022-08-16 2022-09-13 北京理工大学 Fuzzy evidence track association method based on data cleaning and generation of countermeasure network

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2000017103A2 (en) * 1998-09-18 2000-03-30 Massachusetts Institute Of Technology Inventory control
US20100253489A1 (en) * 2009-04-02 2010-10-07 Gm Global Technology Operations, Inc. Distortion and perspective correction of vector projection display
CA2953385A1 (en) * 2014-06-30 2016-01-07 Evolving Machine Intelligence Pty Ltd A system and method for modelling system behaviour
CN111930110A (en) * 2020-06-01 2020-11-13 西安理工大学 Intent track prediction method for generating confrontation network by combining society
CN111931902A (en) * 2020-07-03 2020-11-13 江苏大学 Countermeasure network generation model and vehicle track prediction method using the same

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2000017103A2 (en) * 1998-09-18 2000-03-30 Massachusetts Institute Of Technology Inventory control
US20100253489A1 (en) * 2009-04-02 2010-10-07 Gm Global Technology Operations, Inc. Distortion and perspective correction of vector projection display
CA2953385A1 (en) * 2014-06-30 2016-01-07 Evolving Machine Intelligence Pty Ltd A system and method for modelling system behaviour
CN111930110A (en) * 2020-06-01 2020-11-13 西安理工大学 Intent track prediction method for generating confrontation network by combining society
CN111931902A (en) * 2020-07-03 2020-11-13 江苏大学 Countermeasure network generation model and vehicle track prediction method using the same

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ADAM HOUENOU, ET AL: "A track-to-track association method for automotive perception systems", 《2012 IEEE INTELLIGENT VEHICLES SYMPOSIUM》 *
郑浩，等: "基于拓扑结构的多元特征航迹关联算法", 《指挥信息***与技术》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115048371A (en) * 2022-08-16 2022-09-13 北京理工大学 Fuzzy evidence track association method based on data cleaning and generation of countermeasure network

Similar Documents

Publication Publication Date Title
CN109508360B (en) Geographical multivariate stream data space-time autocorrelation analysis method based on cellular automaton
CN107506444B (en) Machine learning system associated with interrupted track connection
Wu et al. Long-term 4D trajectory prediction using generative adversarial networks
CN113435644A (en) Emergency prediction method based on deep bidirectional long-short term memory neural network
CN109886356A (en) A kind of target tracking method based on three branch's neural networks
Xie et al. Dual-channel and bidirectional neural network for hypersonic glide vehicle trajectory prediction
CN114819054A (en) Power electronic system state monitoring method based on physical information neural network
CN111898746B (en) Deep learning method for continuous relevance of broken flight path
CN103296995B (en) Any dimension high-order (>=4 rank) tasteless conversion and Unscented Kalman Filter method
CN112465113A (en) Generating type interrupted track continuing correlation method
Berneti Design of fuzzy subtractive clustering model using particle swarm optimization for the permeability prediction of the reservoir
CN117313795A (en) Intelligent building energy consumption prediction method based on improved DBO-LSTM
CN109583751A (en) The failure decision-making technique of payload
CN112613542A (en) Bidirectional LSTM-based enterprise decontamination equipment load identification method
CN115359197A (en) Geological curved surface reconstruction method based on spatial autocorrelation neural network
Xu et al. Probabilistic human motion prediction via a Bayesian neural network
CN114510961A (en) Ship behavior intelligent monitoring algorithm based on recurrent neural network and Beidou positioning
Shen et al. An interval analysis scheme based on empirical error and mcmc to quantify uncertainty of wind speed
Zhang et al. Research on intelligent health management technology of opto-electronic equipment
He et al. An Automatic Reflective Clothing Detection Algorithm Based on YOLOv5 for Work Type Recognition
Yan Application of Image Segmenting Technology Based on Fuzzy C-means Algorithm in Competition Video Referee
CN111523090B (en) Number time-varying multi-target tracking method based on Gaussian mixture probability hypothesis density
Shang BLE Complex Environment Location Method and System Based on Machine Learning Algorithm
CN117150738B (en) Action direction pre-judging method under complex scene
Bi A method for modulation recognition of maritime search and rescue communication signal based on neural network

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
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210309

WD01 Invention patent application deemed withdrawn after publication