CN105184322B - A kind of multidate image sorting technique based on incremental integration study - Google Patents
A kind of multidate image sorting technique based on incremental integration study Download PDFInfo
- Publication number
- CN105184322B CN105184322B CN201510582557.5A CN201510582557A CN105184322B CN 105184322 B CN105184322 B CN 105184322B CN 201510582557 A CN201510582557 A CN 201510582557A CN 105184322 B CN105184322 B CN 105184322B
- Authority
- CN
- China
- Prior art keywords
- sample
- training
- incremental
- strong classifier
- multidate
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
A kind of multidate image sorting technique based on incremental integration study, is related to Multitemporal Remote Sensing Images sorting technique field.The present invention is to solve the problems, such as that multidate image nicety of grading is low.The present invention introduces incremental learning on the basis of integrated study, constructs multidate grader, realizes continuously on-line study.First, then the image data of each phase obtains strong classifier C respectively as the basic kernel function of support vector machines algorithm by Ensemble Learning Algorithms0；Secondly, new training data is introduced, original training dataset is updated, obtains incremental data set, then strong classifier C is obtained by Ensemble Learning Algorithms1；New training data is introduced successively, and strong classifier C is obtained by Ensemble Learning Algorithmsn；Last each strong classifier is added to obtain final classification device, is used for the classification of test sample.The present invention classifies suitable for multidate image.
Description
Technical field
The present invention relates to Multitemporal Remote Sensing Images sorting technique fields.
Background technology
As the speed of remote sensing science and technology gathered data is continuously increased, the temporal resolution of remotely-sensed data also constantly carries
Height, research contents have been progressed into the period of multidate image by Mono temporal image period.Multidate image can be reflected in one
Determine each target in region has the characteristics that data volume is big, dimension is high and state is constantly updated in the state change of different moments.
By the research to multi-temporal remote sensing image, land cover pattern situation of change etc. can be monitored, resource, ring to realization to the earth
Border etc. is monitored.Therefore research multidate image sorting technique, has great importance.
Current method can not efficiently solve the problem of sample diversity and complex boundary in multidate image so that
The nicety of grading of multi-temporal remote sensing image is relatively low.
Invention content
The present invention is provided a kind of based on incremental integration study to solve the problems, such as that multidate image nicety of grading is low
Multidate image sorting technique.
A kind of multidate image sorting technique based on incremental integration study, it is realized by following steps：
Step 1: reading multidate image data, handmarking is carried out to it, monitoring data is obtained, according to monitoring data
It determines marker samples, is randomly choosed to obtain training sample and increment sample set by the sample in marking, remaining is as test sample；
Step 2: integrating strong classifier C by Ensemble Learning Algorithms0, the iterations that incremental learning is arranged are N, N
For positive integer；
Step 3: stochastical sampling obtains increment sample in increment sample set, i.e. newly-increased training sample, using newly-increased
Sample replaces the sample of the distribution of weights minimum in Ensemble Learning Algorithms, updates the weight of training sample and makes it to be uniformly distributed；
Step 4: update training dataset, and determine strong classifier C by Ensemble Learning Algorithmsn, n=1,2 ..., N；
Step 5: step 3 is repeated to step 4, until reaching the iterations N of incremental learning, by each strong classifier
CnIt is added, constitutes final classification device；
Step 6: classifying to the test sample of step 1 using final classification device, and survey is provided using voting mechanism
The classification results of sample sheet.
Described in step 2 and step 4 by Ensemble Learning Algorithms integrate strong classifier be utilize Adaboost collection
It is realized at algorithm, and the Weak Classifier in the Adaboost Integrated Algorithms is support vector machines algorithm.
The specific method of strong classifier is obtained using Adaboost Integrated Algorithms is：
It is opposite when then each using the training data under each phase as the kernel function in support vector machines
Weak Classifier of the support vector machines for the Kernel answered as Adaboost algorithm；According to Adaboost algorithm to instruction
Practice data and be iterated formula training, if iterations are T；T is positive integer；
First, the distribution of weights of training sample is set as being uniformly distributed；
Then, all training samples are sampled according to sample rate, the sample after being sampled, to the sample after sampling
Originally it is trained to obtain Weak Classifier 1；
Finally tested with all training samples；
For then improving corresponding weight by the sample of mistake point in test, the sample correctly classified then reduces its weight, right
Training sample after adjustment weight is still sampled according to corresponding sample rate, is trained to obtain weak point to the sample after sampling
Class device 2；And so on, iteration T times.
In the present invention, incremental learning instructs the sample of continuous renewal on the basis of retaining original training pattern
Practice study, forms a continuous learning system, improve the learning ability of system.The present invention passes through to new sample
It practises, solves the problems, such as to be learnt in face of diversity sample, while the sample of complex boundary can be learnt, overcome
The classification problems such as sample diversity and complex boundary present in multidate image, can effectively improve the precision of classification.
Description of the drawings
Fig. 1 is a kind of flow signal of multidate image sorting technique learnt based on incremental integration of the present invention
Figure.
Specific implementation mode
Specific implementation mode one, a kind of multidate image sorting technique based on incremental integration study, it is by following steps
It realizes：
Step 1: reading multidate image data, handmarking is carried out to it, monitoring data is obtained, according to monitoring data
It determines marker samples, is randomly choosed to obtain training sample and increment sample set by the sample in marking, remaining is as test sample；
Step 2: integrating strong classifier C by Ensemble Learning Algorithms0, the iterations that incremental learning is arranged are N；
Step 3: stochastical sampling obtains increment sample in increment sample set, i.e. newly-increased training sample, using newly-increased
Sample replaces the sample of the distribution of weights minimum in Ensemble Learning Algorithms, updates the weight of training sample and makes it to be uniformly distributed；
Step 4: update training dataset, and determine strong classifier C by Ensemble Learning Algorithmsn, n=1,2 ..., N；
Step 5: step 3 is repeated to step 4, until reaching the iterations of incremental learning, by each strong classifier Cn
It is added, constitutes final classification device；
Step 6: classifying to the test sample of step 1 using final classification device, and survey is provided using voting mechanism
The classification results of sample sheet.
Specific implementation mode two, present embodiment are that one kind described in specific implementation mode one is based on incremental integration
The multidate image sorting technique of habit further limits, being integrated by Ensemble Learning Algorithms described in step 2 and step 4
To strong classifier be using Adaboost Integrated Algorithms realize, and the Weak Classifier in the Adaboost Integrated Algorithms be support to
Amount machine SVM algorithm.
Specific implementation mode three, present embodiment are that one kind described in specific implementation mode two is based on incremental integration
The multidate image sorting technique of habit further limits, and the specific method of strong classifier is obtained using Adaboost Integrated Algorithms
It is：
It is opposite when then each using the training data under each phase as the kernel function in support vector machines
Weak Classifier of the support vector machines for the Kernel answered as Adaboost algorithm；According to Adaboost algorithm to instruction
Practice data and be iterated formula training, if iterations are T；T is positive integer；
First, the distribution of weights of training sample is set as being uniformly distributed；
Then, all training samples are sampled according to sample rate according to weight distribution, the sample after being sampled,
Sample after sampling is trained to obtain Weak Classifier 1；
Finally tested with all training samples；
For then improving corresponding weight by the sample of mistake point in test, the sample correctly classified then reduces its weight, right
Training sample after adjustment weight is still sampled according to corresponding sample rate, is trained to obtain weak point to the sample after sampling
Class device 2；And so on, iteration T times.
Claims (3)
1. a kind of multidate image sorting technique based on incremental integration study, it is characterized in that：It is realized by following steps：
Step 1: reading multidate image data, handmarking is carried out to it, is obtained monitoring data, is determined according to monitoring data
Marker samples are randomly choosed to obtain training sample and increment sample set by the sample in marking, remaining is as test sample；
Step 2: integrating strong classifier C by Ensemble Learning Algorithms0, the iterations that incremental learning is arranged are N, and N is just
Integer；
Step 3: stochastical sampling obtains increment sample, i.e. newly-increased training sample in increment sample set, newly-increased sample is utilized
Instead of the sample of the distribution of weights minimum in Ensemble Learning Algorithms, updating the weight of training sample makes it to be uniformly distributed；
Step 4: update training dataset, and determine strong classifier C by Ensemble Learning Algorithmsn, n=1,2 ..., N；
Step 5: step 3 is repeated to step 4, until reaching the iterations N of incremental learning, by each strong classifier CnPhase
Add, constitutes final classification device；
Step 6: classifying to the test sample of step 1 using final classification device, and test specimens are provided using voting mechanism
This classification results.
2. a kind of multidate image sorting technique based on incremental integration study according to claim 1, it is characterised in that
Described in step 2 and step 4 to integrate strong classifier by Ensemble Learning Algorithms be real using Adaboost Integrated Algorithms
It is existing, and the Weak Classifier in the Adaboost Integrated Algorithms is support vector machines algorithm.
3. a kind of multidate image sorting technique based on incremental integration study according to claim 2, it is characterised in that
The specific method of strong classifier is obtained using Adaboost Integrated Algorithms is：
It is corresponding when then each using the training data under each phase as the kernel function in support vector machines
Weak Classifier of the support vector machines of Kernel as Adaboost algorithm；According to Adaboost algorithm to training number
According to formula training is iterated, if iterations are T；T is positive integer；
First, the distribution of weights of training sample is set as being uniformly distributed；
Then, all training samples are sampled according to sample rate, the sample after being sampled, to the sample after sampling into
Row training obtains Weak Classifier 1；
Finally tested with all training samples；
For then improving corresponding weight by the sample of mistake point in test, the sample correctly classified then reduces its weight, to adjustment
Training sample after weight is still sampled according to corresponding sample rate, is trained to obtain Weak Classifier to the sample after sampling
2；And so on, iteration T times.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510582557.5A CN105184322B (en) | 2015-09-14 | 2015-09-14 | A kind of multidate image sorting technique based on incremental integration study |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510582557.5A CN105184322B (en) | 2015-09-14 | 2015-09-14 | A kind of multidate image sorting technique based on incremental integration study |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105184322A CN105184322A (en) | 2015-12-23 |
| CN105184322B true CN105184322B (en) | 2018-11-02 |
Family
ID=54906386
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510582557.5A Active CN105184322B (en) | 2015-09-14 | 2015-09-14 | A kind of multidate image sorting technique based on incremental integration study |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105184322B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106548136A (en) * | 2016-10-19 | 2017-03-29 | 中科院成都信息技术股份有限公司 | A kind of wireless channel scene classification method |
| CN107508866B (en) * | 2017-08-08 | 2020-10-02 | 重庆大学 | Method for reducing transmission consumption of mobile equipment end neural network model updating |
| CN107563441A (en) * | 2017-09-01 | 2018-01-09 | 深圳市唯特视科技有限公司 | A kind of urban target detection method based on grader |
| CN108830312B (en) * | 2018-06-01 | 2022-09-09 | 苏州中科天启遥感科技有限公司 | Integrated learning method based on sample adaptive expansion |
| CN111488917A (en) * | 2020-03-19 | 2020-08-04 | 天津大学 | Garbage image fine-grained classification method based on incremental learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101937510A (en) * | 2010-09-14 | 2011-01-05 | 南京信息工程大学 | Fast incremental learning method based on quasi-Haar and AdaBoost classifier |
| CN103258217A (en) * | 2013-05-15 | 2013-08-21 | 中国科学院自动化研究所 | Pedestrian detection method based on incremental learning |
| CN103780501A (en) * | 2014-01-03 | 2014-05-07 | 濮阳职业技术学院 | Peer-to-peer network traffic identification method of inseparable-wavelet support vector machine |
| CN104102837A (en) * | 2014-07-11 | 2014-10-15 | 天津工业大学 | Incremental learning integrated algorithm for soft measurement modeling |
| CN104462797A (en) * | 2014-11-24 | 2015-03-25 | 天津工业大学 | Increment integration algorithm used for procedure parameter online testing |
-
2015
- 2015-09-14 CN CN201510582557.5A patent/CN105184322B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101937510A (en) * | 2010-09-14 | 2011-01-05 | 南京信息工程大学 | Fast incremental learning method based on quasi-Haar and AdaBoost classifier |
| CN103258217A (en) * | 2013-05-15 | 2013-08-21 | 中国科学院自动化研究所 | Pedestrian detection method based on incremental learning |
| CN103780501A (en) * | 2014-01-03 | 2014-05-07 | 濮阳职业技术学院 | Peer-to-peer network traffic identification method of inseparable-wavelet support vector machine |
| CN104102837A (en) * | 2014-07-11 | 2014-10-15 | 天津工业大学 | Incremental learning integrated algorithm for soft measurement modeling |
| CN104462797A (en) * | 2014-11-24 | 2015-03-25 | 天津工业大学 | Increment integration algorithm used for procedure parameter online testing |
Non-Patent Citations (5)
| Title |
|---|
| A review on time series data mining;Fu T等;《Engineering Applications of Artificial Intelligence》;20111231;全文 * |
| Sequential cascade classification of image time series by exploiting multiple pairwise change detection;Demir B等;《IEEE》;20131231;全文 * |
| 一种用于软测量建模的增量学习集成算法;田慧欣等;《控制与决策》;20150831;第30卷(第8期);全文 * |
| 基于时间序列环境卫星影像的作物分类识别;李鑫川等;《农业工程学报》;20131231;全文 * |
| 基于神经网络集成的增量学习;林民龙;《万方数据》;20121130;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105184322A (en) | 2015-12-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105184322B (en) | A kind of multidate image sorting technique based on incremental integration study | |
| CN110211097B (en) | Crack image detection method based on fast R-CNN parameter migration | |
| de Lima et al. | Petrographic microfacies classification with deep convolutional neural networks | |
| CN104102929B (en) | Hyperspectral remote sensing data classification method based on deep learning | |
| CN105243398B (en) | The method of improvement convolutional neural networks performance based on linear discriminant analysis criterion | |
| CN104484681B (en) | Hyperspectral Remote Sensing Imagery Classification method based on spatial information and integrated study | |
| CN104680542B (en) | Remote sensing image variation detection method based on on-line study | |
| Yu et al. | Application of hierarchical clustering, singularity mapping, and Kohonen neural network to identify Ag-Au-Pb-Zn polymetallic mineralization associated geochemical anomaly in Pangxidong district | |
| CN107644212B (en) | Hyperspectral remote sensing image classification method, device and storage device | |
| CN109815979B (en) | Weak label semantic segmentation calibration data generation method and system | |
| CN105844228A (en) | Remote sensing image cloud detection method based on convolution nerve network | |
| CN103745233B (en) | The hyperspectral image classification method migrated based on spatial information | |
| CN104331899B (en) | A kind of SAR image registration method and device | |
| CN103886342A (en) | Hyperspectral image classification method based on spectrums and neighbourhood information dictionary learning | |
| CN104463249A (en) | Remote sensing image airport detection method based on weak supervised learning frame | |
| Xu et al. | Deep learning of rock microscopic images for intelligent lithology identification: Neural network comparison and selection | |
| CN105023024B (en) | A kind of Classifying Method in Remote Sensing Image and system based on regularization set metric learning | |
| CN105913090B (en) | SAR image objective classification method based on SDAE-SVM | |
| CN105160351B (en) | Semi-supervised hyperspectral classification method based on anchor point sparse graph | |
| Bhuyan et al. | Generating multi-temporal landslide inventories through a general deep transfer learning strategy using HR EO data | |
| CN112132014A (en) | Target re-identification method and system based on non-supervised pyramid similarity learning | |
| Verbovšek et al. | GIS-assisted classification of litho-geomorphological units using Maximum Likelihood Classification, Vipava Valley, SW Slovenia | |
| CN104268557B (en) | Polarization SAR sorting technique based on coorinated training and depth S VM | |
| Trott et al. | Random forest rock type classification with integration of geochemical and photographic data | |
| CN104463207A (en) | Knowledge self-encoding network and polarization SAR image terrain classification method thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |