CN110414616B - Remote sensing image dictionary learning and classifying method utilizing spatial relationship - Google Patents
Remote sensing image dictionary learning and classifying method utilizing spatial relationship Download PDFInfo
- Publication number
- CN110414616B CN110414616B CN201910711662.2A CN201910711662A CN110414616B CN 110414616 B CN110414616 B CN 110414616B CN 201910711662 A CN201910711662 A CN 201910711662A CN 110414616 B CN110414616 B CN 110414616B
- Authority
- CN
- China
- Prior art keywords
- dictionary
- pixel
- training
- spatial
- sparse
- 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 43
- 238000012549 training Methods 0.000 claims abstract description 54
- 238000002372 labelling Methods 0.000 claims abstract description 13
- 238000012706 support-vector machine Methods 0.000 claims abstract description 10
- 238000005457 optimization Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/28—Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (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)
- Image Processing (AREA)
Abstract
The invention discloses a remote sensing image dictionary learning classification method utilizing a spatial relationship, which comprises the following steps: firstly, regarding a p-neighborhood set of each pixel as a training unit, and extracting and expressing spatial relation information of each training unit; and then, local neighbor spatial relation information is introduced to construct a dictionary learning model based on local neighbor region joint representation, and the characteristics extracted from each training unit are utilized to train by an online dictionary updating mechanism to obtain an optimal dictionary set. And finally, carrying out sparse coding on the p-neighborhood set associated with each pixel based on the optimal dictionary obtained by training, and training a linear support vector machine model based on the obtained sparse discriminant coefficient characteristics and the labeling information to classify unlabeled pixels. The method carries out joint sparse expression on the pixels in the local neighboring area of the remote sensing image, so that the constructed dictionary learning model can fully perceive potential spatial relationship information in the image, and the aim of accurately identifying the ground object target is fulfilled.
Description
Technical Field
The invention relates to a remote sensing image dictionary learning classification method utilizing a spatial relationship, which can be used for a ground surface refined interpretation task under a complex hyperspectral image scene and belongs to the technical field of remote sensing big data intelligent interpretation.
Background
At present, a novel machine learning method and a novel model become hot research directions in remote sensing image data processing, sparse expression is used as a novel machine learning technology, and the ground object information of the remote sensing image can be effectively extracted by utilizing the high redundancy of massive high-dimensional data and the sparsity of the signals of interest. The existing remote sensing image classification method based on sparse representation generally directly constructs a dictionary set from a labeled training sample, and determines the type of the non-labeled pel ground object on the image through a class minimum residual criterion and dictionary corresponding labeling information. The sparse representation model constructed in this way is severely dependent on the selection quality of the dictionary set, i.e. when the number of selected dictionaries is large, a lower operation efficiency is usually incurred due to the large dictionary size. When the number of the selected dictionaries is small, partial pels cannot be effectively sparsely expressed due to the small size of the dictionaries, and thus cannot be effectively and correctly classified.
Disclosure of Invention
The invention aims to: aiming at the problems and the defects existing in the prior art, the invention provides a remote sensing image dictionary learning classification method utilizing a spatial relationship, and in view of the spatial homogeneity distribution characteristic of a remote sensing image, namely that spatially adjacent pixels are generally formed by similar materials, the probability of belonging to the same ground object category in the remote sensing image classification is higher. By means of dictionary learning thought, a complete dictionary set is learned from the whole image by utilizing an online two-step dictionary updating strategy, so that full expression of all pixels is realized, and excellent classification performance can be obtained under the condition of limited training samples.
The technical scheme is as follows: a method for learning and classifying the dictionary of remote sensing image by using the space relation includes such steps as determining the size of a proper space window according to the space resolution of remote sensing image, regarding the p-neighborhood set of each pixel as a training unit, and extracting and expressing the space relation information of each training unit. And then, embedding the image local neighbor spatial relationship information into a dictionary learning model to construct a dictionary learning model based on local neighbor region joint representation, and training by utilizing the features extracted from each training unit through an online dictionary updating mechanism to obtain an optimal dictionary. And finally, performing sparse coding on the p-neighborhood set associated with each pixel of the remote sensing image by using the trained optimal dictionary to obtain sparse discrimination coefficient characteristics, selecting part of samples by using the labeling layer to train a linear support vector machine model, and classifying unlabeled samples. Specifically, the training process and the classifying process can be divided.
The training process comprises the following steps:
step 100, traversing the whole remote sensing image pixel by pixel, and executing a spatial neighbor extraction operation on each pixel to obtain a spatial neighbor set of each pixel;
step 101, extracting local spatial relation features of each p-neighborhood set (namely a pixel space neighborhood set) to obtain local neighbor features corresponding to each pixel;
102, constructing a dictionary learning model based on spatial relation by using the total reconstruction error minimum criterion of training data (pixels available on an image);
step 103, training an optimal (complete) dictionary set by an online dictionary updating mechanism by utilizing stream data;
the classification process specifically comprises the following steps:
step 200, acquiring a joint representation of each pixel based on a local neighbor region by utilizing the learned optimal dictionary set, and acquiring sparse discrimination coefficient characteristics of each p-neighbor set;
and step 201, selecting part of labeling samples (pixels with labeling information), training a linear support vector machine by using sparse discrimination coefficient characteristics, and classifying remote sensing images by using a learning obtained model.
The specific method for traversing the whole remote sensing image pixel by pixel and executing the spatial neighbor extraction operation on each pixel in the step 100 is as follows: according to the characteristics of image space resolution and the like, determining a local neighbor window with proper size, and aiming at any pixel y of an image i Obtaining its corresponding p-neighborhood set Y G，i ＝[y i，0 ，y i，1 ，...，y i，p ]Wherein y is i，0 Represents the center pel of the window (i.e., the current pel y i )，y i，j Representing the jth spatial neighboring pixel associated with the center pixel.
Wherein,,representing the center pixel y i With its jth spatial neighboring pixel y i，j The self-adaptive weight between sigma represents Gaussian kernel frequency bandwidth, and the normalization processing is carried out on the weight parameter of the center pixel, namely +.> Then, local neighbor adaptive characteristics of each p-neighborhood set are obtained according to the local adaptive weights>The proper window size is determined according to the spatial resolution of the input image, and the window of the high spatial resolution image (such as 2 m) is properly selected to be larger (such as 10 x 10 windows); the low spatial resolution image (e.g., 20 m) is selected to be smaller (e.g., 3*3 window). The selection principle ensures that the selected adjacent pixels are spatially related to each other, and the spatially related pixels are input into a dictionary learning model for joint expression so as to train to obtain a better dictionary.
In step 102, the specific method for constructing the dictionary learning model based on the spatial relationship by using the minimum reconstruction error criterion of the training data is as follows: assume a set of training dataAnd constructing a dictionary learning model based on the joint representation of the local neighbor areas according to the minimum reconstruction error criterion of the training data by utilizing the spatial relation information contained in each p-neighborhood set of the training data, wherein the specific formula is as follows:
wherein,,the dictionary set based on the local neighbor features is represented, and the initial state of the dictionary set is composed of the same number of feature features selected from each class. Alpha i The representation is based on the dictionary->Sparse coding coefficient of S (α) i ) The representation sparse regularization constraint term may be l 0 -norm, l 1 -norms and l 2 -norms, λ denotes the reconstruction error term and S (α i ) Balance parameters between.
The specific practice of training the optimal (complete) dictionary set by using the stream data through the online dictionary updating mechanism in step 103 is as follows: dividing training data into stream data forms, using online two-stepAnd (3) carrying out optimization strategy, namely carrying out joint sparse coding and dictionary updating based on the local neighbor region in turn, and learning from stream data to obtain an optimal dictionary set. Wherein, the sparse coding stage utilizes the dictionary set obtained by the last iteration trainingCalculating the current coding coefficient alpha t The specific formula is expressed as follows:
the above formula can be expressed by the basis of l 1 The sparse optimization problem solving of the norm constraint, the specific optimization method comprises a minimum angle regression method and an alternate direction multiplier method, and the right part of the equal sign of the optimization formula, namely the unknown variable coding coefficient alpha is solved t Dictionary based on current iterationSolving->The coding coefficient alpha of (2) t . Wherein alpha is t Representing the coding coefficients to be solved ∈ ->Representing the actually calculated coding coefficients. When obtaining coding coefficient alpha t After that, the current iteration dictionary set +.>The method can be obtained by approximate solution of the following formula:
wherein the statistical matrixAnd-> Existing information of previous iterations of the dictionary learning model based on the joint representation of the local neighbor regions is stored, tr representing the trace of the matrix. Dictionary updates the corresponding optimization problem can be solved by a block coordinate descent algorithm. Training the obtained optimal dictionary ++when the algorithm stopping criterion is satisfied>I.e. dictionary corresponding to the current iteration->
The classification process specifically comprises the following steps:
the specific method for obtaining the joint representation of each pixel based on the local neighbor region by using the learned optimal dictionary set in the step 201 is as follows: optimal dictionary based on training in training processThe characteristic associated with the spatial p-neighborhood set corresponding to any pixel on the remote sensing image can be obtained based on the local neighbor region joint coding model>The corresponding coding coefficient alpha (namely sparse discriminant coding feature) is specifically calculated according to the following formula:
step 202, training a linear Support Vector Machine (SVM) by using sparse discriminant coefficient features, and classifying remote sensing images by using a learning model comprises the following specific steps: and randomly selecting part labeling samples of each type, training a linear support vector machine model based on the sparse discriminant coefficient characteristics, and classifying the sparse discriminant coefficient characteristics associated with each pixel of the image according to the learning obtaining model.
The effective effects are as follows: compared with the prior art, the remote sensing image dictionary learning and classifying method utilizing the spatial relationship can fully utilize the spatial relationship information of the local neighbor region to train and obtain the complete dictionary set in the implementation process, and meanwhile, an online dictionary updating mechanism based on stream data is adopted to alternately perform sparse coding and dictionary updating based on the local neighbor relationship, so that an optimal dictionary can be trained from the whole image, and the training mode is more flexible.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a flow chart of a local spatial relationship information extraction process of the present invention;
FIG. 3 is a flow chart of the dictionary training process based on local spatial relationship information of the present invention.
Detailed Description
The present invention is further illustrated below in conjunction with specific embodiments, it being understood that these embodiments are meant to be illustrative of the invention only and not limiting the scope of the invention, and that modifications of the invention, which are equivalent to those skilled in the art to which the invention pertains, will fall within the scope of the invention as defined in the claims appended hereto.
The remote sensing image dictionary learning and classifying method utilizing the spatial relationship aims to train a complete dictionary by utilizing the spatial relationship information of the local neighbors according to the homogeneity and heterogeneity of the local spatial neighbors of the remote sensing image, sparse coding is carried out on each pixel neighborhood of the remote sensing image based on the optimal dictionary to obtain sparse discrimination coefficient characteristics, and the obtained discrimination characteristics are utilized to train a linear classifier to realize the identification of the central pixels of each neighborhood set.
A flow chart of a remote sensing image dictionary learning classification method using spatial relationship is shown in figure 1. The device needs to input a remote sensing image and a corresponding labeling layer, wherein the dimension of the remote sensing image is MxN x B (wherein MxN is the space dimension of the remote sensing image, B is the wave band number of the remote sensing image), and the dimension of the labeling layer is MxN. Label (C)The injection sample is mainly used for the initialization of the training dictionary and the selection of the training sample in the classification stage. The specific flow is as follows: firstly, selecting a rule window to traverse the whole image of a remote sensing image, taking each pixel point as a center to obtain each pixel p-neighborhood set, and extracting spatial relationship information of each p-neighborhood set to obtain a characteristic image capable of representing local spatial relationship informationAnd the dictionary learning model based on local region joint expression is realized by fusing local neighbor spatial relationship information in a sparse coding stage by means of dictionary learning algorithm ideas. The model obtained by training the online random optimization idea is assumed to have a larger number of training data, even an infinite number of training data, each sample is input into the model to be trained one by one, and the initial dictionary is constructed by randomly selecting the same number of samples according to each type. Taking the dictionary obtained by the last iterative calculation as an optimal dictionary +.>And according to the optimum dictionary->And carrying out joint sparse expression based on the local neighbor region on the p-neighborhood set corresponding to each pixel of the image to obtain sparse discrimination coefficient characteristics capable of reflecting the local spatial neighbor relation. And finally, randomly selecting a certain number of samples according to the labeling layer, inputting sparse discrimination features of the corresponding samples to train a linear support vector machine model, classifying unlabeled samples of the whole image by using the trained model, and drawing and outputting.
The basic idea of the process of extracting the local spatial relationship information is that the process is as shown in fig. 2, the process is that the whole image traversal is carried out according to the input whole remote sensing image, the corresponding local adjacent areas are divided for each pixel, the local spatial information extraction and expression are carried out for each local adjacent area, and the characteristics capable of reflecting the spatial relationship information of each local adjacent are extracted. The specific flow is as follows: first, a suitable window size (sxs) is determined based on information such as the spatial resolution of the remote sensing image) The p-neighborhood set of the current picture element can be obtained. And then, selecting a proper distance measure based on a Gaussian kernel function to determine the similarity weight between the central pixel and the p-neighborhood of the central pixel, and normalizing the weight of the current pixel and each p-neighborhood of the central pixel. And calculating the characteristics capable of reflecting the current p-neighborhood set spatial relationship information according to the normalized similarity weightAnd finally, outputting a characteristic image which can reflect the spatial relation information of each pixel on the remote sensing image according to the full-image traversing result.
A flow chart of a dictionary training process based on local spatial relationship information is shown in fig. 3, and the basic idea is that the dictionary learning model based on local region joint expression is trained by utilizing the spatial relationship information characteristics extracted by the p-neighborhood set corresponding to each pixel of a remote sensing image by utilizing the stream data through an online two-step optimization mechanism, and the operations of local region joint coding and dictionary updating are sequentially carried out until a stopping criterion is met. The specific flow is as follows: firstly, extracting p-neighborhood set from each pixel of remote sensing image, and calculating to obtain a series of information features reflecting spatial relationThe obtained characteristics->Randomly initialized to meet independent co-distribution assumptions. Then, according to the image labeling layer, selecting the same number of labeling pixels in each class, constructing an initial training dictionary in a stacking mode, and adding +.>And updating the current dictionary model according to a pixel-by-pixel input mode. Current sample of input->Dictionary based on preamble training>Coding to obtain coding coefficient alpha t . From the obtained alpha t Updating the statistical matrix A t ，B t And further calculate the current dictionary ++>Then, inputting the next sample, continuing training according to the operation until the stopping criterion is met, and taking the dictionary obtained in the last iteration as the optimal dictionary +.>And outputting.
Claims (3)
1. A remote sensing image dictionary learning and classifying method utilizing spatial relation is characterized by comprising the following steps:
step 100, traversing the whole remote sensing image pixel by pixel, and executing a spatial neighbor extraction operation on each pixel to obtain a spatial neighbor set of each pixel;
step 101, extracting local spatial relation features of each p-neighborhood set to obtain local neighbor features corresponding to each pixel;
102, constructing a dictionary learning model based on a spatial relationship by utilizing a minimum criterion of a reconstruction error of training data;
step 103, training an optimal dictionary set by using stream data through an online dictionary updating mechanism;
step 200, acquiring a joint representation of each pixel based on a local neighbor region by utilizing the learned optimal dictionary set, and acquiring sparse discrimination coefficient characteristics of each p-neighbor set;
step 201, selecting a part of labeling samples, training a linear support vector machine by using sparse discriminant coefficient characteristics, and classifying remote sensing images by using a learning obtaining model;
the specific method for traversing the whole remote sensing image pixel by pixel and executing the spatial neighbor extraction operation on each pixel in the step 100 is as follows: determining a local neighbor window according to the characteristics of the spatial resolution of the image, and aiming at any image of the imageElement y i Obtaining its corresponding p-neighborhood set Y G,i ＝[y i,0 ,y i,1 ,...,y i,p ]Wherein y is i,0 Representing the central picture element of the window, i.e. the current picture element y i ，y i,j Representing a j-th spatial neighboring pixel associated with the center pixel;
step 101 is to extract local spatial relation features of each p-neighborhood set, and the specific method for obtaining the local neighbor features corresponding to each pixel is as follows: first, the similarity between the center pixel and each neighbor pixel is determined on each p-neighborhood set using a Gaussian kernel function, i.e
Wherein,,representing the center pixel y i With its jth spatial neighboring pixel y i,j The self-adaptive weight between sigma represents Gaussian kernel frequency bandwidth, and the normalization processing is carried out on the weight parameter of the center pixel, namely +.> Then, local neighbor adaptive characteristics of each p-neighborhood set are obtained according to the local adaptive weights>
Step 102, a specific method for constructing a dictionary learning model based on spatial relationship by using the minimum reconstruction error criterion of training data is as follows: assume a set of training dataUtilizing the space implied by each p-neighborhood set of training dataAnd the relation information is used for constructing a dictionary learning model based on the joint representation of the local neighbor areas according to the minimum reconstruction error criterion of the training data, and the specific formula is as follows:
wherein,,representing a dictionary set based on local neighbor features, wherein the initial state of the dictionary set is composed of the same number of feature features selected from each class; alpha i The representation is based on the dictionary->Sparse coding coefficient of S (α) i ) Representing sparse regularization constraint terms;
the specific practice of training the optimal dictionary set by using the stream data through the online dictionary updating mechanism in step 103 is as follows: dividing training data into stream data forms, and sequentially and alternately carrying out joint sparse coding and dictionary updating based on local neighbor areas by utilizing an online two-step optimization strategy to learn from the stream data to obtain an optimal dictionary set; wherein, the sparse coding stage utilizes the dictionary set obtained by the last iteration trainingCalculating the current coding coefficient alpha t The specific formula is expressed as follows:
the function is based on l 1 -a sparse optimization problem solution of norm constraint, the specific optimization method comprising a minimum angle regression method and an alternating direction multiplier method; when obtaining coding coefficient alpha t Thereafter, the current iteration dictionary setThe method is obtained by approximate solution of the following formula:
wherein the statistical matrixAnd-> Storing existing information of previous iterations of a dictionary learning model based on the joint representation of the local neighbor regions; the dictionary updating corresponding optimization problem is solved through a block coordinate descent algorithm; training the obtained optimal dictionary ++when the algorithm stopping criterion is satisfied>I.e. dictionary corresponding to the current iteration->
2. The method of claim 1, wherein the step 200 of obtaining the local neighbor region-based joint representation of each pixel by using the learned optimal dictionary set comprises the following steps: optimal dictionary based on training in training processObtaining the characteristic associated with the spatial p-neighborhood set corresponding to any pixel on the remote sensing image based on the local neighbor region joint coding model by the following method>The corresponding coding coefficient alpha, namely the sparse discriminant coding feature, is specifically calculated according to the following formula:
3. the method for learning and classifying a remote sensing image dictionary by using spatial relationships according to claim 1, wherein the step 201 is implemented by training a linear support vector machine by using sparse discriminant coefficient features and classifying the remote sensing image by using a learning model, and comprises the following steps: and randomly selecting part labeling samples of each type, training a linear support vector machine model based on the sparse discriminant coefficient characteristics, and classifying the sparse discriminant coefficient characteristics associated with each pixel of the image according to the learning obtaining model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910711662.2A CN110414616B (en) | 2019-08-02 | 2019-08-02 | Remote sensing image dictionary learning and classifying method utilizing spatial relationship |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910711662.2A CN110414616B (en) | 2019-08-02 | 2019-08-02 | Remote sensing image dictionary learning and classifying method utilizing spatial relationship |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110414616A CN110414616A (en) | 2019-11-05 |
| CN110414616B true CN110414616B (en) | 2023-07-25 |
Family
ID=68365447
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910711662.2A Active CN110414616B (en) | 2019-08-02 | 2019-08-02 | Remote sensing image dictionary learning and classifying method utilizing spatial relationship |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110414616B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160387B (en) * | 2019-11-28 | 2022-06-03 | 广东工业大学 | Graph model based on multi-view dictionary learning |
| CN111860178B (en) * | 2020-06-22 | 2021-03-23 | 中国科学院空天信息创新研究院 | Small sample remote sensing target detection method and system based on weight dictionary learning |
| CN112163616B (en) * | 2020-09-25 | 2022-04-22 | 电子科技大学 | Local sparse constraint transformation RCS sequence feature extraction method |
| CN112507908B (en) * | 2020-12-15 | 2024-05-31 | 国网陕西省电力公司电力科学研究院 | Collaborative remote sensing aerial photographing system and method |
| CN112487231B (en) * | 2020-12-17 | 2021-08-27 | 中国矿业大学(北京) | Automatic image labeling method based on double-image regularization constraint and dictionary learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440500A (en) * | 2013-08-30 | 2013-12-11 | 西北工业大学 | Hyperspectral remote sensing image classifying and recognizing method |
| CN105095913A (en) * | 2015-08-11 | 2015-11-25 | 湖北大学 | Remote sensing image classification method and system based on neighbor regular joint sparse representation |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9152881B2 (en) * | 2012-09-13 | 2015-10-06 | Los Alamos National Security, Llc | Image fusion using sparse overcomplete feature dictionaries |
-
2019
- 2019-08-02 CN CN201910711662.2A patent/CN110414616B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440500A (en) * | 2013-08-30 | 2013-12-11 | 西北工业大学 | Hyperspectral remote sensing image classifying and recognizing method |
| CN105095913A (en) * | 2015-08-11 | 2015-11-25 | 湖北大学 | Remote sensing image classification method and system based on neighbor regular joint sparse representation |
Non-Patent Citations (1)
| Title |
|---|
| 高光谱遥感影像稀疏表示与字典学习分类研究;甘乐;《中国博士学位论文全文数据库基础科学辑》;20190215;第74-98页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110414616A (en) | 2019-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414616B (en) | Remote sensing image dictionary learning and classifying method utilizing spatial relationship | |
| CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
| CN111860612B (en) | Unsupervised hyperspectral image hidden low-rank projection learning feature extraction method | |
| CN110414377B (en) | Remote sensing image scene classification method based on scale attention network | |
| CN111695467B (en) | Spatial spectrum full convolution hyperspectral image classification method based on super-pixel sample expansion | |
| CN108596055B (en) | Airport target detection method of high-resolution remote sensing image under complex background | |
| CN113160192A (en) | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background | |
| Masci et al. | Multi-scale pyramidal pooling network for generic steel defect classification | |
| CN108229551B (en) | Hyperspectral remote sensing image classification method based on compact dictionary sparse representation | |
| CN111639587B (en) | Hyperspectral image classification method based on multi-scale spectrum space convolution neural network | |
| CN111695456B (en) | Low-resolution face recognition method based on active discriminant cross-domain alignment | |
| CN115496928B (en) | Multi-modal image feature matching method based on multi-feature matching | |
| Zhang et al. | Road recognition from remote sensing imagery using incremental learning | |
| CN106874862B (en) | Crowd counting method based on sub-model technology and semi-supervised learning | |
| CN109635726B (en) | Landslide identification method based on combination of symmetric deep network and multi-scale pooling | |
| CN110188827A (en) | A kind of scene recognition method based on convolutional neural networks and recurrence autocoder model | |
| CN112270317A (en) | Traditional digital water meter reading identification method based on deep learning and frame difference method | |
| CN109002771B (en) | Remote sensing image classification method based on recurrent neural network | |
| CN110717531A (en) | Method for detecting classified change type based on uncertainty analysis and Bayesian fusion | |
| CN114972885A (en) | Multi-modal remote sensing image classification method based on model compression | |
| CN110991374B (en) | Fingerprint singular point detection method based on RCNN | |
| CN113657414B (en) | Object identification method | |
| CN108960005B (en) | Method and system for establishing and displaying object visual label in intelligent visual Internet of things | |
| Rafi et al. | Attention-based domain adaptation for hyperspectral image classification | |
| CN116993760A (en) | Gesture segmentation method, system, device and medium based on graph convolution and attention mechanism |
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 |