CN115496291A - Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value - Google Patents

Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value Download PDF

Info

Publication number
CN115496291A
CN115496291A CN202211236102.4A CN202211236102A CN115496291A CN 115496291 A CN115496291 A CN 115496291A CN 202211236102 A CN202211236102 A CN 202211236102A CN 115496291 A CN115496291 A CN 115496291A
Authority
CN
China
Prior art keywords
sample
data
data set
meteorological temperature
augmented
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
CN202211236102.4A
Other languages
Chinese (zh)
Inventor
马廷淮
杨彬
荣欢
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.)
Nanjing University of Information Science and Technology
Original Assignee
Nanjing University of Information Science and Technology
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 Nanjing University of Information Science and Technology filed Critical Nanjing University of Information Science and Technology
Priority to CN202211236102.4A priority Critical patent/CN115496291A/en
Publication of CN115496291A publication Critical patent/CN115496291A/en
Pending legal-status Critical Current

Links

Images

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
    • 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
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • 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
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Business, Economics & Management (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Economics (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Remote Sensing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a clustering type data augmented meteorological temperature prediction method based on high-precision residual defect values, which comprises the following steps of: s1, creating a sample set: repairing the residual values of the input original meteorological temperature data set, and dividing a sample set by combining a sliding window, an input time sequence length and a prediction time sequence length to form a complete sample set; s2, clustering: taking the value of each sample in the complete sample set from the temperature dimension to obtain data information representing the sample, and reducing the dimension to 3 dimensions by using principal component analysis; after the K value is selected, each sample is endowed with a class number represented by the sample by using a K-MEANS algorithm; s3, data augmentation: each sample is augmented according to the category to which the sample belongs; s4, training a time sequence predictor; and S5, testing the timing sequence predictor. Compared with methods such as an average value method, a mode method and the like, the prediction accuracy of the prediction method is greatly improved.

Description

Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value
Technical Field
The invention belongs to the technical field of meteorological information prediction, and particularly relates to a clustering type sample augmented meteorological temperature prediction model based on high-precision residual values, which is suitable for repairing a large number of residual data sets and solving the overfitting problem of a time sequence prediction model.
Background
High-precision repair and long-time-sequence prediction of incomplete information are two important basic subjects in artificial intelligence, and under many conditions, data values under multiple timestamps of a certain feature in the future need to be rapidly and accurately predicted from mass data, and the process is called time sequence prediction. Timing prediction is widely applied to the fields of machine learning and the like. The prediction algorithms widely adopted at present are mainly RNN and GRU. These timing prediction models can only solve the short timing prediction problem because too long timing can cause the gradient explosion or disappearance problem. The CNN + LSTM algorithm provides an effective method for solving the problems of characteristic space learning and long time sequence prediction and solves the problem of multi-variable multi-step time sequence prediction.
In practical application, abnormal values or missing values of data input into the predictor can occur due to various factors such as sensors and the like, and the abnormal values have important influence on the prediction accuracy, so that a high-accuracy data repairing system is needed. Data restoration is widely applied to the fields of data processing, data mining, machine learning and the like, and the restoration method mainly adopted at present is to analyze the statistical characteristics of missing data and then fill up the missing data by adopting data capable of representing the characteristics, such as mean and mode; however, this method cannot repair accurately, and the number of missing values directly affects the final repair effect. The machine learning regression problem algorithm-regression decision tree provides an effective method for solving the problem and solving the problem of high-precision data set restoration.
Decision regression trees are decision tree models for regression. The regression tree divides the input space by a heuristic method, namely traversing all input variables, finding the optimal segmentation point of the optimal segmentation variable, dividing the input space into two parts, and repeating the operation.
In time sequence prediction, a common method is to directly input sensor data into a prediction model, however, a neural network is difficult to learn the correlation among multiple features in a training process, so that the combination of CNN and LSTM can improve the prediction accuracy. CNN is used for multivariate feature extraction and LSTM is used for long-time prediction. Convolutional neural networks CNN perform well in many applications, such as image classification, object detection and medical image analysis. The main idea behind CNN is that it can take local features from higher layer inputs and shift them to lower layers to obtain more complex features; the convolutional neural network includes convolutional layers, pooling layers, and fully-connected layers. Long-short term memory is an improvement on the recurrent neural network. LSTM proposes memory blocks instead of traditional RNN cells in solving the gradient vanishing and gradient explosion problems.
Disclosure of Invention
The technical problem to be solved is as follows: aiming at the technical problems that long-time sequence prediction is carried out on a meteorological data set with a large number of deficiency values and overfitting needs to be solved, the invention provides a clustering type data augmentation meteorological temperature prediction method based on high-precision residual deficiency values, and the precision of the method is greatly improved compared with methods such as an average value method and a mode method.
The technical scheme is as follows:
a clustering type data augmented meteorological temperature prediction method based on high-precision residual values comprises the following steps:
s1, creating a sample set:
performing incomplete value restoration on an input original meteorological temperature data set, defining the restored original meteorological temperature data set as an integral data set, performing sample set division by combining a sliding window, an input time sequence length and a prediction time sequence length to form an integral sample set, and dividing the integral sample set into a training sample set and a testing sample set;
s2, clustering:
taking out the data information representing the sample from the temperature dimension of each sample in the complete sample set, and reducing the dimension to 3 dimensions by using principal component analysis to ensure that only 3 pieces of data information of each sample are available; after the K value is selected, each sample is endowed with the class number represented by the sample by using a K-MEANS algorithm, and the class number to which each sample belongs is stored;
s3, data augmentation:
amplifying each sample according to the category to which the sample belongs, storing the amplification data as an amplification sample set, and merging the amplification sample set and the complete sample set into a training sample set;
s4, training a time sequence predictor:
constructing a meteorological temperature predictor, importing a training sample set into the meteorological temperature predictor to learn spatial information and extract characteristics, and using MSE as a loss function;
the method comprises the following steps that a one-dimensional convolution is adopted to be practical for the features of each time step in a single sample, the principle that the number of channels is doubled and the number of features is halved is kept in the convolution process, data are tiled in the last module of a meteorological temperature predictor, and a Dense layer and a RELU activation function are used for halving the number of the features; in the convolution process, a residual error network is adopted to repair the network so as to solve the problem of gradient explosion; using LSTM to predict long time sequence in the last template of weather temperature predictor;
s5, testing a time sequence predictor:
and after the meteorological temperature predictor is trained, detecting the model performance of the meteorological temperature predictor by adopting a test data set.
Further, in step S1, the process of performing residual value restoration on the input original meteorological temperature data set includes the following sub-steps:
carrying out abnormal value retrieval on an input original meteorological temperature data set, and marking the obtained abnormal value position and the corresponding time point;
the entire raw meteorological temperature data set is divided into three parts, by label: an original data set, a good data set and a incomplete data set; the intact data set is a data set consisting of non-abnormal values in the original meteorological temperature data set, and the incomplete data set is a data set consisting of abnormal values in the original meteorological temperature data set;
putting the intact data set into a regression tree or a gradient lifting regression tree for model training to obtain a repair model;
and importing the incomplete data set into a repair model for repair, and covering the repaired numerical value to the original incomplete position in the original data set to obtain a repaired complete data set.
Further, in step S1, according to 7: a scale of 3 divides the complete sample set into a training sample set and a test sample set.
Further, in step S2, all samples are displayed as 3D images using matplotlib, and all samples in different clusters display dots of the same color.
Further, in step S3, the new data after augmentation is (X) new ,Y new )：
X new ＝0.7*X+0.1*X 1 +0.1*X 2 +0.1*X 3
Y new ＝0.7*Y+0.1*Y 1 +0.1*Y 2 +0.1*Y 3
Wherein (X, Y) is an amplified sample, (X) 1 ,Y 1 )、(X 2 ,Y 2 )、(X 3 ,Y 3 ) Three samples are randomly selected that belong to the same category as the augmented sample.
Has the beneficial effects that:
firstly, the clustering-type data augmented meteorological temperature prediction method based on the high-precision residual defect value trains different restoration models according to the situation of the defect value at each time point by using a complete data set, so that the defect value can be accurately restored.
Secondly, the clustering type data augmentation meteorological temperature prediction method based on the high-precision residual defect value utilizes an algorithm for carrying out data augmentation in a cluster after K-MEANS clustering, and solves the LSTM overfitting problem; compared with direct noise increase or European distance increase, the method has obvious improvement on interpretability and the prediction effect after the increase.
Drawings
FIG. 1 is a flowchart of a clustering-type data augmented weather temperature prediction method based on high-precision residual values according to an embodiment of the present invention;
FIG. 2 is a flow chart of K-MEANS based clustering;
FIG. 3 is a flow chart of K-MEANS based clustering data augmentation;
FIG. 4 is a detailed flowchart of the clustering-type data augmented weather temperature prediction method based on the high-precision residual value according to the embodiment of the present invention.
Detailed Description
The following examples are presented to enable one of ordinary skill in the art to more fully understand the present invention and are not intended to limit the invention in any way.
FIG. 4 is a detailed flowchart of the clustering-type data augmented weather temperature prediction method based on the high-precision residual value according to the embodiment of the present invention. In the embodiment, all data under time nodes with complete data are extracted and input to a gradient lifting regression tree (GBDTR) for model training, so that a high-precision data restoration system is formed finally; and then, data under the time nodes with the missing data are put into a data repairing system for high-precision data repairing, so that the incomplete data set is restored into a complete data set. And then cutting the repaired data set according to the length of a sliding window, the length of an input time sequence and the length of a predicted time sequence, and dividing the data set into a training set and a testing set according to the proportion of 70% to 30%. The samples are then subjected to Principal Component Analysis (PCA) dimensionality reduction to three dimensions and clustered according to an appropriate number of clusters, each sample being subjected to sample augmentation in its own cluster only, the augmented samples being given a high weight and the auxiliary augmented samples being given a low weight, but the sum of which must be equal to 1. And finally, combining the augmented sample and the training set into a new training set and inputting the new training set into a meteorological temperature prediction system. The method combines a machine learning decision regression tree algorithm and optimizes the regression tree algorithm by using gradient boosting; and the solution of under-fitting of the time sequence prediction model is solved by using clustering type data augmentation; finally, a one-dimensional convolution (CONV 2D) is used in the predictor to solve the problems of spatial feature learning and long-term short-term memory (LSTM) to solve long-time-sequence prediction.
Referring to fig. 1, the method comprises the steps of:
step 10, inputting the whole data set, searching abnormal values of the data set to obtain positions of the abnormal values, and marking corresponding time points of the positions; and the data set is divided into three parts according to the marks: raw data set, good data set, data set with incomplete data.
And 20, putting the complete data set into a regression tree or a gradient lifting regression tree for model training, wherein the hyper-parameters of the regression tree can be obtained by using default parameters in sklern. And after the decision regression tree training is finished, putting the incomplete data set into a repair model, and returning the obtained repair value to the original incomplete position in the original data set.
Step 30, defining the repaired original data set as a complete data set, performing sample division by integrating a sliding window, an input time sequence length and a prediction time sequence length to form a complete sample set, and dividing the complete sample set into a training sample set and a testing sample set according to a ratio of 7.
Step 40, as shown in fig. 2, extracting the data information representing each sample from the temperature dimension value of each sample in the complete sample set, and reducing the dimension to 3 dimensions (beneficial to visualization) by using principal component analysis; each sample can represent the whole sample by using 3 pieces of data after the dimension is reduced to 3, and each sample is given to a class number represented by the sample by using a K-MEANS algorithm after a proper K value is selected; displaying all samples by using a 3D image by using matplotlib, displaying dots with the same color by using all samples in different clusters, and rotating the 3D image by using a gyroscope to check whether the samples are reasonably clustered under a certain visual angle;
step 50, as shown in fig. 3, after the appropriate K value is found, the class number to which each sample belongs is stored, and each sample is augmented according to the class to which it belongs: the augmented sample data (X, Y) is selected from three sample data (X1, Y1), (X2, Y2), (X3, Y3) randomly, and the augmented new data: (X) new ＝0.7*X+0.1*X 1 +0.1*X 2 +0.1*X 3 ；Y new ＝0.7*Y+0.1*Y 1 +0.1*Y 2 +0.1*Y 3 ) And after the execution is finished, the augmentation data is stored as an augmentation sample set for use by the prediction model. And combining the augmentation sample set and the training sample set into a training sample set, wherein the test sample set is still unchanged.
And step 60, using one-dimensional convolution to use the characteristics of each time stamp in a single sample, keeping the principle that the number of channels is doubled and the number of characteristics is halved in the convolution process, tiling data in the last module, using a Dense layer and a RELU activation function to halve the number of characteristics, solving the problem of gradient explosion by using ResNet if the problem exists, and solving long-time prediction by using LSTM in the last module of the predictor.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to those skilled in the art without departing from the principles of the present invention may be apparent to those skilled in the relevant art and are intended to be within the scope of the present invention.

Claims (5)

1. A clustering type data augmented meteorological temperature prediction method based on high-precision residual values is characterized by comprising the following steps:
s1, creating a sample set:
performing incomplete value restoration on an input original meteorological temperature data set, defining the restored original meteorological temperature data set as an integral data set, performing sample set division by combining a sliding window, an input time sequence length and a prediction time sequence length to form an integral sample set, and dividing the integral sample set into a training sample set and a testing sample set;
s2, clustering:
taking out data information representing the sample from the temperature dimension value of each sample in the complete sample set, and reducing the dimension to 3 dimensions by using principal component analysis to ensure that only 3 pieces of data information of each sample exist; after the K value is selected, each sample is endowed with the class number represented by the sample by using a K-MEANS algorithm, and the class number to which each sample belongs is stored;
s3, data augmentation:
amplifying each sample according to the category to which the sample belongs, storing the amplification data as an amplification sample set, and merging the amplification sample set and the complete sample set into a training sample set;
s4, training a time sequence predictor:
constructing a meteorological temperature predictor, importing a training sample set into the meteorological temperature predictor to learn spatial information and extract features, and using MSE as a loss function;
the method comprises the following steps that a one-dimensional convolution is adopted to be practical for the features of each time step in a single sample, the principle that the number of channels is doubled and the number of features is halved is kept in the convolution process, data are tiled in the last module of a meteorological temperature predictor, and a Dense layer and a RELU activation function are used for halving the number of the features; in the convolution process, a residual error network is adopted to repair the network so as to solve the problem of gradient explosion; using LSTM to predict long time sequence in the last template of weather temperature predictor;
s5, testing a timing sequence predictor:
and after the meteorological temperature predictor is trained, detecting the model performance of the meteorological temperature predictor by adopting a test data set.
2. The clustering-type data augmented meteorological temperature prediction method based on the high-precision residual values according to claim 1, wherein in the step S1, the process of repairing the residual values of the input original meteorological temperature data set comprises the following sub-steps:
carrying out abnormal value retrieval on an input original meteorological temperature data set, and marking the obtained abnormal value position and the corresponding time point;
the entire raw meteorological temperature data set is divided into three parts, by label: an original data set, a good data set and a defective data set; the intact data set is a data set consisting of non-abnormal values in the original meteorological temperature data set, and the incomplete data set is a data set consisting of abnormal values in the original meteorological temperature data set;
putting the intact data set into a regression tree or a gradient lifting regression tree for model training to obtain a repair model;
and importing the incomplete data set into a repair model for repair, and covering the repaired numerical value on the original incomplete position in the original data set to obtain a repaired complete data set.
3. The clustering-type data augmented meteorological temperature prediction method based on the high-precision residual value according to claim 1, characterized in that in step S1, according to 7: a ratio of 3 divides the complete sample set into a training sample set and a test sample set.
4. The method for predicting the clustering-type data augmented weather temperature based on the high-precision residual value according to claim 1, wherein in step S2, all samples are displayed by using a 3D image by using matplotlib, and all samples in different clusters display dots of the same color.
5. The method for predicting the meteorological temperature augmented by the clustered data based on the high-precision residual value according to claim 1, wherein in the step S3, the new augmented data is (X) new ,Y new )：
X new ＝0.7*X+0.1*X 1 +0.1*X 2 +0.1*X 3
Y new ＝0.7*Y+0.1*Y 1 +0.1*Y 2 +0.1*Y 3
Wherein (X, Y) is an amplified sample, (X) 1 ,Y 1 )、(X 2 ,Y 2 )、(X 3 ,Y 3 ) Three samples were randomly selected that were of the same class as the sample being augmented.
CN202211236102.4A 2022-10-10 2022-10-10 Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value Pending CN115496291A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211236102.4A CN115496291A (en) 2022-10-10 2022-10-10 Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211236102.4A CN115496291A (en) 2022-10-10 2022-10-10 Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value

Publications (1)

Publication Number Publication Date
CN115496291A true CN115496291A (en) 2022-12-20

Family

ID=84474973

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211236102.4A Pending CN115496291A (en) 2022-10-10 2022-10-10 Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value

Country Status (1)

Country Link
CN (1) CN115496291A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117932245A (en) * 2024-03-21 2024-04-26 华南理工大学 Financial data missing value completion method, device and storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117932245A (en) * 2024-03-21 2024-04-26 华南理工大学 Financial data missing value completion method, device and storage medium
CN117932245B (en) * 2024-03-21 2024-06-11 华南理工大学 Financial data missing value completion method, device and storage medium

Similar Documents

Publication Publication Date Title
CN110852316B (en) Image tampering detection and positioning method adopting convolution network with dense structure
CN108830285B (en) Target detection method for reinforcement learning based on fast-RCNN
Xu et al. A weakly supervised surface defect detection based on convolutional neural network
CN113436169B (en) Industrial equipment surface crack detection method and system based on semi-supervised semantic segmentation
CN112200121B (en) Hyperspectral unknown target detection method based on EVM and deep learning
JP2017062713A (en) Identifier creation circuit, identifier creation method, and program
CN108734109B (en) Visual target tracking method and system for image sequence
WO2019026134A1 (en) Information processing device and information processing method
CN108171119B (en) SAR image change detection method based on residual error network
CN112365497A (en) High-speed target detection method and system based on Trident Net and Cascade-RCNN structures
CN113313166B (en) Ship target automatic labeling method based on feature consistency learning
CN110909657A (en) Method for identifying apparent tunnel disease image
CN114266794A (en) Pathological section image cancer region segmentation system based on full convolution neural network
CN109002792B (en) SAR image change detection method based on layered multi-model metric learning
CN114898472A (en) Signature identification method and system based on twin vision Transformer network
CN115496291A (en) Clustering type data augmented meteorological temperature prediction method based on high-precision residual defect value
CN113362277A (en) Workpiece surface defect detection and segmentation method based on deep learning
CN114119460A (en) Semiconductor image defect identification method, semiconductor image defect identification device, computer equipment and storage medium
CN110717602B (en) Noise data-based machine learning model robustness assessment method
CN111832616A (en) Method and system for identifying airplane model by using remote sensing image of multiple types of depth maps
Silva et al. Online weighted one-class ensemble for feature selection in background/foreground separation
CN111242028A (en) Remote sensing image ground object segmentation method based on U-Net
CN113077438B (en) Cell nucleus region extraction method and imaging method for multi-cell nucleus color image
CN117115147B (en) Textile detection method and system based on machine vision
CN117557827A (en) Plate shape anomaly detection method based on self-coding cascade forests

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