CN109542993A - A kind of Remote Sensing Data Processing method based on rough set - Google Patents
A kind of Remote Sensing Data Processing method based on rough set Download PDFInfo
- Publication number
- CN109542993A CN109542993A CN201811369570.2A CN201811369570A CN109542993A CN 109542993 A CN109542993 A CN 109542993A CN 201811369570 A CN201811369570 A CN 201811369570A CN 109542993 A CN109542993 A CN 109542993A
- Authority
- CN
- China
- Prior art keywords
- data
- screening
- remote sensing
- carries out
- garbled
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Instructional Devices (AREA)
Abstract
It is not able to satisfy the demand of the Fast Attribute reduction of remotely-sensed data for the prior art, proposes a kind of remote sensing data mining method based on rough set;Attribute reduction is carried out using parallel method;Reduce the output of intermediate result;Go out useful use of information rough set Fast Attribute reduction from remote sensing data mining and excavates remotely-sensed data;The present invention can effectively improve remotely-sensed data attribute reduction speed.
Description
Technical field
The invention belongs to Remote Sensing Data Processing fields, and in particular to a kind of Remote Sensing Data Processing method based on rough set.
Background technique
In recent ten years, Hi-spatial resolution remote sensing image is widely used for agricultural, forestry, ocean and environmental monitoring etc.
Field has huge economic value and social benefit.However, due to the scale of construction of Hi-spatial resolution remote sensing image
(Volume) greatly, data type (Variety) is more, abundant information, and interpretation analysis process is complicated, is also difficult to standard so far
Really, automatic terrain classification efficiently is carried out to Hi-spatial resolution remote sensing image.How to the big number of high spatial resolution remote sense
Become one of technological difficulties and the bottleneck for influencing its large-scale application according to progress terrain classification.With middle low resolution remote sensing images phase
Than Hi-spatial resolution remote sensing image texture is more abundant, shape is more obvious, and spatial relationship is more complicated.Existing technology
Usually described using spectrum, shape and textural characteristics in Hi-spatial resolution remote sensing image class different characteristics.However, this
It is characterized in low-level image feature, it is difficult to describe the geometry and structural information of atural object in high spatial resolution images comprehensively.In recent years,
Word packet model (Bag-of-Word, BOW) and topic model (Topic model) in text analyzing and scene understanding are drawn
Enter remote sensing fields.These methods extract the statistical information or semantic information of local feature by word packet model, and divide accordingly
The theme in Hi-spatial resolution remote sensing image is analysed, to achieve the purpose that classification.Existing feature extracting method is system mostly
Count feature, it is difficult to the accurately essential information of description ground class, it is difficult to realize Hi-spatial resolution remote sensing image automatic interpretation.How
For the complexity of the diversity of sensor, the variability of image-forming condition and ground target, high spatial resolution remote sense is extracted
The deep structure information of atural object in big data imperfectly describes ground class feature as far as possible, is high spatial resolution remote sense big data
In terrain classification key.In order to excavate better feature, people have to put into a large amount of energy and go research one good
Feature.And good feature exploitation generally requires the understanding for having very deep to problem, needs to grope repeatedly.Therefore it is required to instantly certainly
It is dynamic to generate suitable feature.Existing remote sensing satellite data processing system framework is generally stored by lower data application layer, data
Layer, data process&analysis layer, upper layer data application layer composition, whole system mainly pass through single system computing cluster and realize data
Processing and analysis.But as remote sensing satellite transmitting is more and more intensive, load data and application diversity are more and more significant, together
When remote sensing satellite data storage size rapidly increase, in particular near real-time processing remote sensing application, data to be treated
Amount is multiplied, and user is more more and more intense to the high-timeliness demand of data processing and application.
Summary of the invention
The remote sensing application that the present invention be directed to handle in particular near real-time, data volume to be treated are multiplied,
User is more more and more intense to the high-timeliness demand of data processing and application, a kind of designed remotely-sensed data based on rough set
Processing method.
A kind of Remote Sensing Data Processing method based on rough set, comprising the following steps:
M1 sets up GPS positioning information in the earth's surface known resource point for needing to handle remote sensing information;
M2, remotely-sensed data type choose multi-wavelength data, all-wave data as a kind of garbled data, and chosen spectrum band value is as two
Class garbled data chooses geography information as Feature Selection data;
M3 counts selected data block statistical nature according to Feature Selection data;
M4 sorts out selected data block statistical nature, a kind of screening of correlation data block statistical nature removal according to Feature Selection data
Similarity is lower than 64% data in data;
M5 carries out characterization processing to two class garbled datas;
M6 repeats step M4 and M5, removes the redundancy feature of selected data block representated by all types of Feature Selection data;
M7, counting all garbled data features in M6 is screening collection;
M8 carries out the topographical features or the screening of marine resources characteristic of data block according to the screening collection of two class garbled datas;
M9, the screening collection for carrying out a kind of garbled data to the data in step M8 screen, and carry out data block confirmation;
M10 is the screening collection of 1 and a kind of garbled data according to screening collection the selection result of the step M9 result to two class garbled datas
The block that the selection result is 0 carries out regulation extraction;
M11 carries out an iteration with the reversed screening collection rule rejected in step M7 of extracting of the rule of step M10;
Screening collection in M12, replacement step M8, M9 is that screening collection collects with an iteration screening, and subsequent screening is preferential to carry out double sieves
Selected works screening;
M13 repeats step M8 to M11, until having screened all data blocks;
M14, exports the selection result or rotation continues to screen from the first data block.
Preferably, the step M2 the following steps are included:
A1 removes geography information as Feature Selection data
A2 carries out expert knowledge library foundation, and using expert knowledge library wave band data vector set as Feature Selection data;
A3 is normalized the screening collection of step A2, is set as open set;
A4, copy step A3 result screening collection are used as the additional screening collection into step M7.
Preferably, the screening mode in the step M8 is to be carried out based on greedy search algorithm to screening collection condition
Matching, and matching sequence screening is carried out to data block.
Preferably, the step M9 the following steps are included:
B1 collects sieve number of entries according to screening to establish description maximum boundary domain;
B2, with the major class in critical value arrangement Boundary Region;
B3 enables probability match with Expressive Features and to match total probability greater than setting confidence level;
B4 carries out boundary filtering with all Expressive Features for meeting confidence level;
B5 filters out the data for meeting maximum boundary domain in all data blocks, is set as primary credible maximum side the selection result;
B6, the regular type of replacement screening collection are according to adaptation Boundary Region;
B7, with the confidence level judgement of the adaptation Boundary Region straight line step B3 of step B6;
B8 carries out postsearch screening with primary credible maximum side the selection result of the judging result of step B7 to step B5 and is exported
As a result.
Preferably, the expert knowledge library includes change value of the vegetation growth index on plural wave band, ocean neck
Domain is for marine resources in the change value of the plural wave band at telemetry end in time change.
Preferably, in the step M3 data block split the following steps are included:
C1 converts building granulating decomposition model according to SI-DWT;
C2 carries out alignment fractionation to data block, and stops SI-DWT transformation when second level is decomposed;
C3 splits particle data to second level and carries out characterization statement.
Substantial effect of the invention is to excavate remotely-sensed data using the reduction of rough set Fast Attribute, effectively improve distant
Feel data attribute reduction speed, and can simple cycle continuous updating rough set matching value accelerate subsequent excavation speed, to original
There are data to carry out continuing excavation.
Specific embodiment
Below by specific embodiment, technical scheme of the present invention will be further explained in detail.
Embodiment 1
A kind of Remote Sensing Data Processing method based on rough set, comprising the following steps:
M1 sets up GPS positioning information in the earth's surface known resource point for needing to handle remote sensing information;
M2, remotely-sensed data type choose multi-wavelength data, all-wave data as a kind of garbled data, and chosen spectrum band value is as two
Class garbled data chooses geography information as Feature Selection data;
M3 counts selected data block statistical nature according to Feature Selection data;
M4 sorts out selected data block statistical nature, a kind of screening of correlation data block statistical nature removal according to Feature Selection data
Similarity is lower than 64% data in data;
M5 carries out characterization processing to two class garbled datas;
M6 repeats step M4 and M5, removes the redundancy feature of selected data block representated by all types of Feature Selection data;
M7, counting all garbled data features in M6 is screening collection;
M8 carries out the topographical features or the screening of marine resources characteristic of data block according to the screening collection of two class garbled datas;
M9, the screening collection for carrying out a kind of garbled data to the data in step M8 screen, and carry out data block confirmation;
M10 is the screening collection of 1 and a kind of garbled data according to screening collection the selection result of the step M9 result to two class garbled datas
The block that the selection result is 0 carries out regulation extraction;
M11 carries out an iteration with the reversed screening collection rule rejected in step M7 of extracting of the rule of step M10;
Screening collection in M12, replacement step M8, M9 is that screening collection collects with an iteration screening, and subsequent screening is preferential to carry out double sieves
Selected works screening;
M13 repeats step M8 to M11, until having screened all data blocks;
M14, exports the selection result or rotation continues to screen from the first data block.
The step M2 the following steps are included:
A1 removes geography information as Feature Selection data
A2 carries out expert knowledge library foundation, and using expert knowledge library wave band data vector set as Feature Selection data;
A3 is normalized the screening collection of step A2, is set as open set;
A4, copy step A3 result screening collection are used as the additional screening collection into step M7.
Screening mode in the step M8 is to be matched to screening collection condition, and right based on greedy search algorithm
Data block carries out matching sequence screening.
Greedy search algorithm is not that can obtain total optimization solution to all problems, it is important to the choosing of greedy search algorithm
Select, greedy search algorithm must have markov property, i.e. the pervious process of some state will not influence later state, only with work as
Preceding state is related.The total optimization solution that can usually prove problem first and is coveted since greedy search
After greedy search, former problem reduction is the smaller similar subproblem of a scale.Then, it is proved with mathematical induction, by each
Greedy search is walked, a total optimization solution of problem finally can be obtained.
The step M9 the following steps are included:
B1 collects sieve number of entries according to screening to establish description maximum boundary domain;
B2, with the major class in critical value arrangement Boundary Region;
B3 enables probability match with Expressive Features and to match total probability greater than setting confidence level;
B4 carries out boundary filtering with all Expressive Features for meeting confidence level;
B5 filters out the data for meeting maximum boundary domain in all data blocks, is set as primary credible maximum side the selection result;
B6, the regular type of replacement screening collection are according to adaptation Boundary Region;
B7, with the confidence level judgement of the adaptation Boundary Region straight line step B3 of step B6;
B8 carries out postsearch screening with primary credible maximum side the selection result of the judging result of step B7 to step B5 and is exported
As a result.
Confidence level refers to that particular individual treats the degree that particular proposition authenticity is believed, that is, probability is believed individual
Read rational measurement.What probability the confidence level explanation of probability shows event itself there is no, and why event is assigned with generally
Rate is possessed conviction evidence in the number of people brain for assign probability.Confidence level refers to that population parameter value falls in sample statistics value
Probability in a certain area;And confidence interval refers under a certain confidence level, error model between sample statistics value and population parameter value
It encloses.Confidence interval is bigger, and confidence level is higher.
And maximum boundary domain and adaptation Boundary Region are then that result can be generated in data for the special characteristic in set
Maximum and minimum value, matching when due to data completely random thus introduce confidence level come weigh maximum boundary domain with
The matching for adapting to Boundary Region is horizontal.
The expert knowledge library includes change value of the vegetation growth index on plural wave band, and marine field is ocean money
Source is in time change in the change value of the plural wave band at telemetry end.
In the step M3 data block split the following steps are included:
C1 converts building granulating decomposition model according to SI-DWT;
C2 carries out alignment fractionation to data block, and stops SI-DWT transformation when second level is decomposed;
C3 splits particle data to second level and carries out characterization statement.
Granulation calculates the calculating and operation for referring to executing to information.And the basis of Granule Computing is mentioned based on rough set
A kind of detailed-oriented partition structure out, the practical Granule Computing of the present invention facilitate follow-up data to screen for splitting data block roughly
Work is matched, guarantees that the same screening that edge effect can be avoided to bring in screening for exporting discontinuous cutting is lost in screening
Effect, or secondary granulation is carried out to basic data by way of final election granulation, the mode of primary screening result is overlapped to avoid counting
According to edge screening effect.
Claims (6)
1. a kind of Remote Sensing Data Processing method based on rough set, which comprises the following steps:
M1 sets up GPS positioning information in the earth's surface known resource point for needing to handle remote sensing information;
M2, remotely-sensed data type choose multi-wavelength data, all-wave data as a kind of garbled data, and chosen spectrum band value is as two
Class garbled data chooses geography information as Feature Selection data;
M3 counts selected data block statistical nature according to Feature Selection data;
M4 sorts out selected data block statistical nature, a kind of screening of correlation data block statistical nature removal according to Feature Selection data
Similarity is lower than 64% data in data;
M5 carries out characterization processing to two class garbled datas;
M6 repeats step M4 and M5, removes the redundancy feature of selected data block representated by all types of Feature Selection data;
M7, counting all garbled data features in M6 is screening collection;
M8 carries out the topographical features or the screening of marine resources characteristic of data block according to the screening collection of two class garbled datas;
M9, the screening collection for carrying out a kind of garbled data to the data in step M8 screen, and carry out data block confirmation;
M10 is the screening collection of 1 and a kind of garbled data according to screening collection the selection result of the step M9 result to two class garbled datas
The block that the selection result is 0 carries out regulation extraction;
M11 carries out an iteration with the reversed screening collection rule rejected in step M7 of extracting of the rule of step M10;
Screening collection in M12, replacement step M8, M9 is that screening collection collects with an iteration screening, and subsequent screening is preferential to carry out double sieves
Selected works screening;
M13 repeats step M8 to M11, until having screened all data blocks;
M14, exports the selection result or rotation continues to screen from the first data block.
2. a kind of Remote Sensing Data Processing method based on rough set according to claim 1, which is characterized in that the step
Rapid M2 the following steps are included:
A1 removes geography information as Feature Selection data
A2 carries out expert knowledge library foundation, and using expert knowledge library wave band data vector set as Feature Selection data;
A3 is normalized the screening collection of step A2, is set as open set;
A4, copy step A3 result screening collection are used as the additional screening collection into step M7.
3. a kind of Remote Sensing Data Processing method based on rough set according to claim 1, which is characterized in that the step
Screening mode in rapid M8 is to be matched to screening collection condition, and carry out matching row to data block based on greedy search algorithm
Sequence screening.
4. a kind of Remote Sensing Data Processing method based on rough set according to claim 1, which is characterized in that the step
Rapid M9 the following steps are included:
B1 collects sieve number of entries according to screening to establish description maximum boundary domain;
B2, with the major class in critical value arrangement Boundary Region;
B3 enables probability match with Expressive Features and to match total probability greater than setting confidence level;
B4 carries out boundary filtering with all Expressive Features for meeting confidence level;
B5 filters out the data for meeting maximum boundary domain in all data blocks, is set as primary credible maximum side the selection result;
B6, the regular type of replacement screening collection are according to adaptation Boundary Region;
B7, with the confidence level judgement of the adaptation Boundary Region straight line step B3 of step B6;
B8 carries out postsearch screening with primary credible maximum side the selection result of the judging result of step B7 to step B5 and is exported
As a result.
5. a kind of Remote Sensing Data Processing method based on rough set according to claim 2, which is characterized in that described is special
Family's knowledge base includes change value of the vegetation growth index on plural wave band, marine field be marine resources in time change
The change value of the plural wave band at telemetry end.
6. a kind of Remote Sensing Data Processing method based on rough set according to claim 2, which is characterized in that the step
In rapid M3 data block split the following steps are included:
C1 converts building granulating decomposition model according to SI-DWT;
C2 carries out alignment fractionation to data block, and stops SI-DWT transformation when second level is decomposed;
C3 splits particle data to second level and carries out characterization statement.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811369570.2A CN109542993B (en) | 2018-11-16 | 2018-11-16 | Remote sensing data processing method based on rough set |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811369570.2A CN109542993B (en) | 2018-11-16 | 2018-11-16 | Remote sensing data processing method based on rough set |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109542993A true CN109542993A (en) | 2019-03-29 |
CN109542993B CN109542993B (en) | 2022-07-22 |
Family
ID=65847779
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811369570.2A Active CN109542993B (en) | 2018-11-16 | 2018-11-16 | Remote sensing data processing method based on rough set |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109542993B (en) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120183225A1 (en) * | 2010-11-24 | 2012-07-19 | Indian Statistical Institute | Rough wavelet granular space and classification of multispectral remote sensing image |
-
2018
- 2018-11-16 CN CN201811369570.2A patent/CN109542993B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120183225A1 (en) * | 2010-11-24 | 2012-07-19 | Indian Statistical Institute | Rough wavelet granular space and classification of multispectral remote sensing image |
CN103052962A (en) * | 2010-11-24 | 2013-04-17 | 印度统计学院 | Rough wavelet granular space and classification of multispectral remote sensing image |
Non-Patent Citations (2)
Title |
---|
王婧等: "基于粗糙集规则提取的面向对象树种分类方法", 《遥感信息》 * |
陈敏: "基于粗糙集核计算的遥感图像波段优选算法", 《宁德师专学报(自然科学版)》 * |
Also Published As
Publication number | Publication date |
---|---|
CN109542993B (en) | 2022-07-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhong et al. | An end-to-end dense-inceptionnet for image copy-move forgery detection | |
CN108363714A (en) | A kind of method and system for the ensemble machine learning for facilitating data analyst to use | |
CN110135167B (en) | Edge computing terminal security level evaluation method for random forest | |
CN102208037B (en) | Hyper-spectral image classification method based on Gaussian process classifier collaborative training algorithm | |
CN113922985B (en) | Network intrusion detection method and system based on ensemble learning | |
CN109299185B (en) | Analysis method for convolutional neural network extraction features aiming at time sequence flow data | |
CN111143838B (en) | Database user abnormal behavior detection method | |
CN110458201A (en) | A kind of remote sensing image object-oriented classification method and sorter | |
CN102243628A (en) | Mineralizing case reasoning model and method | |
CN113326862A (en) | Audit big data fusion clustering and risk data detection method, medium and equipment | |
CN117453764A (en) | Data mining analysis method | |
CN109542949B (en) | Formal vector-based decision information system knowledge acquisition method | |
CN110765329A (en) | Data clustering method and electronic equipment | |
Shi et al. | Optimization and data mining for fracture prediction in geosciences | |
CN109542993A (en) | A kind of Remote Sensing Data Processing method based on rough set | |
CN115001820A (en) | Data processing method and device, storage medium and electronic equipment | |
Huiling et al. | Coal-Rock Interface Recognition Method Based on Image Recognition | |
CN114969511A (en) | Content recommendation method, device and medium based on fragments | |
CN107239693A (en) | A kind of analysis method and system based on password coding rule | |
Dang-Nhu | Evaluating disentanglement of structured representations | |
Rajeswari et al. | Chi-square mapreduce model for agricultural data | |
CN111291376A (en) | Web vulnerability verification method based on crowdsourcing and machine learning | |
Xie et al. | Models and Features with Covariate Shift Adaptation for Suspicious Network Event Recognition | |
Dai et al. | Vibration Signal Denoising for Coal-Gangue Impacting the Tail Beam of Hydraulic Support Based on Improved ALO and VMD | |
CN114488261B (en) | LHAASO experiment-based low-energy section gamma proton high-energy particle identification method |
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 |