CN108846338A - Polarization characteristic selection and classification method based on object-oriented random forest - Google Patents
Polarization characteristic selection and classification method based on object-oriented random forest Download PDFInfo
- Publication number
- CN108846338A CN108846338A CN201810561139.1A CN201810561139A CN108846338A CN 108846338 A CN108846338 A CN 108846338A CN 201810561139 A CN201810561139 A CN 201810561139A CN 108846338 A CN108846338 A CN 108846338A
- Authority
- CN
- China
- Prior art keywords
- feature
- classification
- random forest
- decomposed
- sample
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- 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
- G06F18/211—Selection of the most significant subset of features
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of polarization characteristic selection and classification method based on object-oriented random forest, solve the problems, such as the feature selecting and image classification when numerous polarization characteristics participate in classification.This method carries out multi-scale division to characteristic set using object-oriented method, carries out random forest modeling to the sample object after segmentation, and calculate the importance of each feature, carries out feature set optimization to selection algorithm using before sequence.The present invention improves model training efficiency and nicety of grading using the random forest method of object-oriented.Using the building for carrying out optimal feature subset before sequence to selection algorithm this stopping criterion for iteration of combination precision highest, avoid falling into locally optimal solution.The algorithm can provide qualitative reference while improving nicety of grading for reasonably optimizing feature set.
Description
Technical field
The invention belongs to field of image processings, relate generally to polarimetric SAR image feature extraction and classifying, specifically a kind of base
In the polarization characteristic selection of object-oriented random forest and classification method.
Background technique
In recent years, polarimetric SAR image was increasingly used in the extraction of earth's surface information.Method based on goal decomposition
It is the important means that analysis and information extraction are carried out to polarimetric SAR image.For more complicated geomorphological environment, only according to certain pole
Change decomposition means or certain several characteristic parameter is difficult effectively to be distinguished all ground surface types.Then a variety of polarization point are integrated
Resolving Algorithm has become the effective way for solving the problems, such as this in conjunction with a variety of polarization characteristic parameters.But in practical applications, excessively
There may be interdepending between characteristic parameter, it is easy to cause a series of problem, such as：Analysis feature, training pattern are taken
Between it is too long, cause " dimension disaster ", model is complicated, and calculation amount increases, nicety of grading decline etc..Therefore, how to comform multiple features
The feature good to atural object target classification effect is selected in parameter, and calculation amount and information redundancy are reduced while guaranteeing nicety of grading
It will be another critical issue.
Random forest (Random Forest, RF) is a kind of superior performance that machine learning field grows up in recent years
Classifier, be that decision tree is freely combined, the algorithm while overcoming single decision tree deficiency, be not necessarily to people
Overfitting phenomenon will not occur in work beta pruning, and improve model training efficiency, and prediction effect is preferable.Random forests algorithm mentions
The calculation method of variable importance has been supplied, but how also to have needed further to consider according to importance quantitative selection optimized parameter.
Based on this, the present invention proposes a kind of polarization characteristic selection and classification method based on object-oriented random forest, and this method is to divide
Cutting object is that unit chooses sample, numerous polarization characteristic parameters of sample is extracted, according to the size of their importance values, using sequence
To selection (Sequential Forward Selection, SFS) algorithm and in conjunction with this iteration ends item of precision highest before arranging
Part carries out feature set optimization, which not only allows for the space relationship between pixel, and high-efficient, the optimal characteristics that will be obtained
Subset is in the classification of polarimetric SAR image, nicety of grading to be greatly improved.
Summary of the invention
The purpose of the present invention is to provide a kind of feature extracting methods based on a variety of polarization decomposing algorithms, by object-oriented
Divide the selection combined with random forest modeling for polarization characteristic, qualitative reference is provided for reasonably optimizing feature set, using sequence
Optimal feature subset is constructed to selection algorithm before arranging, and is up to stopping criterion for iteration with precision, avoids falling into locally optimal solution,
Improve nicety of grading.
The present invention in order to solve the above technical problems, using following technical scheme：The present invention provides a kind of based on object-oriented
The polarization characteristic of random forest selects and classification method, and this approach includes the following steps：
Step 1, full polarimetric SAR is pre-processed, multiple look processing and certain filtering algorithm removal is respectively adopted
Speckle noise in image improves the improvement of visual effect of image；
Step 2,20 kinds of goal decomposition algorithms are based on to pretreated polarization image to decompose.
Step 3, multi-scale division is carried out to characteristic set using object-oriented method, as unit of each object, chosen
A certain number of training samples carry out random forest modeling to it；
Step 4, the importance that feature is calculated according to sample, is ranked up the size of importance values；
Step 5, the highest feature of importance values is added to destination subset to selection algorithm using before sequence；
Step 6, the feature of destination subset is based on random forest method to classify, and calculates overall accuracy；
Step 7, optimal polarization characteristic parameter is selected according to the classification overall accuracy calculated every time, constitutes optimal characteristics
Collection.
Further, it in the step 2, in order to sufficiently excavate the scattered information of polarimetric SAR image, is more polarized
Characteristic parameter constructs initial polarization characteristic set, carries out polarization decomposing using 20 kinds of goal decomposition algorithms, is Pauli respectively
Decompose, Krogager decompose, Huynen decompose, Barnes1 decompose, Barnes2 decompose, Holm1 decompose, Holm2 decompose,
VanZly3 is decomposed, Cloude is decomposed, H/A/Alpha is decomposed, Freeman2 is decomposed, Freeman3 is decomposed, Yamaguchi3 points
Solution, Yamaguchi4 are decomposed, Neumann is decomposed, Touzi is decomposed, An_Yang3 is decomposed, An_Yang4 is decomposed, Arii3_NNED
It decomposes, Arii3_ANNED is decomposed.By the available 93 polarization characteristic parameters of the above decomposition algorithm, in addition 3 matrix elements
Element --- S11, S12 and S22 obtains the set comprising 96 polarization characteristics.
Further, in the step 3, the selection of best segmental scale can be obtained by the method for many experiments not
With the segmentation result under scale, and by visual observation, interpretation is chosen.Then random forest modeling is carried out according to selected sample, it is first
K self-service sample sets are first randomly selected from original training data concentration with putting back to using bootstrap (Bootstrap), utilize this
K sample set constructs k decision tree.In this process, the sample not being extracted every time forms the outer data (Out-of- of k bag
Bag, OOB)；Next, being equipped with N number of feature, then randomly select n feature (n≤N), leads at each node of every one tree
Cross and calculate the information content that each feature contains, the strongest feature of classification capacity is selected to be divided, such decision tree certain
One leaf node or be that can not continue division or all samples of the inside all point to the same classification, each tree is all
Without beta pruning, grow it to the maximum extent；All decision trees are finally formed into random forest, it, will after random forest building
In new sample input classifier, vote for every decision tree of each sample its classification, classification results press decision
Votes are set to determine.
Further, in the step 4, serial number, the meter of each feature importance are write from 1 to 96 to each characteristic parameter
It calculates using the method based on outer data (Out of Bag, the OOB) error of bag, is equipped with self-service sample b=1,2 ..., B, variable Xj
The importance measures based on OOB errorCalculating process is as follows, the outer data of bag when finding b=1 firstWith tree TbIt is rightClassify, and records correct classification numberFor variable Xj, j=1,2 ..., N are rightIn XjValue carry out
Disturbance, the data set after disturbance are denoted asThen using tree TbIt is rightClassify, and records correct classification numberIt is right
In b=2,3 ..., B, above procedure is repeated, then variable XjThe variable importance calculation formula based on OOB error it is as follows：
The importance of each characteristic parameter is calculated by above formula, and is ranked up from big to small according to its value.
The invention adopts the above technical scheme, compared with prior art, has the following technical effects：
Model training efficiency and nicety of grading are improved using the random forest method of object-oriented in the present invention.Using sequence
The building for carrying out optimal feature subset before arranging to selection algorithm this stopping criterion for iteration of combination precision highest, avoids falling into part
Optimal solution.The algorithm can provide qualitative reference while improving nicety of grading for reasonably optimizing feature set.
Detailed description of the invention
Fig. 1 is flow chart of the invention；
Fig. 2 is polarization characteristic parameter importance ranking；
Fig. 3 is the polarization characteristic number of parameters for participating in classification and the relationship of classification overall accuracy；
Fig. 4 is the final result classified using the mentioned classification method of the present invention and other methods.
Specific embodiment
Technical solution of the present invention is described in further detail below with reference to specific attached drawing.Following embodiment is only used for
It is illustrated more clearly that technical solution of the present invention, and not intended to limit the protection scope of the present invention.
Polarization characteristic selection and classification method based on object-oriented random forest as shown in Figure 1, includes the following steps：
Step 1, full polarimetric SAR is pre-processed, multiple look processing and certain filtering algorithm removal is respectively adopted
Speckle noise in image improves the improvement of visual effect of image.
For the validity of qualitative and quantitative analysis the method for the present invention, experimental data is complete using the ALOS PALSAR of Japan
Polarize image, and the incident angle of image is 23.858 °, range resolution 9.37m, azimuth resolution 3.57m.First
Image is filtered and mostly view etc. pre-processes, wherein filtering uses window size to filter for the Refined Lee of 3 × 3 sizes
Device, multiple look processing orientation and distance to ratio be 6: 1.
It should be noted that method of the invention is applicable not only to ALOS PALSAR full polarimetric SAR data used in experiment,
Other airborne or spaceborne full polarimetric SAR datas are also equally applicable to, only the parameters such as the size of filter window and the ratio of multiple look processing
Selection have difference, need to be chosen for different data sources.
Step 2, in order to sufficiently excavate the scattered information of polarimetric SAR image, more polarization characteristic parameters are obtained, building is most
First polarization characteristic set carries out polarization decomposing using 20 kinds of goal decomposition algorithms, is that Pauli is decomposed, Krogager divides respectively
Solution, Huynen are decomposed, Barnes1 is decomposed, Barnes2 is decomposed, Holm1 is decomposed, Holm2 is decomposed, VanZly3 is decomposed, Cloude
Decompose, H/A/Alpha decompose, Freeman2 decompose, Freeman3 decompose, Yamaguchi3 decompose, Yamaguchi4 decompose,
Neumann is decomposed, Touzi is decomposed, An_Yang3 is decomposed, An_Yang4 is decomposed, Arii3_NNED is decomposed, Arii3_ANNED points
Solution.By the available 93 polarization characteristic parameters of the above decomposition algorithm, in addition 3 matrix elements --- S11, S12 and S22,
Obtain the set comprising 96 polarization characteristics.It should be noted that：Polarization decomposing algorithm can select not Tongfang as needed
The decomposition algorithm of method, use is different, and obtained characteristic parameter is also different.
Step 3, multi-scale division is carried out to characteristic set using object-oriented method, as unit of each object, chosen
A certain number of training samples carry out random forest modeling to it.The selection of best segmental scale can pass through the side of many experiments
Method obtains the segmentation result under different scale, and interpretation is chosen by visual observation.The ALOS used for experiment
PALSAR data are compared by many experiments and are found, when dividing scale is 15, segmentation effect is best, so subsequent experimental is all
It is based on the segmentation result under the scale factor.Then random forest modeling is carried out according to selected sample, uses bootstrap first
(Bootstrap) k self-service sample sets are randomly selected from original training data concentration with putting back to, utilize this k sample set structure
Build k decision tree.In this process, the sample not being extracted every time forms the outer data (Out-of-Bag, OOB) of k bag；It connects down
Come, be equipped with N number of feature, then randomly selected at each node of every one tree n feature (n≤N), by calculating each feature
The information content contained selects the strongest feature of classification capacity to be divided, some leaf node of such decision tree is wanted
It is that can not continue division or all samples of the inside all point to the same classification, each tree all without beta pruning, makes it
It grows to the maximum extent；All decision trees are finally formed into random forest, after random forest building, new sample is inputted into classification
It in device, votes for every decision tree of each sample its classification, classification results are determined by decision tree votes.
Step 4, serial number is write from 1 to 96 to each characteristic parameter, the calculating of feature importance is used based on the outer data of bag
The method of (Out of Bag, OOB) error is equipped with self-service sample b=1,2 ..., B, variable XjBased on the important of OOB error
Property measurementCalculating process is as follows, the outer data of bag when finding b=1 firstWith tree TbIt is rightClassify, and records
Correct classification numberFor variable Xj, j=1,2 ..., N are rightIn XjValue disturbed, the data set after disturbance
It is denoted asThen using tree TbIt is rightClassify, and records correct classification numberFor b=2,3 ..., B, repeat
Above procedure, then variable XjThe variable importance calculation formula based on OOB error it is as follows：
The importance of each characteristic parameter is calculated by above formula, and (figure is ranked up according to its value from big to small
2)。
Step 5, using, to selection algorithm, target signature subset being added in the maximum feature of importance values every time before sequence,
Classified using the feature in target signature subset, and calculates overall classification accuracy.A feature is added every time, gradually changes
Generation.
Step 6, after each iteration, the feature of destination subset is based on random forest method and is classified, analysis changes every time
The relationship (Fig. 3) of nicety of grading and polarization characteristic number of parameters that generation calculates.For the ease of quantitative analysis and analysis, the present invention is adopted
The classification of the method for the present invention is evaluated with overall accuracy (OA), producer's precision (PA) and user's precision (UA) (table 1 and table 2)
As a result (Fig. 4 a) and QUEST traditional decision-tree classification results (Fig. 4 b).
The nicety of grading of 1 the method for the present invention of table
The nicety of grading of 2 QUEST traditional decision-tree of table
It can be seen that compared with decision Tree algorithms from table 1, table 2 and Fig. 4, the method for the present invention has better effect：Pass through
Discovery is compared with ground truth, in the classification results obtained using the mentioned method of the present invention, the mistake of atural object divides phenomenon
It is effectively improved (oval and rectangular area in figure), the similar region of scattering mechanism has obtained effective differentiation, final result
It is more nearly with real surface, classification overall accuracy improves 11% or more, Kappa coefficient and improves 0.14.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the principle of the present invention, it can also make several improvements and deform, these improvement and deformations are also answered
It is considered as protection scope of the present invention.
Claims (4)
1. polarization characteristic selection and classification method based on object-oriented random forest, it is characterised in that include the following steps：
Step 1, full polarimetric SAR is pre-processed, multiple look processing and certain filtering algorithm removal image is respectively adopted
In speckle noise, improve the improvement of visual effect of image；
Step 2,20 kinds of goal decomposition algorithms are based on to pretreated polarization image to decompose；
Step 3, multi-scale division is carried out to characteristic set using object-oriented method, as unit of each object, chosen certain
The training sample of quantity carries out random forest modeling to it；
Step 4, the importance that feature is calculated according to sample, is ranked up the size of importance values；
Step 5, the highest feature of importance values is added to destination subset to selection algorithm using before sequence；
Step 6, the feature of destination subset is based on random forest method to classify, and calculates overall accuracy；
Step 7, optimal polarization characteristic parameter is selected according to the classification overall accuracy calculated every time, constitutes optimal feature subset.
2. the polarization characteristic selection based on object-oriented random forest and classification method, feature exist according to claim 1
In：
In the step 2, in order to sufficiently excavate the scattered information of polarimetric SAR image, more polarization characteristic parameters are obtained, are constructed
Initial polarization characteristic set carries out polarization decomposing using 20 kinds of goal decomposition algorithms, is Pauli decomposition, Krogager respectively
Decompose, Huynen decompose, Barnes1 decompose, Barnes2 decompose, Holm1 decompose, Holm2 decompose, VanZly3 decompose,
Cloude is decomposed, H/A/Alpha is decomposed, Freeman2 is decomposed, Freeman3 is decomposed, Yamaguchi3 is decomposed, Yamaguchi4
It decomposes, Neumann is decomposed, Touzi is decomposed, An_Yang3 is decomposed, An_Yang4 is decomposed, Arii3_NNED is decomposed, Arii3_
ANNED is decomposed, by the available 93 polarization characteristic parameters of the above decomposition algorithm, in addition 3 matrix elements --- S11, S12
And S22, obtain the set comprising 96 polarization characteristics.
3. the polarization characteristic selection based on object-oriented random forest and classification method, feature exist according to claim 1
In：
In the step 3, the selection of best segmental scale can obtain the segmentation under different scale by the method for many experiments
As a result, simultaneously interpretation is chosen by visual observation, random forest modeling is then carried out according to selected sample, using bootstrap
(Bootstrap) k self-service sample sets are randomly selected from original training data concentration with putting back to, utilize this k sample set structure
K decision tree is built, in this process, the sample not being extracted every time forms the outer data (Out-of-Bag, OOB) of k bag；It connects down
Come, be equipped with N number of feature, then randomly selected at each node of every one tree n feature (n≤N), by calculating each feature
The information content contained selects the strongest feature of classification capacity to be divided, some leaf node of such decision tree is wanted
It is that can not continue division or all samples of the inside all point to the same classification, each tree all without beta pruning, makes it
It grows to the maximum extent；All decision trees are finally formed into random forest, after random forest building, new sample is inputted into classification
It in device, votes for every decision tree of each sample its classification, classification results are determined by decision tree votes.
4. the polarization characteristic selection based on object-oriented random forest and classification method, feature exist according to claim 1
In：
In the step 4, serial number is write from 1 to 96 to each characteristic parameter, the calculating of feature importance is used based on number outside bag
According to the method for (Out of Bag, OOB) error, it is equipped with self-service sample b=1,2 ..., B, variable XjThe weight based on OOB error
The property wanted is measuredCalculating process is as follows, the outer data of bag when finding b=1 firstWith tree TbIt is rightClassify, and remembers
The correct classification number of recordFor variable Xj, j=1,2 ..., N are rightIn XjValue disturbed, the data after disturbance
Collection is denoted asThen using tree TbIt is rightClassify, and records correct classification numberFor b=2,3 ..., B, weight
Above procedure is answered, then variable XjThe variable importance calculation formula based on OOB error it is as follows：
The importance of each characteristic parameter is calculated by above formula, and is ranked up from big to small according to its value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810561139.1A CN108846338B (en) | 2018-05-29 | 2018-05-29 | Polarization feature selection and classification method based on object-oriented random forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810561139.1A CN108846338B (en) | 2018-05-29 | 2018-05-29 | Polarization feature selection and classification method based on object-oriented random forest |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108846338A true CN108846338A (en) | 2018-11-20 |
| CN108846338B CN108846338B (en) | 2022-04-15 |
Family
ID=64210205
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810561139.1A Active CN108846338B (en) | 2018-05-29 | 2018-05-29 | Polarization feature selection and classification method based on object-oriented random forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108846338B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472321A (en) * | 2018-12-03 | 2019-03-15 | 北京工业大学 | A kind of prediction towards time series type surface water quality big data and assessment models construction method |
| CN109726826A (en) * | 2018-12-19 | 2019-05-07 | 东软集团股份有限公司 | Training method, device, storage medium and the electronic equipment of random forest |
| CN110096967A (en) * | 2019-04-10 | 2019-08-06 | 同济大学 | A kind of road anger driver's hazardous act characteristic variable screening technique based on random forests algorithm |
| CN110717495A (en) * | 2019-09-30 | 2020-01-21 | 北京工业大学 | Solid waste incineration working condition identification method based on multi-scale color moment features and random forest |
| CN111107092A (en) * | 2019-12-23 | 2020-05-05 | 深圳供电局有限公司 | Attack recognition method based on random forest algorithm and energy storage coordination control device |
| EP3787229A1 (en) | 2019-09-02 | 2021-03-03 | Siemens Aktiengesellschaft | Method and device for automatically selecting analysis strings for feature extraction |
| CN112446522A (en) * | 2019-09-02 | 2021-03-05 | 中国林业科学研究院资源信息研究所 | Grass yield estimation method and device facing multi-scale segmentation and storage medium |
| CN113095426A (en) * | 2021-04-22 | 2021-07-09 | 西安交通大学 | Encrypted traffic classification method, system, equipment and readable storage medium |
| CN114187533A (en) * | 2022-02-15 | 2022-03-15 | 西南交通大学 | GB-InSAR (GB-InSAR) atmospheric correction method based on random forest time sequence classification |
| CN116206203A (en) * | 2023-03-08 | 2023-06-02 | 中国石油大学(华东) | Oil spill detection method based on SAR and Dual-EndNet |
| CN118154936A (en) * | 2024-02-01 | 2024-06-07 | 北京格致博雅生物科技有限公司 | Machine learning-based variety identification and classification method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778836A (en) * | 2016-11-29 | 2017-05-31 | 天津大学 | A kind of random forest proposed algorithm based on constraints |
| CN107563425A (en) * | 2017-08-24 | 2018-01-09 | 长安大学 | A kind of method for building up of the tunnel operation state sensor model based on random forest |
| CN107766883A (en) * | 2017-10-13 | 2018-03-06 | 华中师范大学 | A kind of optimization random forest classification method and system based on weighted decision tree |
-
2018
- 2018-05-29 CN CN201810561139.1A patent/CN108846338B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778836A (en) * | 2016-11-29 | 2017-05-31 | 天津大学 | A kind of random forest proposed algorithm based on constraints |
| CN107563425A (en) * | 2017-08-24 | 2018-01-09 | 长安大学 | A kind of method for building up of the tunnel operation state sensor model based on random forest |
| CN107766883A (en) * | 2017-10-13 | 2018-03-06 | 华中师范大学 | A kind of optimization random forest classification method and system based on weighted decision tree |
Non-Patent Citations (1)
| Title |
|---|
| A. MARCANO-CEDEÑO等: "Feature selection using Sequential Forward Selection and classification applying Artificial Metaplasticity Neural Network", 《IECON 2010 - 36TH ANNUAL CONFERENCE ON IEEE INDUSTRIAL ELECTRONICS SOCIETY》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472321B (en) * | 2018-12-03 | 2023-01-31 | 北京工业大学 | Time series type earth surface water quality big data oriented prediction and evaluation model construction method |
| CN109472321A (en) * | 2018-12-03 | 2019-03-15 | 北京工业大学 | A kind of prediction towards time series type surface water quality big data and assessment models construction method |
| CN109726826A (en) * | 2018-12-19 | 2019-05-07 | 东软集团股份有限公司 | Training method, device, storage medium and the electronic equipment of random forest |
| CN109726826B (en) * | 2018-12-19 | 2021-08-13 | 东软集团股份有限公司 | Training method and device for random forest, storage medium and electronic equipment |
| CN110096967A (en) * | 2019-04-10 | 2019-08-06 | 同济大学 | A kind of road anger driver's hazardous act characteristic variable screening technique based on random forests algorithm |
| EP3787229A1 (en) | 2019-09-02 | 2021-03-03 | Siemens Aktiengesellschaft | Method and device for automatically selecting analysis strings for feature extraction |
| CN112446522A (en) * | 2019-09-02 | 2021-03-05 | 中国林业科学研究院资源信息研究所 | Grass yield estimation method and device facing multi-scale segmentation and storage medium |
| WO2021043570A1 (en) | 2019-09-02 | 2021-03-11 | Siemens Aktiengesellschaft | Method and device for automatically selecting analysis chains for extracting features |
| CN110717495A (en) * | 2019-09-30 | 2020-01-21 | 北京工业大学 | Solid waste incineration working condition identification method based on multi-scale color moment features and random forest |
| CN110717495B (en) * | 2019-09-30 | 2024-01-26 | 北京工业大学 | Solid waste incineration working condition identification method based on multi-scale color moment characteristics and random forest |
| CN111107092A (en) * | 2019-12-23 | 2020-05-05 | 深圳供电局有限公司 | Attack recognition method based on random forest algorithm and energy storage coordination control device |
| CN113095426A (en) * | 2021-04-22 | 2021-07-09 | 西安交通大学 | Encrypted traffic classification method, system, equipment and readable storage medium |
| CN113095426B (en) * | 2021-04-22 | 2023-03-31 | 西安交通大学 | Encrypted traffic classification method, system, equipment and readable storage medium |
| CN114187533A (en) * | 2022-02-15 | 2022-03-15 | 西南交通大学 | GB-InSAR (GB-InSAR) atmospheric correction method based on random forest time sequence classification |
| CN116206203A (en) * | 2023-03-08 | 2023-06-02 | 中国石油大学(华东) | Oil spill detection method based on SAR and Dual-EndNet |
| CN116206203B (en) * | 2023-03-08 | 2023-08-18 | 中国石油大学(华东) | Oil spill detection method based on SAR and Dual-EndNet |
| CN118154936A (en) * | 2024-02-01 | 2024-06-07 | 北京格致博雅生物科技有限公司 | Machine learning-based variety identification and classification method and system |
| CN118154936B (en) * | 2024-02-01 | 2024-09-10 | 北京格致博雅生物科技有限公司 | Machine learning-based variety identification and classification method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108846338B (en) | 2022-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108846338A (en) | Polarization characteristic selection and classification method based on object-oriented random forest | |
| Aldino et al. | Implementation of K-means algorithm for clustering corn planting feasibility area in south lampung regency | |
| Jin et al. | Stem–leaf segmentation and phenotypic trait extraction of individual maize using terrestrial LiDAR data | |
| CN110348399B (en) | Hyperspectral intelligent classification method based on prototype learning mechanism and multidimensional residual error network | |
| CN107229917B (en) | A kind of several remote sensing image general character well-marked target detection methods based on iteration cluster | |
| CN110135494A (en) | Feature selection method based on maximum information coefficient and Gini index | |
| CN109492593B (en) | Hyperspectral image classification method based on principal component analysis network and space coordinates | |
| CN104462184B (en) | A kind of large-scale data abnormality recognition method based on two-way sampling combination | |
| CN104392250A (en) | Image classification method based on MapReduce | |
| CN112052755A (en) | Semantic convolution hyperspectral image classification method based on multi-path attention mechanism | |
| CN104463203B (en) | High-spectrum remote sensing semisupervised classification method based on the scoring of atural object classification degree of membership | |
| CN105184314B (en) | Wrapper formula EO-1 hyperion band selection methods based on pixel cluster | |
| CN102930275A (en) | Remote sensing image feature selection method based on Cramer's V index | |
| CN113420640A (en) | Mangrove hyperspectral image classification method and device, electronic equipment and storage medium | |
| CN110390358A (en) | A kind of deep learning method based on feature clustering | |
| CN110334777A (en) | A kind of unsupervised attribute selection method of weighting multi-angle of view | |
| Naeini et al. | Improving the dynamic clustering of hyperspectral data based on the integration of swarm optimization and decision analysis | |
| CN111507297A (en) | Radar signal identification method and system based on measurement information matrix | |
| CN107704872A (en) | A kind of K means based on relatively most discrete dimension segmentation cluster initial center choosing method | |
| CN104732246B (en) | A kind of semi-supervised coorinated training hyperspectral image classification method | |
| CN104143009B (en) | Competition and cooperation clustering method based on the maximal clearance cutting of dynamic encompassing box | |
| Zhu et al. | Exploring soybean flower and pod variation patterns during reproductive period based on fusion deep learning | |
| CN111488520B (en) | Crop planting type recommendation information processing device, method and storage medium | |
| CN114065831A (en) | Hyperspectral image classification method based on multi-scale random depth residual error network | |
| Ibrahim et al. | On feature selection methods for accurate classification and analysis of emphysema ct images |
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 |