CN109784374A - Multi-angle of view clustering method based on adaptive neighbor point - Google Patents
Multi-angle of view clustering method based on adaptive neighbor point Download PDFInfo
- Publication number
- CN109784374A CN109784374A CN201811559259.4A CN201811559259A CN109784374A CN 109784374 A CN109784374 A CN 109784374A CN 201811559259 A CN201811559259 A CN 201811559259A CN 109784374 A CN109784374 A CN 109784374A
- Authority
- CN
- China
- Prior art keywords
- matrix
- adjacent map
- angle
- follows
- sample data
- 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
Links
Abstract
The present invention provides a kind of multi-angle of view clustering methods based on adaptive neighbor point.Firstly, calculating the eigenmatrix under sample different perspectives to be clustered, line number of going forward side by side Data preprocess solves the constructed function model with order constraint using pretreated data and initialization weight, obtains initial adjacent map;Then, the multi-angle of view clustering function model based on self study weight is constructed, and is iterated solution, obtains optimal adjacent map;The connected component of optimal adjacent map, i.e., final cluster result are obtained finally, solving.The method of the present invention can use the information of sample different perspectives to be processed, obtain better Clustering Effect;In addition, the method for the present invention can effectively avoid introducing the hyper parameter of no calligraphy learning, for unsupervised cluster task, there is better robustness.
Description
Technical field
The invention belongs to machine learning techniques fields, and in particular to a kind of multi-angle of view cluster side based on adaptive neighbor point
Method.
Background technique
Multi-angle of view study is one of the hot research field in machine learning in recent years, is widely used in scene analysis, figure
As multiple fields such as classification and Web information processings.In some cases, doing cluster with the feature at single visual angle can achieve difference
But if strong man's meaning comes from multiple Viewing-angle informations as a result, reasonably integrating, then can obtain preferably as a result, such method
It is referred to as multi-angle of view clustering method.
Currently, existing multi-angle of view clustering method is divided into three classes on the whole: method based on tensor, based on son
The method in space and multi-angle of view clustering method based on figure.Wherein, the more other two methods of multi-angle of view clustering method based on figure
Better effect can be obtained, following steps are generally comprised: firstly, the feature for each visual angle is distinguished composition and calculated
Then similarity matrix with weight or penalizes item to obtain instruction vector to integrate these similarity matrixs, finally, referring to these
Show that vector for input, is clustered using K Mean Method.Kumar et al. is in document " Kumar A, Rai P, Daume H.Co-
regularized multi-view spectral clustering[A].Advances in Neural Information
A kind of collaboration specification Spectral Clustering is proposed in Processing Systems [C] .2011.1413-1421. ", this method is
First is introduced into the collaboration specification thought in semi-supervised learning in spectral clustering, so that the cluster result at each visual angle reaches
To consistent.Li et al. people is in document " Li Y, Nie F, Huang H, et al.Large-Scale Multi-View Spectral
A kind of extensive more views are proposed in Clustering via Bipartite Graph [C] .AAAI.2015.2750-2756. "
Angular spectrum clustering method, this method are not coupled adjoint point to be distributed for all the points, but choose one when composition
It represents a little, then carrys out composition according to this connection for representing point and other points, greatly reduce computation complexity, be suitable for
Large-scale data set.Cai et al. is in document " Cai X, Nie F, Huang H, et al.Heterogeneous image
feature integration via multi-modal spectral clustering[A].Computer Vision
and Pattern Recognition(CVPR),2011[C].IEEE Conference on.IEEE,2011.1977-
A kind of multi-mode Spectral Clustering is proposed in 1984. ", integrates multi-mode number using a unified Laplacian Matrix
According to, and non-negative orthogonality constraint is added to carry out the robustness of improvement method to it.Although based on the method for figure compared with preferable cluster can be obtained
As a result, but also have its limitation: firstly, similarity matrix calculating and subsequent cluster process be separation, and the latter very according to
Rely the former, therefore, insecure similarity matrix will lead to bad cluster result, also, since result also relies on similar K
The number of neighbour in neighbour, parameter settings, traditional composition k nearest neighbor or the Gaussian function such as selection of bandwidth in Gaussian function
Method can hardly obtain best figure;Second, with weight or the multi-angle of view integration method of item is penalized to have parameter to need to adjust
Section, this is very unfavorable for unsupervised cluster task.
Summary of the invention
For overcome the deficiencies in the prior art, the present invention provides a kind of multi-angle of view cluster side based on adaptive neighbor point
Method.Firstly, calculating the eigenmatrix under sample different perspectives to be clustered, line number of going forward side by side Data preprocess utilizes pretreated number
The constructed function model with order constraint is solved according to initialization weight, obtains initial adjacent map;Then, building is based on self-study
The multi-angle of view clustering function model of weight is practised, and is iterated solution, obtains optimal adjacent map;Optimal neighbour is obtained finally, solving
The connected component of map interlinking, i.e., final cluster result.The method of the present invention can use the information of sample different perspectives to be processed, take
Obtain better Clustering Effect;In addition, the method for the present invention can effectively avoid introducing the hyper parameter of no calligraphy learning, for unsupervised
Cluster task has better robustness.
A kind of multi-angle of view clustering method based on adaptive neighbor point, it is characterised in that steps are as follows:
Step 1: for sample to be clustered, the eigenmatrix of its different perspectives being calculated using visual signature operator, remembersFor the eigenmatrix at v-th of visual angle, v=1 ..., V, V is total number viewpoints, and n is sample data
Number,For feature vector of i-th of sample data under v-th of visual angle, the visual signature operator include SIFT, HOG,
LBP,ColorMatrix;
Then, the sample data under different perspectives is pre-processed, obtains pretreated sample data, it may be assumed that
Wherein,It indicatesThe mean value of middle all elements, σ () indicate standard deviation;
The weight for initializing each visual angle isIt is as follows to construct the function model with order constraint:
Wherein, S ∈ Rn×nIndicate similarity matrix, i.e. adjacent map between sample data, sijI-th in representing matrix S
Row j column element, siThe i-th row vector in representing matrix S, Section 2For penalty term, α > 0, LsIt is corresponding for similarity matrix S
Laplacian Matrix, c indicate clustering cluster number, rank () expression seek rank of matrix;
Above-mentioned function model is solved using method of Lagrange multipliers, obtains similarity matrix S, i.e., it is initial adjacent
Figure.
Step 2: multi-angle of view clustering function model of the building based on self study weight is as follows:
Wherein, the calculation formula of parameter alpha are as follows:
Wherein, fiFor the corresponding instruction vector of i-th of sample data, n instruction vector constitutes oriental matrix F, self study
Weight wvCalculation formula are as follows:
Function model (3) are solved using following alternative manner, obtain final optimal adjacent map S, specifically:
Step a: the initial adjacent map S and its corresponding Laplacian Matrix L obtained using step 1S, by following public affairs
Formula initializes oriental matrix F:
Step b: fixed wvAnd F, adjacent map S is updated as follows:
Wherein, diFor by dijThe n-dimensional vector of composition, j=1 ..., n;
Step c: fixed S updates weight parameter w by formula (6)v, and update instruction matrix F as follows:
Step d: repeating step b-c, until the difference for the functional value that twice adjacent calculation obtains less than 0.0001, stops meter
It calculates, the S obtained at this time is final optimal adjacent map.
Step 3: using being documented in document " Tarjan.Depth-first search and linear graph
Algorithms [C] .Symposium on Switching&Automata Theory.IEEE, the Tarjan's in 2008. "
Algorithm calculates the connected component in the optimal adjacent map S that step 2 obtains, as final cluster result.
The beneficial effects of the present invention are: being difficult to determine without introducing due to clustering frame using the multi-angle of view of self study weight
Plan and the hyper parameter sensitive to data set, it is more stronger than previous methods practicability;It, can be effective due to continuous iteration optimization adjacent map
Remote item bring adverse effect is avoided, greatly promotes method to the robustness of abnormal point.
Detailed description of the invention
Fig. 1 is a kind of multi-angle of view clustering method basic flow chart based on adaptive neighbor point of the invention
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples, and the present invention includes but are not limited to following implementations
Example.
As shown in Figure 1, being realized substantially the present invention provides a kind of multi-angle of view clustering method based on adaptive neighbor point
Process is as follows:
1, feature collection, data normalization and the initial adjacent map of building
For sample to be clustered, the present invention first collects its feature from different perspectives, i.e., according to data self attributes
Be calculated the feature of its different perspectives using visual signature operator, the visual signature operator include SIFT, HOG, LBP,
These visual signature operators for generally using of ColorMatrix.Enabling V is collected Viewing-angle information number,For the eigenmatrix at v-th of visual angle,N is the number of sample data,
Dv is the feature dimensions at v-th of visual angle,For feature vector of i-th of sample under v-th of visual angle,
It is feature vectorJ-th of element, j=1 ..., dv.The eigenmatrix number of samples at i.e. each visual angle is identical, but its feature
Number is different.
Firstly, being pre-processed to initial data, i.e., to each visual angle of each sample dataAccording to following formula
It is handled, obtains pretreated sample data
Wherein,It indicatesPlace vectorThe mean value of middle all elements, σ () indicate standard deviation.
Enable each sample data that may only be connected with other 9 sample datas, similarity s therebetweenijIt indicates, j
≠ i, j=1 ..., n.There is biggish similarity apart from close sample data, apart from remote sample data similarity with regard to small.In order to
Simplicity carries out distance metric using the Euclidean distance between two sample datas, initializes the weight at each visual angleThen
Similarity s can be obtained by solving following problemsij:
Wherein, S ∈ Rn×nIndicate similarity matrix, i.e. adjacent map between sample data, the i-th row j column element in matrix S
As sij, Section 2For penalty term, α > 0, effect is that function is avoided to fall into trivial solution, i.e., for some sample number
According to being only assigned similarity with the sample data of its arest neighbors is 1, and the similarity of other sample datas is zero.
Method of the tradition based on figure is finally required using K mean cluster method, but it is sensitive to initial cluster center, is
This problem is avoided, the present invention constrains model according to the following characteristics of similarity matrix, it may be assumed that if similarity matrix
S is nonnegative matrix, then its corresponding Laplacian Matrix LS0 characteristic value tuple be equal to similarity matrix S in connection point
The number of amount.Wherein, LS=D-S, D are to angle matrix, each element is to correspond to row in similarity matrix S on diagonal line
The sum of all row elements.If Laplacian Matrix LSMeet order constraint rank (LS)=n-c, c indicate the number of clustering cluster, that
It can use the structure chart with good nature and data point directly clustered.Based on considerations above, problem (2) are increased
Order constraint, specifically:
Wherein, LsFor Laplacian Matrix, rank of matrix is sought in rank () expression.It is solved using method of Lagrange multipliers public
Formula (12), obtains similarity matrix S, i.e., initial adjacent map.
2, adaptive neighbor point multi-angle of view data fusion seeks optimal adjacent map
Cluster is unsupervised task, data to be clustered do not have any label, can not be learnt if had in method
Parameter, then its ease for use will substantially reduce.Task is clustered for multi-angle of view, previous method requires to introduce additional hyper parameter
To distribute weight for each visual angle, to merge the information at each visual angle.It is poly- that the invention proposes a kind of self study weight multi-angle of view
Class method, i.e., above-mentioned weight parameter wvIt can be learnt, optimization is constantly updated in calculating process.The weight table learnt
It is shown as:
Based on self study weight, similarity matrix can be obtained by solving following function model:
Wherein, the same similarity matrix is shared at each visual angle, very well satisfies cluster between different perspectives in this way
Consistency.Weighting parameter is not defined in model of the invention explicitly, the calculation formula of the parameter alpha in Section 2 are as follows:
Wherein, fiFor the corresponding instruction vector of i-th of sample data, n instruction vector constitutes oriental matrix F.
Using following alternative manner solution formula (14), final optimal adjacent map S is obtained, specifically:
(1) the initial adjacent map S and its corresponding Laplacian Matrix L obtained using step 1S, as follows just
Beginningization oriental matrix F:
(2) fixed wvAnd F, adjacent map S is updated as follows:
Wherein, diFor by dijThe n-dimensional vector of composition, j=1 ..., n.
(3) fixed S updates weight parameter w by formula (13)v, and update instruction matrix F as follows:
It repeats step (2) (3), until the adjacent functional value twice of objective function (formula 14) constantly to successively decrease in solution procedure
Difference less than 0.0001, stop calculate, the S obtained at this time is final optimal adjacent map.
3, final cluster result is calculated
The optimal adjacent map that step 2 obtains is nonnegative matrix, corresponding Laplacian Matrix LS0 characteristic value tuple
Equal to the number of connected component in S.Using being documented in document " Tarjan.Depth-first search and linear
Graph algorithms [C] .Symposium on Switching&Automata Theory.IEEE, in 2008. "
Tarjan's algorithm calculates the connected component in optimal adjacent map S, and connected component is the maximal connected subgraph in adjacent map S, will
Adjacent map is divided into different connected subgraphs, and the sample in each connected subgraph has identical label, marks between different connected subgraphs
Label are different, i.e., are divided into all sample datas in different clustering clusters, thus obtain cluster result.
Claims (1)
1. a kind of multi-angle of view clustering method based on adaptive neighbor point, it is characterised in that steps are as follows:
Step 1: for sample to be clustered, the eigenmatrix of its different perspectives being calculated using visual signature operator, remembersFor the eigenmatrix at v-th of visual angle, v=1 ..., V, V is total number viewpoints, and n is sample data
Number,For feature vector of i-th of sample data under v-th of visual angle, the visual signature operator include SIFT, HOG,
LBP,ColorMatrix;
Then, the sample data under different perspectives is pre-processed, obtains pretreated sample data, it may be assumed that
Wherein,It indicatesThe mean value of middle all elements, σ () indicate standard deviation;
The weight for initializing each visual angle isIt is as follows to construct the function model with order constraint:
Wherein, S ∈ Rn×nIndicate similarity matrix, i.e. adjacent map between sample data, sijThe i-th row j column in representing matrix S
Element, siThe i-th row vector in representing matrix S, Section 2For penalty term, α > 0, LsFor the corresponding drawing of similarity matrix S
This matrix of pula, c indicate the number of clustering cluster, and rank of matrix is sought in rank () expression;
Above-mentioned function model is solved using method of Lagrange multipliers, obtains similarity matrix S, i.e., initial adjacent map;
Step 2: multi-angle of view clustering function model of the building based on self study weight is as follows:
Wherein, the calculation formula of parameter alpha are as follows:
Wherein, fiFor the corresponding instruction vector of i-th of sample data, n instruction vector constitutes oriental matrix F, self study weight wv
Calculation formula are as follows:
Function model (3) are solved using following alternative manner, obtain final optimal adjacent map S, specifically:
Step a: the initial adjacent map S and its corresponding Laplacian Matrix L obtained using step 1S, as follows initially
Change oriental matrix F:
Step b: fixed wvAnd F, adjacent map S is updated as follows:
Wherein, diFor by dijThe n-dimensional vector of composition, j=1 ..., n;
Step c: fixed S updates weight parameter w by formula (6)v, and update instruction matrix F as follows:
Step d: repeating step b-c, until the difference for the functional value that twice adjacent calculation obtains less than 0.0001, stops calculating, this
When obtained S be final optimal adjacent map;
Step 3: using being documented in document " Tarjan.Depth-first search and linear graph
Algorithms [C] .Symposium on Switching&Automata Theory.IEEE, the Tarjan's in 2008. "
Algorithm calculates the connected component in the optimal adjacent map S that step 2 obtains, as final cluster result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811559259.4A CN109784374A (en) | 2018-12-21 | 2018-12-21 | Multi-angle of view clustering method based on adaptive neighbor point |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811559259.4A CN109784374A (en) | 2018-12-21 | 2018-12-21 | Multi-angle of view clustering method based on adaptive neighbor point |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109784374A true CN109784374A (en) | 2019-05-21 |
Family
ID=66497335
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811559259.4A Pending CN109784374A (en) | 2018-12-21 | 2018-12-21 | Multi-angle of view clustering method based on adaptive neighbor point |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109784374A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110232090A (en) * | 2019-05-29 | 2019-09-13 | 北京理工大学 | A kind of integrated detection of multi-angle of view time series abnormal point and method for visualizing |
CN110411724A (en) * | 2019-07-30 | 2019-11-05 | 广东工业大学 | A kind of rotary machinery fault diagnosis method, device, system and readable storage medium storing program for executing |
CN110941978A (en) * | 2019-05-23 | 2020-03-31 | 罗普特科技集团股份有限公司 | Face clustering method and device for unidentified personnel and storage medium |
CN111046745A (en) * | 2019-11-20 | 2020-04-21 | 北京工业大学 | Multi-view clustering method based on soft block diagonal |
CN111104867A (en) * | 2019-11-25 | 2020-05-05 | 北京迈格威科技有限公司 | Recognition model training and vehicle heavy recognition method and device based on component segmentation |
CN111401468A (en) * | 2020-03-26 | 2020-07-10 | 上海海事大学 | Weight self-updating multi-view spectral clustering method based on shared neighbor |
CN111401429A (en) * | 2020-03-12 | 2020-07-10 | 安徽理工大学 | Multi-view image clustering method based on clustering self-adaptive canonical correlation analysis |
CN112307980A (en) * | 2020-10-28 | 2021-02-02 | 北京工业大学 | Image identification method based on incomplete multi-view clustering |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170261264A1 (en) * | 2017-05-25 | 2017-09-14 | Northeastern University | Fault diagnosis device based on common information and special information of running video information for electric-arc furnace and method thereof |
US20170302914A1 (en) * | 2016-04-14 | 2017-10-19 | Gentex Corporation | Vehicle display system providing depth information |
CN107784318A (en) * | 2017-09-12 | 2018-03-09 | 天津大学 | The learning method that a kind of robustness similar diagram for being applied to various visual angles cluster represents |
CN108090485A (en) * | 2017-11-28 | 2018-05-29 | 西安电子科技大学昆山创新研究院 | Display foreground extraction method based on various visual angles fusion |
CN108171279A (en) * | 2018-01-28 | 2018-06-15 | 北京工业大学 | A kind of adaptive product Grassmann manifold Subspace clustering methods of multi-angle video |
CN108764276A (en) * | 2018-04-12 | 2018-11-06 | 西北大学 | A kind of robust weights multi-characters clusterl method automatically |
CN109002854A (en) * | 2018-07-20 | 2018-12-14 | 西安电子科技大学 | Based on hidden expression and adaptive multiple view Subspace clustering method |
-
2018
- 2018-12-21 CN CN201811559259.4A patent/CN109784374A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170302914A1 (en) * | 2016-04-14 | 2017-10-19 | Gentex Corporation | Vehicle display system providing depth information |
US20170261264A1 (en) * | 2017-05-25 | 2017-09-14 | Northeastern University | Fault diagnosis device based on common information and special information of running video information for electric-arc furnace and method thereof |
CN107784318A (en) * | 2017-09-12 | 2018-03-09 | 天津大学 | The learning method that a kind of robustness similar diagram for being applied to various visual angles cluster represents |
CN108090485A (en) * | 2017-11-28 | 2018-05-29 | 西安电子科技大学昆山创新研究院 | Display foreground extraction method based on various visual angles fusion |
CN108171279A (en) * | 2018-01-28 | 2018-06-15 | 北京工业大学 | A kind of adaptive product Grassmann manifold Subspace clustering methods of multi-angle video |
CN108764276A (en) * | 2018-04-12 | 2018-11-06 | 西北大学 | A kind of robust weights multi-characters clusterl method automatically |
CN109002854A (en) * | 2018-07-20 | 2018-12-14 | 西安电子科技大学 | Based on hidden expression and adaptive multiple view Subspace clustering method |
Non-Patent Citations (4)
Title |
---|
FEIPING NIE.ET: "Auto-Weighted Multi-View Learning for Image Clustering and Semi-Supervised Classification", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
FEIPING NIE.ET: "Multi-View Clustering and Semi-Supervised Classification with Adaptive Neighbours", 《PROCEEDINGS OF THE THIRTY-FIRST AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-17)》 * |
刘正等: "基于特征加权和非负矩阵分解的多视角聚类算法", 《电子学报》 * |
李学龙等: "光学影像序列中基于多视角聚类的群组行为分析", 《中国科学:信息科学》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110941978A (en) * | 2019-05-23 | 2020-03-31 | 罗普特科技集团股份有限公司 | Face clustering method and device for unidentified personnel and storage medium |
CN110232090A (en) * | 2019-05-29 | 2019-09-13 | 北京理工大学 | A kind of integrated detection of multi-angle of view time series abnormal point and method for visualizing |
CN110232090B (en) * | 2019-05-29 | 2021-07-09 | 北京理工大学 | Integrated detection and visualization method for abnormal points of multi-view time sequence |
CN110411724B (en) * | 2019-07-30 | 2021-07-06 | 广东工业大学 | Rotary machine fault diagnosis method, device and system and readable storage medium |
CN110411724A (en) * | 2019-07-30 | 2019-11-05 | 广东工业大学 | A kind of rotary machinery fault diagnosis method, device, system and readable storage medium storing program for executing |
CN111046745A (en) * | 2019-11-20 | 2020-04-21 | 北京工业大学 | Multi-view clustering method based on soft block diagonal |
CN111046745B (en) * | 2019-11-20 | 2023-08-04 | 北京工业大学 | Multi-view clustering method based on soft block diagonal angle |
CN111104867A (en) * | 2019-11-25 | 2020-05-05 | 北京迈格威科技有限公司 | Recognition model training and vehicle heavy recognition method and device based on component segmentation |
CN111104867B (en) * | 2019-11-25 | 2023-08-25 | 北京迈格威科技有限公司 | Recognition model training and vehicle re-recognition method and device based on part segmentation |
CN111401429A (en) * | 2020-03-12 | 2020-07-10 | 安徽理工大学 | Multi-view image clustering method based on clustering self-adaptive canonical correlation analysis |
CN111401468A (en) * | 2020-03-26 | 2020-07-10 | 上海海事大学 | Weight self-updating multi-view spectral clustering method based on shared neighbor |
CN111401468B (en) * | 2020-03-26 | 2023-03-24 | 上海海事大学 | Weight self-updating multi-view spectral clustering method based on shared neighbor |
CN112307980A (en) * | 2020-10-28 | 2021-02-02 | 北京工业大学 | Image identification method based on incomplete multi-view clustering |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109784374A (en) | Multi-angle of view clustering method based on adaptive neighbor point | |
Nie et al. | Multi-view clustering and semi-supervised classification with adaptive neighbours | |
Gulhane et al. | Diagnosis of diseases on cotton leaves using principal component analysis classifier | |
Xi et al. | Few-shot learning with class-covariance metric for hyperspectral image classification | |
Senthilnath et al. | Crop stage classification of hyperspectral data using unsupervised techniques | |
Zhu et al. | Learning a discriminant graph-based embedding with feature selection for image categorization | |
CN111178208A (en) | Pedestrian detection method, device and medium based on deep learning | |
CN101777131A (en) | Method and device for identifying human face through double models | |
Yin et al. | Adaptive nonlinear manifolds and their applications to pattern recognition | |
CN107784318A (en) | The learning method that a kind of robustness similar diagram for being applied to various visual angles cluster represents | |
WO2022228348A1 (en) | High-order correlation preserved incomplete multi-view subspace clustering method and system | |
Dehshibi et al. | Cubic norm and kernel-based bi-directional PCA: toward age-aware facial kinship verification | |
Wan et al. | Multi-view clustering by joint spectral embedding and spectral rotation | |
Ji et al. | Semisupervised hyperspectral image classification using spatial-spectral information and landscape features | |
Geng et al. | Local-density subspace distributed clustering for high-dimensional data | |
CN110175631A (en) | A kind of multiple view clustering method based on common Learning Subspaces structure and cluster oriental matrix | |
Su et al. | Graph regularized low-rank tensor representation for feature selection | |
Zhang et al. | Integrating leaf and flower by local discriminant CCA for plant species recognition | |
Zhao et al. | Biologically motivated model for outdoor scene classification | |
CN109766385A (en) | Multiple view Projective Clustering based on self study weight | |
CN116012632A (en) | Feature consistency multi-view clustering algorithm based on local linear embedding | |
Kanzawa | A maximizing model of Bezdek-like spherical fuzzy c-means clustering | |
Bhad et al. | Content based image retrieval a comparative based analysis for feature extraction approach | |
Bhatt et al. | Coarse Clustering and Classification of Images with CNN Features for Participatory Sensing in Agriculture. | |
Wu et al. | Multi-feature sparse constrain model for crop disease recognition. |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190521 |
|
WD01 | Invention patent application deemed withdrawn after publication |