CN118072180B - Hyperspectral image classification method based on keyson hypergraph enhancement - Google Patents
Hyperspectral image classification method based on keyson hypergraph enhancement Download PDFInfo
- Publication number
- CN118072180B CN118072180B CN202410462689.3A CN202410462689A CN118072180B CN 118072180 B CN118072180 B CN 118072180B CN 202410462689 A CN202410462689 A CN 202410462689A CN 118072180 B CN118072180 B CN 118072180B
- Authority
- CN
- China
- Prior art keywords
- hypergraph
- hyperspectral image
- key
- centrality
- sub
- 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
- 238000000034 method Methods 0.000 title claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims description 9
- 230000003595 spectral effect Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 2
- 230000002776 aggregation Effects 0.000 abstract description 5
- 238000004220 aggregation Methods 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a hyperspectral image classification method based on keyson hypergraph enhancement, and belongs to the field of hyperspectral image processing. The invention comprises the following steps: 1. splitting the hyperspectral image into compact and relatively uniform superpixels; 2. each super pixel is regarded as a node, and a hyperspectral image common graph and a global super graph are constructed; 3. constructing a centrality key sub hypergraph through a centrality algorithm; 4. constructing a medium center key sub hypergraph through a medium center algorithm; 5. fusing the degree center sub-hypergraph and the medium center sub-hypergraph to obtain a key sub-hypergraph; 6. and sending the hyperspectral image super-pixel level characteristics, the hyperspectral image global hypergraph and the key sub-hypergraph into the hypergraph convolution under the enhancement of the key sub-hypergraph to train and classify. The invention provides the concept of the hyperspectral image key sub hypergraph for the first time, can improve the interpretability of the network to the hyperspectral image ground feature characteristics, promotes the aggregation and updating of information, and effectively improves the classification performance.
Description
Technical Field
The invention belongs to the field of hyperspectral image processing, and particularly relates to a hyperspectral image classification method based on keyson hypergraph enhancement.
Background
In recent years, with the continuous development of hyperspectral technology and sensor platforms, we can acquire hyperspectral image data with rich spectrum-space information, and realize 'qualitative' to 'quantitative' remote sensing observation. The development is expected to realize more refined hyperspectral image classification, and provides more key and accurate information support for a plurality of fields such as environmental protection, agricultural evaluation, mineral exploration, land management and the like. However, the recognition performance of the model is limited due to the problems of large area, complex ground feature structure, high-dimensional nonlinearity of spectrum information, redundancy and the like of the remote sensing image observation scene. Therefore, how to efficiently represent and learn hyperspectral images under such conditions remains a challenging task.
At present, a deep learning method is mainly adopted in the research of hyperspectral image classification, and feature extraction, selection and classifier are combined in an end-to-end frame so as to automatically extract deep nonlinear feature representation and semantic information in hyperspectral images. Wherein the hypergraph convolutional network can effectively learn and infer non-euclidean data, and model correlation between complex data is widely focused. However, due to the complex scene structure, the constructed hypergraph structure has poor interpretation capability on hyperspectral image ground feature characteristics, and is easily interfered by noise during information aggregation. Therefore, how to improve the interpretability of the hyperspectral image ground feature features by the hypergraph, promote the aggregation and updating of information, and effectively promote the application of the common hypergraph network in hyperspectral image classification is a technical problem to be solved urgently at present.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a hyperspectral image classification method based on the enhancement of a keyson hypergraph, which can improve the interpretation of the feature characteristics of the hyperspectral image and promote the aggregation and updating of information under the enhancement of the keyson hypergraph.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
a hyperspectral image classification method based on keyson hypergraph enhancement comprises the following steps:
(1) Acquiring a hyperspectral image and dividing the hyperspectral image into compact and relatively uniform superpixels;
(2) Each super pixel is regarded as a node, and a hyperspectral image common graph and a global super graph are constructed;
(3) Determining a high centrality key structure of the hyperspectral image through a centrality algorithm, and constructing a high centrality key sub-hypergraph;
(4) Determining a high-medium central key structure of the hyperspectral image through a medium central algorithm, and constructing a medium central key sub-hypergraph;
(5) Fusing the degree center sub-hypergraph and the medium center sub-hypergraph to obtain a hyperspectral image key sub-hypergraph;
(6) And sending the hyperspectral image super-pixel level characteristics, the hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph into the hypergraph convolution under the enhancement of the key sub-hypergraph to train and classify.
As a preferred embodiment of the present invention, in the step (1), the hyperspectral image is segmented into compact and relatively uniform superpixels by a linear iterative clustering algorithm, and a superpixel-level feature representation is acquired; wherein the hyperspectral image is represented asThe superpixel feature is represented asWherein M represents the length of the hyperspectral image, L represents the width, C represents the number of spectral bands,AndThe 1 st block and the N th block super pixel are respectively represented, N is the number of total super pixel blocks, and R is a real number set.
As a preferred embodiment of the present invention, in the step (2), the hyperspectral image is represented by a general chartWherein, the method comprises the steps of, wherein,AndRepresenting the ith block and the jth block super pixels, respectively;
The hyperspectral image global hypergraph is expressed as ; Wherein,Representing a set of super-points,Representing a set of hyperedges that are to be processed,AndRespectively represent the 1 st and N th hypermap nodes of the global hypermap of the hyperspectral image,AndAnd respectively representing the 1 st superside and the N th superside of the global supergraph of the hyperspectral image.
Hyperspectral image plain mapFor subsequent determination of critical nodes and critical structures. The hyperspectral image global hypergraph is used for modeling the higher-order correlation contained in the hyperspectral image.
In the step (3), a high centrality key structure of the hyperspectral image is determined through a centrality algorithm. Centrality is the number of connections one node has with other nodes in the network, and nodes with a high centrality connect more nodes in the network, generally considered critical nodes. Thus, the key nodes can be determined by the centrality of the nodes and the key structure is obtained.
As a preferred embodiment of the present invention, in the step (3), the construction of the isocenter key hypergraph specifically includes:
S3-1: by passing through Calculating the centrality of the nodes, and selecting K nodes with the maximum centrality to form a high centrality key point setWherein, the method comprises the steps of, wherein,Representing nodesIs characterized by a degree of centrality,AndRepresenting the 1 st and the kth highly central keypoints;
S3-2: forming a high centrality superside set by the key points and adjacent nodes of the high centrality key points ,AndRespectively representing the 1 st and the K th height centrality critical supersides;
s3-3: based on the high centrality key point set and the high centrality superside set, a high centrality key sub hypergraph is formed 。
In the step (4), a high-medium central key structure of the hyperspectral image is determined through a medium central algorithm. The betweenness centrality measures the importance of nodes in the network as bridges or information transfer, and key nodes have important reference values in searching hyperspectral network topologies, and are usually measured through the shortest paths.
As a preferred embodiment of the present invention, in the step (4), the median central key hypergraphSpecifically comprising:
S4-1: by passing through Calculating the betweenness centrality of the nodes, and selecting P nodes with the maximum betweenness centrality to form a high betweenness centrality key point setWherein, the method comprises the steps of, wherein,Representing nodesIs characterized by a median center size of (1),Representing slave nodesTo the nodeIs provided with a number of shortest paths of (a),Representing slave nodesTo the nodeThrough a node in the path of (a)Is a function of the number of (3),AndRespectively representing the 1 st and the P-th high-medium central key points;
s4-2: the medium central key points and adjacent nodes thereof form a high medium central superside set ,AndRepresenting the 1 st and P th high-medium central critical supersides;
s4-3: based on the high-bettery central key point set and the high-bettery central superside set, the bettery central key sub-supergraph is formed 。
As a preferred embodiment of the present invention, in the step (5), the hyperspectral image is a key sub-hypergraphRepresented as。
As a preferred embodiment of the present invention, in the step (6), the hyperspectral image is super-pixel-level characterizedGlobal hypergraph of hyperspectral imageAnd hyperspectral image keypoint hypergraphPerforming training and classification by inputting hypergraph convolution under the enhancement of the key sub hypergraph; the hypergraph convolution under the key sub-hypergraph enhancement is expressed as
;
Wherein the method comprises the steps ofAndThe incidence matrixes of the hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph are respectively,Is a matrixIs to be used in the present invention,Is a matrixIs to be used in the present invention,AndThe node degree matrixes of the hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph are respectively,AndRespectively a hyperspectral image global hypergraph and a hyperspectral image key sub-hypergraph,AndThe hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph are respectively the hyperedge weight matrixes of the hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph,For a learnable parameter, Y is the hidden layer representation of the output.
Compared with the prior art, the invention has the beneficial effects that: the invention firstly provides the concept of the hyperspectral image key sub-hypergraph, and constructs the hyperspectral image key sub-hypergraph by providing the concept of the hyperspectral image key sub-structure. Under the enhancement of the hyperspectral image key sub hypergraph, the model can better understand the characteristic attribute of the hyperspectral image feature, and promote the hypergraph convolution network to aggregate and update information, so that the classification performance of the hyperspectral image is improved.
Drawings
FIG. 1 is a flow chart of a method for classifying hyperspectral images based on keysub hypergraph enhancement.
Fig. 2 is a key structural diagram of a hyperspectral image, a network topology, and a hyperspectral image.
FIG. 3 is a graph of the prediction results of different models, with FIG. (a) being a pseudo-color image of a hyperspectral image; (b) A truth label graph, (c) a prediction graph of an SVM model, (d) a prediction graph of a 2DCNN model, (e) a prediction graph of an M3DCNN model, (f) a prediction graph of a HybridSN model, (g) a prediction graph of a GCN model, (h) a prediction graph of an S2HCN model, and (i) a prediction graph of a EHGCN (invention) model.
Detailed Description
For a better description of the objects, technical solutions and advantages of the present invention, the present invention will be further described with reference to the following specific examples.
Example 1
A hyperspectral image classification method based on keyson hypergraph enhancement specifically comprises the following steps:
step1: for hyperspectral images to be acquired And (C) represents the length of the hyperspectral image, the width of the hyperspectral image, and the number of spectral bands. Segmentation of images into compact and relatively uniform superpixels by a linear iterative clustering algorithm, the superpixel features being represented asWherein M represents the length of the hyperspectral image, L represents the width, C represents the number of spectral bands,AndThe 1 st block and the N th block super pixel are respectively represented, N is the number of total super pixel blocks, and R is a real number set.
Step2: constructing hyperspectral image normal map on super-pixel level feature spaceSpecifically expressed as. And construct a hyperspectral image global hypergraphWherein, the method comprises the steps of, wherein,Representing a set of super-points,Representing a set of hyperedges that are to be processed,AndRespectively represent the 1 st and N th hypermap nodes of the global hypermap of the hyperspectral image,AndAnd respectively representing the 1 st superside and the N th superside of the global supergraph of the hyperspectral image. The hyperspectral image common graph is used for determining subsequent key nodes and key structures. The hyperspectral image global hypergraph is used for modeling the higher-order correlation contained in the hyperspectral image.
Step3: and determining a high centrality key structure of the hyperspectral image through a centrality algorithm. Centrality is the number of connections one node has with other nodes in the network. Nodes with a high degree of centrality connect more nodes in the network and are generally considered critical nodes. Thus, the key nodes can be determined by the centrality of the nodes and the key structure is obtained. Specifically, for a nodeThe centrality is defined as: from the calculated centrality K nodes with the greatest centrality are selected to form a high centrality key point setForming a high-centrality superside set by the key points and adjacent nodes thereof. The key point set with high centrality and the hyperedge set with high centrality based on the centrality form a centrality key sub hypergraph。
Step4: and determining a high-medium central key structure of the hyperspectral image through a medium central algorithm. The betweenness centrality measures the importance of nodes in the network as bridges or information transfer, and key nodes have important reference values in searching hyperspectral network topologies, and are usually measured through the shortest paths. Specifically, for a nodeThe median centrality is defined as: wherein, the method comprises the steps of, wherein, Representing slave nodesTo the nodeIs provided with a number of shortest paths of (a),Representing slave nodesTo the nodeThrough a node in the path of (a)Is a number of (3). Similarly, the median centrality obtained from the calculationP nodes with the greatest betweenness are selected to form a key point set based on the high betweenness centralityForming high-medium central superb set by high-medium central key points and adjacent nodes thereofBased on the combination of the high-medium central key point set and the high-medium central superb set, the medium central key sub hypergraph is formed。
Step5: degree-centered sub-hypergraphAnd medium number center sub hypergraphFusion to obtain hyperspectral image key sub hypergraphExpressed as。
Step6: super-pixel level characterization of hyperspectral imagesGlobal hypergraph of hyperspectral imageAnd hyperspectral image keypoint hypergraphHypergraph convolution under input key sub-hypergraph enhancement. Specifically, the hypergraph convolution under key sub-hypergraph enhancement is defined as: Wherein AndThe incidence matrixes of the hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph are respectively,Is a matrixIs to be used in the present invention,Is a matrixIs to be used in the present invention,AndThe node degree matrixes of the hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph are respectively,AndRespectively a hyperspectral image global hypergraph and a hyperspectral image key sub-hypergraph,AndThe hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph are respectively the hyperedge weight matrixes of the hyperspectral image global hypergraph and the hyperspectral image key sub-hypergraph,For a learnable parameter, Y is the hidden layer representation of the output.
Effect example
This example is for the purpose of illustrating that the method for classifying hyperspectral images based on keypoint hypergraph enhancement described in example 1 is feasible.
1. Experimental data:
Pavia University (PaviaU) data set A portion of the hyperspectral data of the image made by an on-board reflectance optical spectrum imager (ROSIS-03) in Germany to Parviea, italy in 2003. The spatial resolution of the resulting image was 1.3m, with a 610 x 340 pixel size and 103 spectral bands. Furthermore, the image contains 9 different ground object categories.
2. Experimental setup
All experiments were carried out using PyTorch framework, optimized using Adam optimizer, weight decay set to 0.0005, activation function using LeakyReLU, total training period number 600, initial learning rate set to 0.001; and the Overall Accuracy (OA), the Average Accuracy (AA) and Kappa coefficient were used as evaluation indexes.
In order to fully verify that the hyperspectral image classification method based on the enhancement of the keyson hypergraph has better classification performance, abbreviated as EHGCN (the invention), the effect example is provided with some advanced models as contrast verification ,Support Vector Machine(SVM)、Two-dimensional Convolutional Neural Net-work (2DCNN)、multi-scale 3D deep convolutional neural network (M3DCNN)、Hybrid Spectral Network (HybridSN)、Graph Convolutional Network (GCN)、spectral-spatial hypergraph convolution neural network (S2HCN).. It is worth noting that the hypergraph is also used for hyperspectral image classification by S2HCN, but the interpretation capability of the hyperspectral image feature is poor, the hyperspectral image feature is easy to be influenced by noise, and the hyperspectral image classification method has important reference value in verifying the effectiveness of the method.
3. Experimental results
Through the above steps, the PaviaU dataset was experimentally verified, and the experimental results are shown in table 1.
Table 1 PaviaU dataset numerical results (%)
Method of | SVM | 2DCNN | M3DCNN | HybridSN | GCN | S2HCN | EHGCN (inventive) |
OA | 69.37±3.93 | 71.82±2.35 | 80.36±3.39 | 76.84±1.43 | 86.44±1.82 | 89.2±1.81 | 92.02±0.13 |
AA | 58.69±1.96 | 71.99±3.63 | 77.48±2.87 | 62.9±3.08 | 78.48±3.54 | 83.27±2.88 | 87.22±0.17 |
Kappa | 58.13±4.89 | 63.68±2.65 | 74.2±4.22 | 69.54±1.77 | 82.14±2.29 | 85.65±2.39 | 89.44±0.16 |
As can be seen from table 1 and fig. 3, the classification effect of the present invention is higher than that of the comparative model in all three indexes. In particular, the classification effect of the invention is higher than that of S2HCN, which shows that the invention creatively enhances the interpretation of hypergraph to hyperspectral ground object characteristics by constructing a key sub-hypergraph structure, and promotes the aggregation and updating of information. At the same time, it can also be seen from fig. 3 that the proposed method has better classification performance, especially on the area a with critical structure.
Finally, it should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the scope of the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted equally without departing from the spirit and scope of the technical solution of the present invention.
Claims (7)
1. The hyperspectral image classification method based on keyson hypergraph enhancement is characterized by comprising the following steps of:
(1) Acquiring a hyperspectral image and dividing the hyperspectral image into compact and relatively uniform superpixels;
(2) Each super pixel is regarded as a node, and a hyperspectral image common graph and a global super graph are constructed;
(3) Determining a high centrality key structure of the hyperspectral image through a centrality algorithm, and constructing a high centrality key sub-hypergraph;
(4) Determining a high-medium central key structure of the hyperspectral image through a medium central algorithm, and constructing a medium central key sub-hypergraph;
(5) Fusing the degree center sub-hypergraph and the medium center sub-hypergraph to obtain a hyperspectral image key sub-hypergraph;
(6) And training and classifying hyperspectral image superpixel level characteristics, hyperspectral image global hypergraph and hyperspectral image key sub-hypergraph input hypergraph convolution under key sub-hypergraph enhancement.
2. The hyperspectral image classification method based on key sub-hypergraph enhancement as claimed in claim 1, wherein in the step (1), the hyperspectral image is segmented into compact and relatively uniform superpixels by a linear iterative clustering algorithm, and a superpixel-level feature representation is obtained; wherein the hyperspectral image is represented asThe superpixel feature is represented asWherein M represents the length of the hyperspectral image, L represents the width, C represents the number of spectral bands,And/>The 1 st block and the N th block super pixel are respectively represented, N is the number of total super pixel blocks, and R is a real number set.
3. The method for classifying hyperspectral images based on keypoint hypergraph enhancement as claimed in claim 1, wherein in the step (2), the hyperspectral image normal graph is expressed as
Wherein/>And/>Representing the ith block and the jth block super pixels, respectively;
The hyperspectral image global hypergraph is expressed as ; Wherein,Representing a set of superpoints,/>Representing a hyperedge set,/>And/>1 St and nth hypergraph nodes respectively representing global hypergraphs of hyperspectral images,/>And/>And respectively representing the 1 st superside and the N th superside of the global supergraph of the hyperspectral image.
4. The hyperspectral image classification method based on keyhypergraph enhancement as claimed in claim 1, wherein in the step (3), the construction of the isocentric keygraph specifically comprises:
S3-1: by passing through Calculating the centrality of the nodes, and selecting K nodes with the maximum centrality to form a high centrality key point set/>Wherein/>Representing nodes/>Centrality of (v)/(v)And/>Representing the 1 st and the kth highly central keypoints;
S3-2: forming a high centrality superside set by the key points and adjacent nodes of the high centrality key points ,/>And/>Respectively representing the 1 st and the K th height centrality critical supersides;
s3-3: based on the high centrality key point set and the high centrality superside set, a high centrality key sub hypergraph is formed 。
5. The method of hyperspectral image classification based on keyhypergraph enhancement as claimed in claim 4 wherein in step (4), the central keyhypergraph is a medium-center keyhypergraphSpecifically comprising:
S4-1: by passing through Calculating the betweenness centrality of the nodes, and selecting P nodes with the maximum betweenness centrality to form a high betweenness centrality key point set/>Wherein/>Representing nodes/>The median center size of/(Representing slave node/>To node/>Shortest path number,/>Representing slave node/>To node/>In the path through node/>Number of,/>And/>Respectively representing the 1 st and the P-th high-medium central key points;
s4-2: the medium central key points and adjacent nodes thereof form a high medium central superside set ,/>And/>Representing the 1 st and P th high-medium central critical supersides;
s4-3: based on the high-bettery central key point set and the high-bettery central superside set, the bettery central key sub-supergraph is formed 。
6. The method for classifying hyperspectral images based on keysub-hypergraph enhancement as claimed in claim 5, wherein in the step (5), hyperspectral images are keysub-hypergraphsExpressed as/>。
7. The method of classifying hyperspectral images based on key sub-hypergraph enhancement as recited in claim 6, wherein in said step (6), hyperspectral images are super-pixel-level characterizedHyperspectral image Global hypergraph/>And hyperspectral image key hypergraph/>Performing training and classification by inputting hypergraph convolution under the enhancement of the key sub hypergraph; the hypergraph convolution under the key sub-hypergraph enhancement is expressed as
;
Wherein the method comprises the steps ofAnd/>Incidence matrices of hyperspectral image global hypergraph and hyperspectral image key sub hypergraph respectively,/>For matrix/>Transpose of/>For matrix/>Transpose of/>And/>Node degree matrix of hyperspectral image global hypergraph and hyperspectral image key sub hypergraph respectively,/>And/>Hyper-edge matrix of hyperspectral image global hypergraph and hyperspectral image key sub-hypergraph respectively,/>And/>Hyper-edge weight matrix of hyperspectral image global hypergraph and hyperspectral image key sub-hypergraph respectively,/>For a learnable parameter, Y is the hidden layer representation of the output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410462689.3A CN118072180B (en) | 2024-04-17 | 2024-04-17 | Hyperspectral image classification method based on keyson hypergraph enhancement |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410462689.3A CN118072180B (en) | 2024-04-17 | 2024-04-17 | Hyperspectral image classification method based on keyson hypergraph enhancement |
Publications (2)
Publication Number | Publication Date |
---|---|
CN118072180A CN118072180A (en) | 2024-05-24 |
CN118072180B true CN118072180B (en) | 2024-06-18 |
Family
ID=91107930
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410462689.3A Active CN118072180B (en) | 2024-04-17 | 2024-04-17 | Hyperspectral image classification method based on keyson hypergraph enhancement |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN118072180B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116912677A (en) * | 2023-06-12 | 2023-10-20 | 湖州师范学院 | Hyperspectral image classification method based on mixed CNN-GCN network under limited training sample |
CN117315381A (en) * | 2023-11-30 | 2023-12-29 | 昆明理工大学 | Hyperspectral image classification method based on second-order biased random walk |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105787516B (en) * | 2016-03-09 | 2019-07-16 | 南京信息工程大学 | A kind of hyperspectral image classification method based on empty spectrum locality low-rank hypergraph study |
CN110717354B (en) * | 2018-07-11 | 2023-05-12 | 哈尔滨工业大学 | Super-pixel classification method based on semi-supervised K-SVD and multi-scale sparse representation |
CN115331063A (en) * | 2022-09-02 | 2022-11-11 | 安徽大学 | Hyperspectral image classification method and system based on dynamic hypergraph convolution network |
CN115471750A (en) * | 2022-09-15 | 2022-12-13 | 西安电子科技大学 | Hyperspectral anomaly detection method for discriminating forest based on superpixel guidance |
CN116778339A (en) * | 2023-07-19 | 2023-09-19 | 青岛大学 | Method and system for selecting hyperspectral wave bands by aid of local view auxiliary discrimination |
CN117456263A (en) * | 2023-11-14 | 2024-01-26 | 湖州师范学院 | Hyperspectral image classification method based on double-branch multi-scale converter network |
-
2024
- 2024-04-17 CN CN202410462689.3A patent/CN118072180B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116912677A (en) * | 2023-06-12 | 2023-10-20 | 湖州师范学院 | Hyperspectral image classification method based on mixed CNN-GCN network under limited training sample |
CN117315381A (en) * | 2023-11-30 | 2023-12-29 | 昆明理工大学 | Hyperspectral image classification method based on second-order biased random walk |
Also Published As
Publication number | Publication date |
---|---|
CN118072180A (en) | 2024-05-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | A graph-cnn for 3d point cloud classification | |
Li et al. | Classification of hyperspectral image based on deep belief networks | |
CN112966114B (en) | Literature classification method and device based on symmetrical graph convolutional neural network | |
CN110827921B (en) | Single cell clustering method and device, electronic equipment and storage medium | |
CN111931505A (en) | Cross-language entity alignment method based on subgraph embedding | |
CN107590515A (en) | The hyperspectral image classification method of self-encoding encoder based on entropy rate super-pixel segmentation | |
CN110473592A (en) | The multi-angle of view mankind for having supervision based on figure convolutional network cooperate with lethal gene prediction technique | |
CN111047078B (en) | Traffic characteristic prediction method, system and storage medium | |
CN113157957A (en) | Attribute graph document clustering method based on graph convolution neural network | |
CN104156943B (en) | Multi objective fuzzy cluster image change detection method based on non-dominant neighborhood immune algorithm | |
CN110993037A (en) | Protein activity prediction device based on multi-view classification model | |
CN104318271A (en) | Image classification method based on adaptability coding and geometrical smooth convergence | |
Zhang et al. | A graph-voxel joint convolution neural network for ALS point cloud segmentation | |
CN117853596A (en) | Unmanned aerial vehicle remote sensing mapping method and system | |
Zhang et al. | Semantic segmentation of spectral LiDAR point clouds based on neural architecture search | |
CN117407781A (en) | Equipment fault diagnosis method and device based on federal learning | |
CN117349494A (en) | Graph classification method, system, medium and equipment for space graph convolution neural network | |
CN118072180B (en) | Hyperspectral image classification method based on keyson hypergraph enhancement | |
CN115953902A (en) | Traffic flow prediction method based on multi-view space-time diagram convolution network | |
CN116758349A (en) | Hyperspectral image classification method based on multi-scale super-pixel node aggregate graph convolution residual network | |
CN116432053A (en) | Multi-mode data representation method based on modal interaction deep hypergraph neural network | |
CN114936327B (en) | Element recognition model acquisition method and device, computer equipment and storage medium | |
CN114265954B (en) | Graph representation learning method based on position and structure information | |
CN115346055A (en) | Multi-kernel width map based neural network feature extraction and classification method | |
Ghosh et al. | A deep learning‐based SAR image change detection using spatial intuitionistic fuzzy C‐means clustering |
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 |