CN110059144B - Trajectory owner prediction method based on convolutional neural network - Google Patents
Trajectory owner prediction method based on convolutional neural network Download PDFInfo
- Publication number
- CN110059144B CN110059144B CN201910266737.0A CN201910266737A CN110059144B CN 110059144 B CN110059144 B CN 110059144B CN 201910266737 A CN201910266737 A CN 201910266737A CN 110059144 B CN110059144 B CN 110059144B
- Authority
- CN
- China
- Prior art keywords
- track
- user
- owner
- trajectory
- layer
- 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.)
- Expired - Fee Related
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- 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
- 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)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Remote Sensing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a trajectory owner prediction method based on a convolutional neural network, wherein a directed weightless graph G (V, E) is formed according to the trajectories of all users, and a low-dimensional real value vector of a trajectory position ID is learned through Node2 Vec; then, slicing the user track, replacing the position ID with a low-dimensional real value vector corresponding to the position ID for the cut indefinite-length track, and intercepting or filling to form a fixed dimension matrix of the track; and then, constructing and training a four-layer convolutional neural network as a prediction model, inputting a track matrix constructed by the latitude and longitude of the position of the user to be detected into the trained prediction model to obtain probability distribution of track owner classification, and finally marking the index of the maximum value in the probability distribution as the number of the track corresponding to the owner.
Description
Technical Field
The invention belongs to the technical field of machine learning, and particularly relates to a trajectory owner prediction method based on a convolutional neural network.
Background
The track owner prediction comprises the steps of extracting and analyzing the characteristics of a track of an unknown owner, and then judging the owner of the track. The trajectory owner prediction is the basis of a plurality of location-based services, has important significance for improving the quality of the location-based services, and a service provider can perform personalized recommendation, preference-based path planning and the like by using the prediction result.
In the conventional trajectory owner prediction method, a trajectory is generally processed as a time series, and then the representation of the time series is learned by using an RNN or the like. Although the method learns the context of the trajectory, in a trajectory sequence, a certain specific position or a combination of certain specific positions is critical to the classification of the trajectory, the features of the trajectory cannot be effectively captured by the existing method, and the features can be better learned by the convolutional neural network, so that a trajectory owner prediction method based on the convolutional neural network is provided.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a trajectory owner prediction method based on a convolutional neural network, and the accuracy of trajectory prediction is improved by improving a trajectory modeling and feature extraction method.
In order to achieve the above object, the present invention provides a trajectory owner prediction method based on a convolutional neural network, which is characterized by comprising the following steps:
(1) data preprocessing
(1.1) counting the longitude and latitude of the historical positions of all user tracks according to a time sequence to form a longitude and latitude set, wherein if a certain longitude and latitude appears repeatedly, the longitude and latitude set is only kept once;
numbering each longitude and latitude in the longitude and latitude set from 1, and giving a unique position ID;
(1.2) replacing the longitude and latitude of the historical positions of the user tracks by the position ID given in the step (1.1) according to the time sequence, and representing each user track by a string of position IDs; at the same time, each user is identified by a unique integer ID, so that the user and the user trajectory can form an owner trajectory in the form of [ user ID, trajectory (location ID, …, location ID) ];
(1.3) forming a directed weightless graph G according to the owner track<V,E>Where V is the set of all location IDs, if a certain user slave ID appearsiTo IDjThen, then<IDi,IDj>Representing a directed edge, all such edges comprising an edge set E of the graph G;
(2) track representation
(2.1) taking the constructed directed weightless graph G as input, and learning a low-dimensional real value vector of each position ID in G through a Node2Vec algorithm;
(2.2) slicing the track of each user in the step (1.2) according to a fixed time interval, thereby dividing each user track into a plurality of position ID sequences, and then carrying out owner identification on the cut position ID sequences by using the user ID;
(2.3) replacing the position ID with a low-dimensional real-valued vector corresponding to the position ID on the cut indefinite-length track, so as to generate a track matrix of each user;
then, constructing a track matrix of each user fixed dimension in an intercepting or filling mode to form a data set, wherein the filled vector is an average value of low-dimensional real-value vectors corresponding to all position IDs;
(3) and constructing a prediction model
Constructing a four-layer convolutional neural network, wherein an input layer of the four-layer convolutional neural network is a track matrix with fixed dimensionality; the convolution layer is provided with three convolution kernels m × embedding _ size, wherein m is a constant, and the embedding size is the dimension of a low-dimensional real value vector output by the Node2 vec; the pooling layer is k-max Pooling, k represents the first k maxima after convolution; the output of the full connection layer is input to a softmax function, and the probability distribution of the track owner is obtained;
(4) training prediction model
(4.1) construction of training set
Taking a track matrix set X of part of user fixed dimensions in the data set and a one-hot category vector set Y corresponding to the track matrix set X as a training set, wherein X is [ Vec _ X ═1,Vec_x2,...,Vec_xn],Vec_xnA trajectory matrix representing the nth user fixed dimension, Y ═ Vec _ t1,Vec_t2,...,Vec_tn],Vec_tnRepresenting a one-hot category vector corresponding to the nth owner, wherein if the nth position is 1, all the rest positions are 0;
(4.2) initializing the prediction model
Initializing a weight matrix W for each convolution kernel in a convolution layerpThe value of (a) is a normal distribution, the mean value thereof is 0, and the variance thereof is 0.1; initializing simultaneously the bias vector B of each convolution kernel in the convolution layerpThe number of elements of each bias vector is 0.1, namely the number of neurons of a corresponding layer; initializing the weight matrix of the full connection layer as W, wherein the dimension of the weight matrix is [ batch _ size × k ] convolution kernel number and category number]Simultaneously initializing the bias vector B value of the full-connection layer to be 0.1, wherein the number of elements is the number of categories; wherein, batch _ size is a constant, p is 1,2,3, and p represents the number of convolution kernels in the convolution layer;
(4.3) inputting the training set into the initialized prediction model, optimizing a loss function by adopting an Adam algorithm, then transmitting the error to the previous layer by utilizing an error back propagation BP algorithm, and updating the weight matrix W of the convolutional layerpOffset vector BpAfter a plurality of iterations, a converged neural network model is obtained, so that a trained prediction model is obtained;
(5) trajectory owner prediction
And (3) constructing a track matrix of the fixed dimension of the user according to the longitude and latitude of the user position to be detected in the steps (1) and (2), inputting the constructed track matrix into a trained prediction model to obtain probability distribution of the track corresponding to all owner categories, wherein the index of the maximum value in the probability distribution is the owner number corresponding to the track.
The invention aims to realize the following steps:
the invention relates to a trajectory owner prediction method based on a convolutional neural network, wherein a directed weightless graph G (V, E) is formed according to the trajectories of all users, and a low-dimensional real value vector of a trajectory position ID is learned through Node2 Vec; then, slicing the user track, replacing the position ID with a low-dimensional real value vector corresponding to the position ID for the cut indefinite-length track, and intercepting or filling to form a fixed dimension matrix of the track; and then, constructing and training a four-layer convolutional neural network as a prediction model, inputting a track matrix constructed by the latitude and longitude of the position of the user to be detected into the trained prediction model to obtain probability distribution of track owner classification, and finally marking the index of the maximum value in the probability distribution as the number of the track corresponding to the owner.
Meanwhile, the trajectory owner prediction method based on the convolutional neural network also has the following beneficial effects:
(1) constructing an owner track sequence into a network, namely a directed graph without authority, and learning a low-dimensional real value vector of each Node in the network through a Node2Vec algorithm;
(2) the track filling vector is the average value of all vectors, and compared with the situation that all the filling vectors are 0, the owner prediction accuracy is greatly improved;
(3) compared with the traditional convolutional neural network, the method improves the accuracy of the track owner prediction.
Drawings
FIG. 1 is a flow chart of a trajectory owner prediction method based on a convolutional neural network according to the present invention;
FIG. 2 is an architectural diagram of a predictive model.
Detailed Description
The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.
Examples
FIG. 1 is a flowchart of a trajectory owner prediction method based on a convolutional neural network.
In this embodiment, as shown in fig. 1, the trajectory owner prediction method based on the convolutional neural network of the present invention includes the following steps:
s1, preprocessing data
S1.1, counting the longitude and latitude of the historical positions of all user tracks according to a time sequence to form a longitude and latitude set, wherein if a certain longitude and latitude appears repeatedly, the longitude and latitude set is only retained once;
as shown in table 1, numbering each longitude and latitude in the longitude and latitude set from 1, and giving a unique position ID;
table 1 is a location ID identification table of latitude and longitude;
numbering | Latitude and |
1 | 39.747652-104.99251 |
2 | 39.891383-105.070814 |
3 | 39.891077-105.068532 |
4 | 39.750469-104.999073 |
…… | …… |
TABLE 1
S1.2, replacing the longitude and latitude of the historical positions of the user tracks by the position ID given in the step S1.1 according to the time sequence, and representing each user track by a string of position IDs; at the same time, each user is identified by a unique integer ID, so that the user and the user trajectory can form an owner trajectory in the form of [ user ID, trajectory (location ID, …, location ID) ]; in this embodiment, as an owner trajectory: [0, (622, 474, 474, 474, 481, 482, 482, 83, 83, 270, 487, 270, 270, 83, 83, 471, … …) ], wherein 0 is a user ID followed by a sequence of location IDs;
s1.3, forming a directed weightless graph G according to the owner track<V,E>Where V is the set of all location IDs, if a certain user slave ID appearsiTo IDjThen, then<IDi,IDj>Representing a directed edge, all such edges comprising an edge set E of the graph G;
s2, track representation
S2.1, taking the constructed directed weightless graph G as input, and learning out a low-dimensional real value vector of each position ID in G through a Node2Vec algorithm; node2Vec is a network representation learning method, and the specific process thereof belongs to the prior art, and is not described herein again.
In this embodiment, the low-dimensional real-valued vector such as position ID number 60 is represented as:
[60(-0.383389,-0.826315,-1.379363,……,-1.839076,1.930556,0.502587)];
s2.2, slicing the track of each user in the step S1.2 according to a fixed time interval, thereby dividing each user track into a plurality of position ID sequences, and then carrying out owner identification on the cut position ID sequences by using the user ID; in this embodiment, the slicing effect is as follows:
[0,(622,474,474,474,481,482,482,83)]
[0,(83,270,487,270,270,83,83,471)]
……
wherein 0 is a user ID, followed by a position ID sequence;
s2.3, replacing the position ID with a low-dimensional real-valued vector corresponding to the position ID for the cut indefinite-length track, and generating a track matrix of each user;
then, constructing a track matrix of each user fixed dimension in an intercepting or filling mode, wherein the filled vector is the average value of low latitude real value vectors corresponding to all position IDs; in the embodiment, the first 30 position IDs are intercepted to construct a track matrix of each user fixed dimension;
s3, constructing a prediction model
Constructing a four-layer convolutional neural network, as shown in fig. 2, wherein an input layer of the four-layer convolutional neural network is a track matrix with fixed dimensions; the convolution layer is provided with three convolution kernels m × embedding _ size, wherein the value of m is 2,3 and 4, the embedding size is the dimensionality of a low latitude real value vector output by Node2vec, and the number of each convolution kernel is 64; the pooling layer is k-max force, where k represents the first k maxima after convolution, where k is 3 in this example; and inputting the output of the full connection layer to a softmax function to obtain the probability distribution of the track owner.
S4 training prediction model
S4.1, constructing a training set
Taking a track matrix set X of part of user fixed dimensions in the data set and a one-hot category vector set Y corresponding to the track matrix set X as a training set, wherein X is [ Vec _ X ═1,Vec_x2,...,Vec_xn],Vec_xnA trajectory matrix representing the nth user fixed dimension, Y ═ Vec _ t1,Vec_t2,...,Vec_tn],Vec_tnRepresenting a one-hot category vector corresponding to the nth owner, wherein if the nth position is 1, all the rest positions are 0;
s4.2, initializing a prediction model
Initializing a weight matrix W for each convolution kernel in a convolution layerpThe value of (a) is positive-space distribution, the mean value is 0, and the variance is 0.1; initializing simultaneously the bias vector B of each convolution kernel in the convolution layerpThe number of elements of each bias vector is 0.1, namely the number of neurons of a corresponding layer; initializing the weight matrix of the full connection layer as W, wherein the dimension of the weight matrix is [ batch _ size × k ] convolution kernel number and category number]Simultaneously initializing the bias vector B value of the full-connection layer to be 0.1, wherein the number of elements is the number of categories; wherein, the batch _ size is a constant and takes a value of 64, p is 1,2,3, and p represents the number of convolution kernels in the convolution layer;
s4.3, inputting the training set into the initialized prediction model, and optimizing a loss function by adopting an Adam algorithm, wherein the optimized loss function is as follows:
wherein N is batch _ size, yjIs Vec _ tjAll elements of (1), ajIs the output value of the softmax function;
then, the error is transmitted to the previous layer by using an error back propagation BP algorithm, and the weight matrix W of the convolutional layer is updatedpOffset vector BpThe weight matrix W and the offset vector B of the full connection layer are subjected to iteration for a plurality of times to obtain a converged neural network model, so that a trained prediction model is obtained;
s5 trajectory owner prediction
And (4) constructing a track matrix of the fixed dimension of the user according to the longitude and latitude of the user position to be detected and the methods of the steps S1 and S2, inputting the constructed track matrix into the trained prediction model to obtain the probability distribution of the track corresponding to all owner categories, wherein the index of the maximum value in the probability distribution is the owner number corresponding to the track.
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all matters of the invention which utilize the inventive concepts are protected.
Claims (2)
1. A trajectory owner prediction method based on a convolutional neural network is characterized by comprising the following steps:
(1) data preprocessing
(1.1) counting the longitude and latitude of the historical positions of all user tracks according to a time sequence to form a longitude and latitude set, wherein if a certain longitude and latitude appears repeatedly, the longitude and latitude set is only kept once;
numbering each longitude and latitude in the longitude and latitude set from 1, and giving a unique position ID;
(1.2) replacing the longitude and latitude of the historical positions of the user tracks by the position ID given in the step (1.1) according to the time sequence, and representing each user track by a string of position IDs; at the same time, each user is identified by a unique integer ID, so that the user and the user trajectory can form an owner trajectory in the form of [ user ID, trajectory (location ID, …, location ID) ];
(1.3) forming a directed weightless graph G according to the owner track<V,E>Where V is the set of all location IDs, if a certain user slave ID appearsiTo IDjThen, then<IDi,IDj>Represents aThe strips have directed edges, all of which form the edge set E of the graph G;
(2) track representation
(2.1) taking the constructed directed weightless graph G as input, and learning a low-dimensional real value vector of each position ID in G through a Node2Vec algorithm;
(2.2) slicing the track of each user in the step (1.2) according to a fixed time interval, thereby dividing each user track into a plurality of position ID sequences, and then carrying out owner identification on the cut position ID sequences by using the user ID;
(2.3) replacing the position ID with a low-dimensional real-value vector corresponding to the position ID for the cut indefinite-length track, so as to generate a track matrix of each user;
then, constructing a track matrix of each user fixed dimension in an intercepting or filling mode to form a data set, wherein the filled vector is an average value of low-dimensional real-value vectors corresponding to all position IDs;
(3) and constructing a prediction model
Constructing a four-layer convolutional neural network, wherein an input layer of the four-layer convolutional neural network is a track matrix with fixed dimensionality; the convolution layer is provided with three convolution kernels m × embedding _ size, wherein m is a constant, and the embedding size is the dimension of a low-dimensional real value vector output by the Node2 vec; the pooling layer is k-max Pooling, k represents the first k maxima after convolution; the output of the full connection layer is input to a softmax function, and the probability distribution of the track owner is obtained;
(4) training prediction model
(4.1) construction of training set
Taking a track matrix set X of part of user fixed dimensions in the data set and a one-hot category vector set Y corresponding to the track matrix set X as a training set, wherein X is [ Vec _ X ═1,Vec_x2,...,Vec_xn],Vec_xnA trajectory matrix representing the nth user fixed dimension, Y ═ Vec _ t1,Vec_t2,...,Vec_tn],Vec_tnRepresenting a one-hot category vector corresponding to the nth owner, wherein if the nth position is 1, all the rest positions are 0;
(4.2) initializing the prediction model
Initializing a weight matrix W for each convolution kernel in a convolution layerpThe value of (a) is a normal distribution, the mean value thereof is 0, and the variance thereof is 0.1; initializing simultaneously the bias vector B of each convolution kernel in the convolution layerpThe number of elements of each bias vector is 0.1, namely the number of neurons of a corresponding layer; initializing the weight matrix of the full connection layer as W, wherein the dimension of the weight matrix is [ batch _ size × k ] convolution kernel number and category number]Simultaneously initializing the bias vector B value of the full-connection layer to be 0.1, wherein the number of elements is the number of categories; wherein, batch _ size is a constant, p is 1,2,3, and p represents the number of convolution kernels in the convolution layer;
(4.3) inputting the training set into the initialized prediction model, optimizing a loss function by adopting an Adam algorithm, then transmitting the error to the previous layer by utilizing an error back propagation BP algorithm, and updating the weight matrix W of the convolutional layerpOffset vector BpAfter a plurality of iterations, a converged neural network model is obtained, so that a trained prediction model is obtained;
(5) trajectory owner prediction
And (3) constructing a track matrix of the fixed dimension of the user according to the longitude and latitude of the user position to be detected in the steps (1) and (2), inputting the constructed track matrix into a trained prediction model to obtain probability distribution of the track corresponding to all owner categories, wherein the index of the maximum value in the probability distribution is the owner number corresponding to the track.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910266737.0A CN110059144B (en) | 2019-04-03 | 2019-04-03 | Trajectory owner prediction method based on convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910266737.0A CN110059144B (en) | 2019-04-03 | 2019-04-03 | Trajectory owner prediction method based on convolutional neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110059144A CN110059144A (en) | 2019-07-26 |
CN110059144B true CN110059144B (en) | 2021-12-10 |
Family
ID=67318263
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910266737.0A Expired - Fee Related CN110059144B (en) | 2019-04-03 | 2019-04-03 | Trajectory owner prediction method based on convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110059144B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021051224A1 (en) * | 2019-09-16 | 2021-03-25 | Intel Corporation | Stochastic trajectory prediction using social graph networks |
CN112405542B (en) * | 2020-11-17 | 2021-09-14 | 中国科学院自动化研究所 | Musculoskeletal robot control method and system based on brain inspiring multitask learning |
CN113222229B (en) * | 2021-04-28 | 2024-02-13 | 南京航空航天大学 | Non-cooperative unmanned aerial vehicle track prediction method based on machine learning |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108153867A (en) * | 2017-12-25 | 2018-06-12 | 山东大学 | User trajectory Forecasting Methodology and device based on temporal regularity |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018022715A1 (en) * | 2016-07-26 | 2018-02-01 | University Of Connecticut | Early prediction of an intention of a user's actions |
-
2019
- 2019-04-03 CN CN201910266737.0A patent/CN110059144B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108153867A (en) * | 2017-12-25 | 2018-06-12 | 山东大学 | User trajectory Forecasting Methodology and device based on temporal regularity |
Non-Patent Citations (2)
Title |
---|
Identifying Human Mobility via Trajectory Embeddings;Qiang Gao等;《Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17)》;20171231;全文 * |
一种基于高斯混合模型的轨迹预测算法;乔少杰等;《软件学报》;20150202;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN110059144A (en) | 2019-07-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110059144B (en) | Trajectory owner prediction method based on convolutional neural network | |
CN108875916B (en) | Advertisement click rate prediction method based on GRU neural network | |
CN112347970B (en) | Remote sensing image ground object identification method based on graph convolution neural network | |
CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
CN109523013B (en) | Air particulate matter pollution degree estimation method based on shallow convolutional neural network | |
CN111966823B (en) | Graph node classification method facing label noise | |
CN112487805B (en) | Small sample Web service classification method based on meta-learning framework | |
CN111242157A (en) | Unsupervised domain self-adaption method combining deep attention feature and conditional opposition | |
CN111125358A (en) | Text classification method based on hypergraph | |
CN111444951B (en) | Sample recognition model generation method, device, computer equipment and storage medium | |
CN112818690B (en) | Semantic recognition method and device combined with knowledge graph entity information and related equipment | |
CN113190688A (en) | Complex network link prediction method and system based on logical reasoning and graph convolution | |
CN111814189B (en) | Distributed learning privacy protection method based on differential privacy | |
CN108596204B (en) | Improved SCDAE-based semi-supervised modulation mode classification model method | |
CN114863091A (en) | Target detection training method based on pseudo label | |
CN114943859B (en) | Task related metric learning method and device for small sample image classification | |
CN113297936A (en) | Volleyball group behavior identification method based on local graph convolution network | |
CN116431597A (en) | Method, electronic device and computer program product for training a data classification model | |
CN114584230B (en) | Predictive channel modeling method based on countermeasure network and long-term and short-term memory network | |
CN113255661A (en) | Bird species image identification method related to bird-involved fault of power transmission line | |
CN105787045B (en) | A kind of precision Enhancement Method for visual media semantic indexing | |
CN111783688B (en) | Remote sensing image scene classification method based on convolutional neural network | |
CN117636183A (en) | Small sample remote sensing image classification method based on self-supervision pre-training | |
CN116415177A (en) | Classifier parameter identification method based on extreme learning machine |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20211210 |
|
CF01 | Termination of patent right due to non-payment of annual fee |