CN111951285A - Optical remote sensing image woodland classification method based on cascade deep convolutional neural network - Google Patents
Optical remote sensing image woodland classification method based on cascade deep convolutional neural network Download PDFInfo
- Publication number
- CN111951285A CN111951285A CN202010807756.2A CN202010807756A CN111951285A CN 111951285 A CN111951285 A CN 111951285A CN 202010807756 A CN202010807756 A CN 202010807756A CN 111951285 A CN111951285 A CN 111951285A
- Authority
- CN
- China
- Prior art keywords
- image
- remote sensing
- neural network
- network
- forest
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 22
- 230000003287 optical effect Effects 0.000 title claims abstract description 19
- 230000011218 segmentation Effects 0.000 claims abstract description 14
- 238000013528 artificial neural network Methods 0.000 claims abstract description 5
- 238000012545 processing Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 15
- 238000002372 labelling Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 8
- 238000013145 classification model Methods 0.000 claims description 6
- 235000017166 Bambusa arundinacea Nutrition 0.000 claims description 5
- 235000017491 Bambusa tulda Nutrition 0.000 claims description 5
- 241001330002 Bambuseae Species 0.000 claims description 5
- 235000015334 Phyllostachys viridis Nutrition 0.000 claims description 5
- 239000011425 bamboo Substances 0.000 claims description 5
- 238000003709 image segmentation Methods 0.000 claims description 5
- 241000218631 Coniferophyta Species 0.000 claims description 4
- 230000000903 blocking effect Effects 0.000 claims description 4
- 239000002689 soil Substances 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 238000000605 extraction Methods 0.000 abstract description 5
- 238000013135 deep learning Methods 0.000 abstract description 3
- 230000007547 defect Effects 0.000 abstract description 2
- 238000011835 investigation Methods 0.000 description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 239000002028 Biomass Substances 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30204—Marker
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an optical remote sensing image woodland classification method based on a cascade deep convolutional neural network, which comprises the following steps of: A. classifying large-granularity ground objects in the image by utilizing an image semantic segmentation depth convolution network, and realizing the division of forest land and non-forest land areas; B. and realizing fine granularity fine division for the woodland area by using the depth convolution neural network facing the image classification. The method can solve the defects of the prior art, can effectively mine the ground feature characteristics of the remote sensing image by performing deep learning, simultaneously gives consideration to efficient learning and extraction of the forest land fine characteristics in the high-resolution image, and can better solve the problem of fine classification of the forest land of the high-resolution remote sensing image.
Description
Technical Field
The invention belongs to the technical field of remote sensing, and particularly relates to an optical remote sensing image woodland classification method based on a cascade deep convolutional neural network.
Background
The remote sensing image forest land fine classification is an important step in forest resource investigation and has important application value. On one hand, the method is an efficient and economic forest land spatial distribution information and change information acquisition means, on the other hand, the method is also an important aid for forest biomass analysis, and especially has great significance for sustainable development and utilization of forest resources.
At present, the forest land fine classification is mostly realized by adopting a traditional remote sensing image classification method, and is usually realized by combining a shallow classifier learning (such as maximum likelihood classification and a support vector machine) method based on manually designed image characteristics (such as image spectrum, texture and the like). However, due to the complexity of the coverage of the earth surface and the forest land, the representation of the earth surface and the forest land in the high-resolution remote sensing image is complicated, and the accurate classification of the manually designed image characteristics and the shallow structure model has great limitation; and the deep-level structure model has stronger expression and modeling capacity due to the complex multilayer nonlinear transformation, so that the deep-level structure model is more suitable for processing complex signals. Therefore, the remote sensing image classification method based on the deep convolutional neural network is the main research direction of the current forest land classification technology. However, for high-resolution imagery, due to the complexity of image feature features and the imbalance among different feature type samples, a method based on a single deep neural network model often cannot well solve the problem of forest land fine classification. The adaptability problem of the depth network to classification processing of large-particle ground objects (such as farmlands, towns, water bodies, vegetation and the like) and small-particle forest land classification processing in the images is solved; particularly, forest landforms in high-resolution remote sensing images have diversity and complexity, which can cause poor fine classification effect, and the fine classification effect is embodied in the following two aspects:
1) for the semantic segmentation deep network, because the types of the forest lands are complex and various and the boundaries of different forest land types are staggered, the artificial labeling is difficult, and better sample data is difficult to establish to train the network; in addition, compared with large-area areas such as farmlands, water areas and the like, fine woodland type areas are small, so that the problem of unbalanced training samples can be solved.
2) For an image classification depth network, different from a traditional close-range image, due to the fact that the feature of the ground feature details in the high-resolution remote sensing image is very rich, the number of categories is too large, and therefore the acquisition of sample data is influenced; in addition, the small-slice (about 80 × 80) method processed by the image classification network uses less context information, and is not beneficial to large-size and large-range remote sensing image classification processing.
Disclosure of Invention
The technical problem to be solved by the invention is to provide an optical remote sensing image woodland classification method based on a cascade deep convolutional neural network, which can solve the defects of the prior art, can give play to the effective excavation of the remote sensing image ground feature by deep learning, simultaneously gives consideration to the efficient learning and extraction of the woodland fine feature in a high-resolution image, and can better solve the problem of the woodland fine classification of the high-resolution remote sensing image.
The subject matter of the present invention includes the following steps,
A. classifying large-granularity ground objects in the image by utilizing an image semantic segmentation depth convolution network, and realizing the division of forest land and non-forest land areas;
B. and realizing fine granularity fine division for the woodland area by using the depth convolution neural network facing the image classification.
Preferably, in the step A, the deep network DeepLab-V3 is trained offline first, and then the deep network is segmented online based on the DeepLab-V3.
Preferably, the offline training of the deep network of deep Lab-V3 includes the steps of,
firstly, cutting a plurality of small-format remote sensing images from a large-format remote sensing image, and then labeling the small-format remote sensing images; then expanding the small-breadth remote sensing image and the true value annotation image; and finally inputting a DeepLab-V3 network for training a segmentation model.
Preferably, the size of the cut small-format remote sensing image is 3000 × 2000; labeling the small-breadth remote sensing image, including farmland, forest land, bare soil, river, town and road; the expansion operation comprises synchronous random cutting, turning, rotation, chrominance change, light and shade change and noise interference; the expansion result is 10 ten thousand pairs of 256 × 256 remote sensing image blocks and truth value labeling image blocks.
Preferably, the deep network online image segmentation based on deep Lab-V3 comprises the following steps,
carrying out block processing on the large-format test image in different window sizes, wherein 30% of blocks are overlapped; and then inputting the test image block into a trained DeepLab-V3 network model for forward operation to obtain a rough terrain classification result.
Preferably, the size of the blocking process window is 256 × 256, 300 × 300, 500 × 500.
Preferably, the step B specifically comprises the following steps,
b1, training an offline Squeezenet lightweight network model and parameters;
b2, image super-pixel processing;
and B3, carrying out online forest land classification based on the SqueezeNet lightweight network.
Preferably, in step B1, different forest lands of known type are first marked and cropped in the high resolution image, the marked types being arbor forest, conifer forest, bamboo forest and shrub, the average image block size of the cropping being about 80 × 80 pixels; and then performing data expansion on the image blocks by means of random cutting, rotation, chrominance conversion and the like to obtain 8-10 ten thousand of small woodland image blocks for training a deep convolutional neural network classification model.
Preferably, in step B2, the forest land area obtained by the first stage processing is first subjected to superpixel calculation to obtain several adjacent image slices; and then taking the minimum bounding rectangle of each super-pixel slice as the input of the deep convolutional neural network classification model.
Preferably, in step B3, online woodland classification is performed based on the SqueezeNet lightweight network; inputting the images in the range of each super-pixel slice area into a network for forward operation, and assigning a predicted label category to each pixel; and finally, splicing and combining the super-pixel type judgment results to obtain a final large-breadth image woodland fine classification result.
The method has the advantages that the two-stage cascade deep neural network forest land fine classification processing framework can effectively solve the problem of forest land fine classification in the large-breadth high-resolution remote sensing image, so that intelligent information service is provided for remote sensing applications such as forest resource investigation, forest land change investigation and the like, the labor cost is saved, and the investigation efficiency is improved; meanwhile, the method can serve the field of land resource investigation.
Drawings
Fig. 1 is a flow chart of forest land fine classification processing proposed by the present invention.
Fig. 2 is an optical remote sensing image of a high-resolution 2 # satellite.
FIG. 3 is a diagram of the fine classification results of the forest lands of FIG. 2, wherein a is a town, b is a road, c is a farmland, d is a water body, e is bare soil, f is a shrub, g is a arbor forest, h is a conifer forest, and i is a bamboo forest.
Detailed Description
The method comprises the following steps of,
A. classifying large-granularity ground objects in the image by utilizing an image semantic segmentation depth convolution network, and realizing the division of forest land and non-forest land areas;
B. and realizing fine granularity fine division for the woodland area by using the depth convolution neural network facing the image classification.
In the step A, firstly, deep network deep DeepLab-V3 is trained offline, and then online image segmentation is carried out based on the deep network DeepLab-V3.
Offline training of the deep network of deep lab-V3 includes the following steps,
firstly, cutting a plurality of small-format remote sensing images from a large-format remote sensing image, and then labeling the small-format remote sensing images; then expanding the small-breadth remote sensing image and the true value annotation image; and finally inputting a DeepLab-V3 network for training a segmentation model.
The size of the cut small-breadth remote sensing image is 3000 × 2000; labeling the small-breadth remote sensing image, including farmland, forest land, bare soil, river, town and road; the expansion operation comprises synchronous random cutting, turning, rotation, chrominance change, light and shade change and noise interference; the expansion result is 10 ten thousand pairs of 256 × 256 remote sensing image blocks and truth value labeling image blocks.
The deep network online image segmentation based on deep lab-V3 comprises the following steps,
carrying out block processing on the large-format test image in different window sizes, wherein 30% of blocks are overlapped; and then inputting the test image block into a trained DeepLab-V3 network model for forward operation to obtain a rough terrain classification result.
The size of the blocking process window is 256 × 256, 300 × 300, 500 × 500.
The step B specifically comprises the following steps,
b1, training an offline Squeezenet lightweight network model and parameters;
b2, image super-pixel processing;
and B3, carrying out online forest land classification based on the SqueezeNet lightweight network.
In step B1, different forest lands of known types are marked and cut in the high-resolution image, the marked forest is arbor forest, conifer forest, bamboo forest and shrub, and the average image block size of cutting is about 80 × 80 pixels; and then performing data expansion on the image blocks by means of random cutting, rotation, chrominance conversion and the like to obtain 8-10 ten thousand of small woodland image blocks for training a deep convolutional neural network classification model.
In step B2, first, performing superpixel calculation on the woodland area obtained by the first stage processing to obtain a plurality of adjacent image slices; and then taking the minimum bounding rectangle of each super-pixel slice as the input of the deep convolutional neural network classification model.
In step B3, classifying forest lands on line based on the SqueezeNet lightweight network; inputting the images in the range of each super-pixel slice area into a network for forward operation, and assigning a predicted label category to each pixel; and finally, splicing and combining the super-pixel type judgment results to obtain a final large-breadth image woodland fine classification result.
Fig. 1 is divided into an offline processing part and an online processing part, wherein the upper part of the figure is a training process (offline processing) of a deep network model, and the lower part is a testing process (online processing) of the network. The testing process is realized by two stages, wherein the first stage is a forest land area extraction stage based on image semantic segmentation, and high-resolution No. 2 optical remote sensing image data is input; the second stage is a forest land type fine division stage, and the input image is forest land area image data after the segmentation processing. Firstly, training network model parameters by using the existing sample data to realize the learning process of a semantic segmentation network and a ground feature classification network model, wherein the two learning processes can adopt an offline parallel processing mode; then, carrying out block segmentation on the original large-breadth high-resolution remote sensing image by using multi-size windows (for example, 256 × 256 pixels, 300 × 300 pixels and 500 × 500 pixels), and pushing image block data to a semantic segmentation network to carry out forest region extraction; performing super-pixel blocking on the forest land area data subjected to semantic segmentation; then, determining a forest land type label for the super pixel block by using an image classification network at the second stage, and realizing fine classification of the forest land; and finally, combining the super pixel blocks with the determined categories to obtain the forest land fine classification result of the whole map.
Fig. 2 is a large-format high-resolution 2 # optical remote sensing image (a high-resolution color image after preprocessing), and the size of the image is 29200 × 27620 pixels.
FIG. 3 is a diagram showing the result of fine classification of a large-area high-resolution No. 2 optical remote sensing image forest land by using the method of the present invention, wherein a type label is arranged on the right side of the classification result diagram. Totally divided into 9 types including farmlands, bare land, water, towns, roads, arbor woodlands, coniferous lands, shrub woodlands and bamboo woodlands. The method can effectively mine the ground feature characteristics of the remote sensing image by performing deep learning, simultaneously gives consideration to efficient learning and extraction of the forest land fine characteristics in the high-resolution image, and can better solve the problem of fine classification of the forest land of the high-resolution remote sensing image.
Claims (10)
1. A method for classifying forest lands of optical remote sensing images based on a cascade deep convolutional neural network is characterized by comprising the following steps,
A. classifying large-granularity ground objects in the image by utilizing an image semantic segmentation depth convolution network, and realizing the division of forest land and non-forest land areas;
B. and realizing fine granularity fine division for the woodland area by using the depth convolution neural network facing the image classification.
2. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 1, wherein: in the step A, firstly, deep network deep DeepLab-V3 is trained offline, and then online image segmentation is carried out based on the deep network DeepLab-V3.
3. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 2, wherein: offline training of the deep network of deep lab-V3 includes the following steps,
firstly, cutting a plurality of small-format remote sensing images from a large-format remote sensing image, and then labeling the small-format remote sensing images; then expanding the small-breadth remote sensing image and the true value annotation image; and finally inputting a DeepLab-V3 network for training a segmentation model.
4. The cascade deep convolutional neural network-based optical remote sensing image woodland classification method as claimed in claim 3, characterized in that: the size of the cut small-breadth remote sensing image is 3000 × 2000; labeling the small-breadth remote sensing image, including farmland, forest land, bare soil, river, town and road; the expansion operation comprises synchronous random cutting, turning, rotation, chrominance change, light and shade change and noise interference; the expansion result is 10 ten thousand pairs of 256 × 256 remote sensing image blocks and truth value labeling image blocks.
5. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 2, wherein: the deep network online image segmentation based on deep lab-V3 comprises the following steps,
carrying out block processing on the large-format test image in different window sizes, wherein 30% of blocks are overlapped; and then inputting the test image block into a trained DeepLab-V3 network model for forward operation to obtain a rough terrain classification result.
6. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 5, wherein: the size of the blocking process window is 256 × 256, 300 × 300, 500 × 500.
7. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 1, wherein: the step B specifically comprises the following steps,
b1, training an offline Squeezenet lightweight network model and parameters;
b2, image super-pixel processing;
and B3, carrying out online forest land classification based on the SqueezeNet lightweight network.
8. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 7, wherein: in step B1, different forest lands of known types are marked and cut in the high-resolution image, the marked forest is arbor forest, conifer forest, bamboo forest and shrub, and the average image block size of cutting is about 80 × 80 pixels; and then performing data expansion on the image blocks by means of random cutting, rotation, chrominance conversion and the like to obtain 8-10 ten thousand of small woodland image blocks for training a deep convolutional neural network classification model.
9. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 7, wherein: in step B2, first, performing superpixel calculation on the woodland area obtained by the first stage processing to obtain a plurality of adjacent image slices; and then taking the minimum bounding rectangle of each super-pixel slice as the input of the deep convolutional neural network classification model.
10. The method for classifying the forest land of the optical remote sensing image based on the cascaded deep convolutional neural network as claimed in claim 7, wherein: in step B3, classifying forest lands on line based on the SqueezeNet lightweight network; inputting the images in the range of each super-pixel slice area into a network for forward operation, and assigning a predicted label category to each pixel; and finally, splicing and combining the super-pixel type judgment results to obtain a final large-breadth image woodland fine classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010807756.2A CN111951285A (en) | 2020-08-12 | 2020-08-12 | Optical remote sensing image woodland classification method based on cascade deep convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010807756.2A CN111951285A (en) | 2020-08-12 | 2020-08-12 | Optical remote sensing image woodland classification method based on cascade deep convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111951285A true CN111951285A (en) | 2020-11-17 |
Family
ID=73332422
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010807756.2A Pending CN111951285A (en) | 2020-08-12 | 2020-08-12 | Optical remote sensing image woodland classification method based on cascade deep convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111951285A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329703A (en) * | 2020-11-20 | 2021-02-05 | 北京林业大学 | Construction method of deep convolutional neural network suitable for identifying remote sensing images of pine wilt disease |
| CN112668911A (en) * | 2020-12-31 | 2021-04-16 | 哈尔滨工业大学 | Water area shoreline pollution risk assessment method and control intelligent decision method and system |
| CN113077441A (en) * | 2021-03-31 | 2021-07-06 | 上海联影智能医疗科技有限公司 | Coronary artery calcified plaque segmentation method and method for calculating coronary artery calcified score |
| CN114140637A (en) * | 2021-10-21 | 2022-03-04 | 阿里巴巴达摩院(杭州)科技有限公司 | Image classification method, storage medium and electronic device |
| CN116310826A (en) * | 2023-03-20 | 2023-06-23 | 中国科学技术大学 | High-resolution remote sensing image forest land secondary classification method based on graphic neural network |
| CN117611909A (en) * | 2023-12-04 | 2024-02-27 | 桂林理工大学 | Wetland vegetation classification method based on deep learning and image spatial resolution |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255334A (en) * | 2018-09-27 | 2019-01-22 | 中国电子科技集团公司第五十四研究所 | Remote sensing image terrain classification method based on deep learning semantic segmentation network |
| CN111047565A (en) * | 2019-11-29 | 2020-04-21 | 南京恩博科技有限公司 | Method, storage medium and equipment for forest cloud image segmentation |
| CN111339921A (en) * | 2020-02-24 | 2020-06-26 | 南京邮电大学 | Insect disease detection unmanned aerial vehicle based on lightweight convolutional neural network and detection method |
-
2020
- 2020-08-12 CN CN202010807756.2A patent/CN111951285A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255334A (en) * | 2018-09-27 | 2019-01-22 | 中国电子科技集团公司第五十四研究所 | Remote sensing image terrain classification method based on deep learning semantic segmentation network |
| CN111047565A (en) * | 2019-11-29 | 2020-04-21 | 南京恩博科技有限公司 | Method, storage medium and equipment for forest cloud image segmentation |
| CN111339921A (en) * | 2020-02-24 | 2020-06-26 | 南京邮电大学 | Insect disease detection unmanned aerial vehicle based on lightweight convolutional neural network and detection method |
Non-Patent Citations (1)
| Title |
|---|
| 蔡烁等: "基于深度卷积网络的高分遥感图像语义分割", 信号处理, vol. 35, no. 12, 31 December 2019 (2019-12-31), pages 2010 - 2016 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329703A (en) * | 2020-11-20 | 2021-02-05 | 北京林业大学 | Construction method of deep convolutional neural network suitable for identifying remote sensing images of pine wilt disease |
| CN112668911A (en) * | 2020-12-31 | 2021-04-16 | 哈尔滨工业大学 | Water area shoreline pollution risk assessment method and control intelligent decision method and system |
| CN113077441A (en) * | 2021-03-31 | 2021-07-06 | 上海联影智能医疗科技有限公司 | Coronary artery calcified plaque segmentation method and method for calculating coronary artery calcified score |
| CN114140637A (en) * | 2021-10-21 | 2022-03-04 | 阿里巴巴达摩院(杭州)科技有限公司 | Image classification method, storage medium and electronic device |
| CN114140637B (en) * | 2021-10-21 | 2023-09-12 | 阿里巴巴达摩院(杭州)科技有限公司 | Image classification method, storage medium and electronic device |
| CN116310826A (en) * | 2023-03-20 | 2023-06-23 | 中国科学技术大学 | High-resolution remote sensing image forest land secondary classification method based on graphic neural network |
| CN116310826B (en) * | 2023-03-20 | 2023-09-22 | 中国科学技术大学 | High-resolution remote sensing image forest land secondary classification method based on graphic neural network |
| CN117611909A (en) * | 2023-12-04 | 2024-02-27 | 桂林理工大学 | Wetland vegetation classification method based on deep learning and image spatial resolution |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111951285A (en) | Optical remote sensing image woodland classification method based on cascade deep convolutional neural network | |
| CN110619282B (en) | Automatic extraction method for unmanned aerial vehicle orthoscopic image building | |
| CN111986099B (en) | Tillage monitoring method and system based on convolutional neural network with residual error correction fused | |
| CN109255334B (en) | Remote sensing image ground feature classification method based on deep learning semantic segmentation network | |
| CN108921799B (en) | Remote sensing image thin cloud removing method based on multi-scale collaborative learning convolutional neural network | |
| CN111274865A (en) | Remote sensing image cloud detection method and device based on full convolution neural network | |
| CN109871798A (en) | A kind of remote sensing image building extracting method based on convolutional neural networks | |
| CN117078943B (en) | Remote sensing image road segmentation method integrating multi-scale features and double-attention mechanism | |
| CN113343789A (en) | High-resolution remote sensing image land cover classification method based on local detail enhancement and edge constraint | |
| CN112232328A (en) | Remote sensing image building area extraction method and device based on convolutional neural network | |
| CN113989681A (en) | Remote sensing image change detection method and device, electronic equipment and storage medium | |
| CN108932474B (en) | Remote sensing image cloud judgment method based on full convolution neural network composite characteristics | |
| CN113569815A (en) | Method for detecting remote sensing image change based on image segmentation and twin neural network | |
| CN116343053B (en) | Automatic solid waste extraction method based on fusion of optical remote sensing image and SAR remote sensing image | |
| CN112419197A (en) | Universal single-time phase and multi-time phase SAR image speckle noise removing method | |
| CN111028244B (en) | Remote sensing image semantic segmentation method based on super-pixel under condition of known sample imbalance | |
| CN113505636A (en) | Mining area change detection method based on attention mechanism and full convolution twin neural network | |
| CN115661032A (en) | Intelligent pavement disease detection method suitable for complex background | |
| CN115862010A (en) | High-resolution remote sensing image water body extraction method based on semantic segmentation model | |
| CN114743103A (en) | Island reef remote sensing image geological classification method based on Deeplabv3+ network model | |
| CN114943902A (en) | Urban vegetation unmanned aerial vehicle remote sensing classification method based on multi-scale feature perception network | |
| CN114387446A (en) | Automatic water body extraction method for high-resolution remote sensing image | |
| CN113158770A (en) | Improved mining area change detection method of full convolution twin neural network | |
| CN115546649B (en) | Single-view remote sensing image height estimation and semantic segmentation multi-task prediction method | |
| CN110853058A (en) | High-resolution remote sensing image road extraction method based on visual saliency detection |
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 |