CN111553222A - Remote sensing ground feature classification post-processing method based on iteration superpixel segmentation - Google Patents
Remote sensing ground feature classification post-processing method based on iteration superpixel segmentation Download PDFInfo
- Publication number
- CN111553222A CN111553222A CN202010316759.6A CN202010316759A CN111553222A CN 111553222 A CN111553222 A CN 111553222A CN 202010316759 A CN202010316759 A CN 202010316759A CN 111553222 A CN111553222 A CN 111553222A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- image
- complexity
- segmentation
- super
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a remote sensing ground feature classification post-processing method based on iteration superpixel segmentation, and belongs to the technical field of intelligent application of remote sensing images. The method comprises the steps of firstly generating a preliminary classification result of a remote sensing image by adopting a deep learning ground feature classification method, then carrying out preliminary super-pixel segmentation on the remote sensing image by adopting an SLIC algorithm according to the resolution ratio and imaging effect of the remote sensing image and setting reasonable parameters, then calculating the complexity of a super-pixel block of each super-pixel block to judge whether secondary or even multi-time iterative segmentation is needed, calculating the complexity of a pattern spot region corresponding to the super-pixel block which does not need to be further segmented, carrying out different post-processing strategies according to the complexity, and finally carrying out category combination by taking the super-pixel with the minimum granularity as a basic unit to output a final post-processing result. The method can optimize the ground feature classification result with weak dependence on artificial parameters, and has better precision improvement and service application capability.
Description
Technical Field
The invention belongs to the technical field of intelligent application of remote sensing images, and particularly relates to a remote sensing ground feature classification post-processing method based on iteration superpixel segmentation.
Background
With the gradual rise of various earth observation satellites of army, civilian and commerce and the use of the earth observation satellites, the remote sensing image data accumulated by each unit gradually reaches the million level at present, and with the concept of the micro-nano satellite, the constellation and the constellation, the subsequent remote sensing image data is increased in an explosive manner. In the face of such huge remote sensing image data, how to rapidly and automatically complete image analysis, especially surface feature type analysis, becomes a task to be completed urgently.
With the great progress of the artificial intelligence technologies such as deep learning and reinforcement learning, the method has wide application prospect in various fields. In the field of remote sensing intelligent processing, the current deep learning is applied to image processing tasks such as ground feature classification, target detection and the like, and various colleges and universities develop corresponding technical researches at home and abroad to produce a plurality of technical achievements. The current deep learning solution which is closer to the remote sensing ground feature classification task is semantic segmentation, and in addition, the current optimal semantic segmentation network design scheme at home and abroad is a full convolution neural network and an improved network structure thereof. Although the full convolution type neural network has a good effect in a remote sensing image ground object classification task, the classification precision of certain typical ground object types (such as water bodies, forests and the like) reaches more than 90%, the compression distortion in the up-sampling and down-sampling process is inevitably caused by the limitation of a convolution structure in a network structure, and the occurrence of errors such as edge blurring, classification holes and the like is inevitable. With the development of the current neural network, although the error phenomenon can be improved to a certain extent by improving the network convolution kernel distribution and other strategies, the controllability is poor due to the black box effect inside the network, and the real service application level is difficult to achieve. Therefore, how to automatically, efficiently and manually and controllably optimize the network classification result is a crucial and urgent problem to be solved.
At present, how to optimize the remote sensing image surface feature classification result is lack of a mature solution way in the industry, and only an attempt is made from an experimental theory stage.
In recent years, a CRF conditional random field is widely researched to optimize a CNN classification result, although the classification result is improved to a certain extent, CRF needs to manually set a plurality of sets of experience parameters such as kernel functions, the optimal experience parameters needed by different images are different, the optimal values need to be determined through a plurality of attempts, and frequent manual parameter setting is unrealistic in the face of the huge remote sensing data volume. The real realization of the automation of the service application should avoid such manual multiple intervention process, so the method of strong dependence of the CRF and other manual parameters cannot meet the current service automation requirement.
Disclosure of Invention
The invention aims to overcome the problems of fuzzy edges, classification holes and other errors of the current massive remote sensing ground object classification results, and provides a remote sensing ground object classification post-processing method based on iteration superpixel segmentation.
The purpose of the invention is realized as follows:
a remote sensing ground feature classification post-processing method based on iteration superpixel segmentation is used for post-processing a ground feature classification preliminary result graph and comprises the following steps:
(1) performing superpixel segmentation on the original remote sensing image by adopting an SLIC (narrow-line segmentation and segmentation in-situ correlation) method to obtain a dense superpixel block;
(2) calculating a complexity parameter M of the super pixel block, comparing the complexity parameter M of the super pixel block with a threshold value U, executing the step (3) if the complexity parameters of all the super pixel blocks are smaller than U, and returning to the step (1) to perform super pixel segmentation again if the complexity parameters of all the super pixel blocks are smaller than U; wherein M is used for describing the complexity of gray scale values and spatial distribution in the super pixel block;
(3) extracting the outline region position of each super pixel block, extracting the pattern spots of the outline region position from the ground feature classification preliminary result image, and calculating the complexity parameter S of each pattern spot; wherein, S is used for describing the complexity of the category distribution in the image spot;
(4) for each pattern spot, comparing the complexity parameter S with a threshold value L, if S is smaller than L, keeping the original type distribution of the pattern spot, otherwise, filling the pattern spot by using the type with the largest occupation ratio in the pattern spot, wherein all the types in the filled pattern spots are the types with the largest occupation ratio;
(5) and (4) combining all the image spots processed in the step (4) to form a vector image with the size of the original image, wherein the vector image is the ground object classification result after the post-processing.
Further, in the step (2), the parameter M is calculated in the following manner:
wherein m in the superpixel blockiIs the number of pixels with a pixel value of i, N is the super-imageNumber of pixels in a pixel block, h is the height of the superpixel patch, wiThe width of the super pixel block when the height is i, p (i, j) is the pixel value with the coordinate position of i, j, aver (i, j, k) is the pixel value mean value of a matrix with i, j as the center and k as the width;
the calculation mode of the parameter S is as follows:
S＝narea*A/N
wherein n isareaThe number of the land feature types contained in the image spot is A, the number of the land features obtained after the contour statistics is carried out on the image spot is A, and the number of pixels in the image spot is N.
Compared with the background technology, the invention has the following advantages:
1. the invention provides a brand-new ground feature classification post-processing method, namely iterative superpixels post-processing ISR (iterative superpixels reprocessing), which can realize the improvement of the post-processing precision of universality without manually setting different experience parameters for different remote sensing images, and realize the real automatic service application.
2. The method has complexity evaluation on different superpixel segmentation results, and automatically judges whether to perform iterative segmentation according to the evaluation results so as to achieve the optimal segmentation superpixel granularity.
3. The invention has different post-processing strategies for different classification effects, thereby preventing the occurrence of over-processing.
Drawings
FIG. 1 is a flow chart of a method of an embodiment of the present invention.
FIG. 2 is a sample diagram of classification results before post-processing in an embodiment of the present invention.
FIG. 3 is a sample diagram of the segmentation after the initial super-pixel segmentation according to the embodiment of the present invention.
FIG. 4 is a diagram of iterative superpixel segmentation results in an embodiment of the present invention.
FIG. 5 is a schematic diagram of a super-pixel with better segmentation effect in the embodiment of the present invention.
FIG. 6 is a schematic diagram of a super-pixel without further segmentation in an embodiment of the present invention.
FIG. 7 is a graph of the results of the final post-processing in an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.
As shown in fig. 1, a remote sensing ground feature classification post-processing method based on iterative superpixel segmentation includes the following steps:
(1) performing superpixel segmentation on an original remote sensing image by adopting a Linear Iterative clustering SLIC (simple Linear Iterative clustering) method to obtain a dense superpixel block;
(2) calculating a complexity parameter M of the super pixel block, comparing the complexity parameter M of the super pixel block with a threshold value U, executing the step (3) if the complexity parameters of all the super pixel blocks are smaller than U, and returning to the step (1) to perform super pixel segmentation again if the complexity parameters of all the super pixel blocks are smaller than U; wherein M is used for describing the complexity of gray scale values and spatial distribution in the super pixel block;
(3) extracting the outline region position of each super pixel block, extracting the pattern spots of the outline region position from the ground feature classification preliminary result image, and calculating the complexity parameter S of each pattern spot; wherein, S is used for describing the complexity of the category distribution in the image spot;
(4) for each pattern spot, comparing the complexity parameter S with a threshold value L, if S is smaller than L, keeping the original type distribution of the pattern spot, otherwise, filling the pattern spot by using the type with the largest occupation ratio in the pattern spot, wherein all the types in the filled pattern spots are the types with the largest occupation ratio;
(5) and (4) combining all the image spots processed in the step (4) to form a vector image with the size of the original image, wherein the vector image is the ground object classification result after the post-processing.
The method carries out post-processing on a ground feature classification preliminary result graph obtained by classifying ground features by adopting a deplab deep learning network, so that the edge of a classification result is optimized, the classification is omitted, and the classification precision is improved.
In the above method, the complexity parameter M of the super pixel block is calculated by the formula,
in the formula, m in the superpixel blockiIs the number of pixels with a pixel value of i, N is the number of pixels in the superpixel block, h is the height of the superpixel patch, wiThe width of the super pixel block when the height is i, p (i, j) is the pixel value with the coordinate position of i, j, aver (i, j, k) is the pixel value mean value of a matrix with i, j as the center and k as the width;
the calculation mode of the complexity parameter S of the image spot is as follows:
S＝narea*A/N
in the formula, nareaThe number of the land feature types contained in the image spot is A, the number of the land features obtained after the contour statistics is carried out on the image spot is A, and the number of pixels in the image spot is N.
Taking a building as an example, the original remote sensing image and the preliminary result of the ground feature classification of the method are shown in fig. 2. According to the method, the SLIC algorithm with few seed nodes is adopted to carry out super-pixel segmentation on an original remote sensing image, and compared with the traditional single SLIC classification (shown in figure 3), the SLIC algorithm with few seed nodes obtains a non-dense super-pixel block (shown in figure 4).
Then, the method sets various combined processing strategies according to the distribution conditions of M and S, so that each super-pixel block and the corresponding image spots are subjected to post-processing, and the processed image spots are obtained. Fig. 5 and 6 are schematic diagrams of superpixels with better and poorer segmentation effects, respectively.
Then, all the processed image patches are combined to form a vector image with the size of the original image, which is the final post-processed ground object classification result, and the comparison result before and after post-processing is shown in fig. 7. Through a comparison test on the classification precision of the land features before and after post-treatment, compared with the classification result of the land features before treatment, the method improves the precision by 5.24%, and has great performance improvement.
According to the method, firstly, a deep learning ground feature classification method is adopted for the remote sensing image to carry out primary classification result generation, and then a SLIC algorithm is adopted to set reasonable parameters according to the resolution ratio and the imaging effect of the remote sensing image to carry out primary superpixel segmentation on the remote sensing image. And then calculating the complexity of the super pixel blocks for each super pixel block to judge whether secondary or even multiple iterative segmentation is needed. And performing the complexity calculation of the image spots on the image spot regions corresponding to the superpixel blocks which do not need to be further divided, performing different post-processing strategies according to the complexity, and finally performing category merging by taking the superpixel with the minimum granularity as a basic unit to output a final post-processing result.
In a word, the method carries out post-processing on the ground feature classification preliminary result graph obtained by classifying the ground features by adopting the deep learning network of deplab, so that the edge of the classification result is optimized, the classification is omitted and the classification precision is improved. The method is based on a super-pixel segmentation method, realizes the segmentation judgment of sub super-pixels and the independent multi-strategy processing of each super-pixel block in an autonomous iteration mode, can optimize the ground feature classification result with artificial parameters weakly dependent, and has better precision improvement and service application capability.
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all matters of the invention which utilize the inventive concepts are protected.
Claims (2)
1. A remote sensing surface feature classification post-processing method based on iteration superpixel segmentation is characterized by being used for post-processing a surface feature classification preliminary result graph and comprising the following steps:
(1) performing superpixel segmentation on the original remote sensing image by adopting an SLIC (narrow-line segmentation and segmentation in-situ correlation) method to obtain a dense superpixel block;
(2) calculating a complexity parameter M of the super pixel block, comparing the complexity parameter M of the super pixel block with a threshold value U, executing the step (3) if the complexity parameters of all the super pixel blocks are smaller than U, and returning to the step (1) to perform super pixel segmentation again if the complexity parameters of all the super pixel blocks are smaller than U; wherein M is used for describing the complexity of gray scale values and spatial distribution in the super pixel block;
(3) extracting the outline region position of each super pixel block, extracting the pattern spots of the outline region position from the ground feature classification preliminary result image, and calculating the complexity parameter S of each pattern spot; wherein, S is used for describing the complexity of the category distribution in the image spot;
(4) for each pattern spot, comparing the complexity parameter S with a threshold value L, if S is smaller than L, keeping the original type distribution of the pattern spot, otherwise, filling the pattern spot by using the type with the largest occupation ratio in the pattern spot, wherein all the types in the filled pattern spots are the types with the largest occupation ratio;
(5) and (4) combining all the image spots processed in the step (4) to form a vector image with the size of the original image, wherein the vector image is the ground object classification result after the post-processing.
2. The remote sensing land feature classification post-processing method based on iterative superpixel segmentation as claimed in claim 1, wherein in said step (2), the calculation mode of parameter M is:
wherein m in the superpixel blockiIs the number of pixels with a pixel value of i, N is the number of pixels in the superpixel block, h is the height of the superpixel patch, wiThe width of the super pixel block when the height is i, p (i, j) is the pixel value with the coordinate position of i, j, aver (i, j, k) is the pixel value mean value of a matrix with i, j as the center and k as the width;
the calculation mode of the parameter S is as follows:
S＝narea*A/N
wherein n isareaThe number of the land feature types contained in the image spot is A, the number of the land features obtained after the contour statistics is carried out on the image spot is A, and the number of pixels in the image spot is N.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010316759.6A CN111553222B (en) | 2020-04-21 | 2020-04-21 | Remote sensing ground feature classification post-processing method based on iteration superpixel segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010316759.6A CN111553222B (en) | 2020-04-21 | 2020-04-21 | Remote sensing ground feature classification post-processing method based on iteration superpixel segmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111553222A true CN111553222A (en) | 2020-08-18 |
| CN111553222B CN111553222B (en) | 2021-11-05 |
Family
ID=72007575
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010316759.6A Active CN111553222B (en) | 2020-04-21 | 2020-04-21 | Remote sensing ground feature classification post-processing method based on iteration superpixel segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111553222B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114898216A (en) * | 2022-06-14 | 2022-08-12 | 湖南省农林工业勘察设计研究总院 | Remote sensing image deep learning segmentation method and system based on superpixels and watershed |
| CN116797598A (en) * | 2023-08-22 | 2023-09-22 | 山东万牧农业科技有限公司郯城分公司 | Image feature-based cultivation feed quality refinement detection method |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289807A (en) * | 2011-07-08 | 2011-12-21 | 西安电子科技大学 | Method for detecting change of remote sensing image based on Treelet transformation and characteristic fusion |
| US20170004472A1 (en) * | 2014-03-07 | 2017-01-05 | Pantry Retail, Inc. | Method for processing transactions at a vending kiosk |
| CN107067405A (en) * | 2017-03-30 | 2017-08-18 | 河海大学 | Based on the preferred Remote Sensing Image Segmentation of yardstick |
| CN107229917A (en) * | 2017-05-31 | 2017-10-03 | 北京师范大学 | A kind of several remote sensing image general character well-marked target detection methods clustered based on iteration |
| CN108647602A (en) * | 2018-04-28 | 2018-10-12 | 北京航空航天大学 | A kind of aerial remote sensing images scene classification method based on image complexity judgement |
| CN109377480A (en) * | 2018-09-27 | 2019-02-22 | 中国电子科技集团公司第五十四研究所 | Arable land use change detection method based on deep learning |
| CN109657610A (en) * | 2018-12-18 | 2019-04-19 | 北京航天泰坦科技股份有限公司 | A kind of land use change survey detection method of high-resolution multi-source Remote Sensing Images |
| CN109712153A (en) * | 2018-12-25 | 2019-05-03 | 杭州世平信息科技有限公司 | A kind of remote sensing images city superpixel segmentation method |
| CN111028244A (en) * | 2019-12-04 | 2020-04-17 | 电子科技大学 | Remote sensing image semantic segmentation method based on super-pixel under condition of known sample imbalance |
-
2020
- 2020-04-21 CN CN202010316759.6A patent/CN111553222B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289807A (en) * | 2011-07-08 | 2011-12-21 | 西安电子科技大学 | Method for detecting change of remote sensing image based on Treelet transformation and characteristic fusion |
| US20170004472A1 (en) * | 2014-03-07 | 2017-01-05 | Pantry Retail, Inc. | Method for processing transactions at a vending kiosk |
| CN107067405A (en) * | 2017-03-30 | 2017-08-18 | 河海大学 | Based on the preferred Remote Sensing Image Segmentation of yardstick |
| CN107229917A (en) * | 2017-05-31 | 2017-10-03 | 北京师范大学 | A kind of several remote sensing image general character well-marked target detection methods clustered based on iteration |
| CN108647602A (en) * | 2018-04-28 | 2018-10-12 | 北京航空航天大学 | A kind of aerial remote sensing images scene classification method based on image complexity judgement |
| CN109377480A (en) * | 2018-09-27 | 2019-02-22 | 中国电子科技集团公司第五十四研究所 | Arable land use change detection method based on deep learning |
| CN109657610A (en) * | 2018-12-18 | 2019-04-19 | 北京航天泰坦科技股份有限公司 | A kind of land use change survey detection method of high-resolution multi-source Remote Sensing Images |
| CN109712153A (en) * | 2018-12-25 | 2019-05-03 | 杭州世平信息科技有限公司 | A kind of remote sensing images city superpixel segmentation method |
| CN111028244A (en) * | 2019-12-04 | 2020-04-17 | 电子科技大学 | Remote sensing image semantic segmentation method based on super-pixel under condition of known sample imbalance |
Non-Patent Citations (4)
| Title |
|---|
| CSILLIK OVIDIU ET.AL: "Fast Segmentation and Classification of Very High Resolution Remote Sensing Data Using SLIC Superpixels", 《 ADVANCES IN OBJECT-BASED IMAGE ANALYSIS—LINKING WITH COMPUTER VISION AND MACHINE LEARNING》 * |
| MOHAMMAD PASHAEI ET.AL: "Fully Convolutional Neural Network for Land Cover Mapping In A Coastal Wetland with Hyperspatial UAS Imagery", 《IGARSS 2019 - 2019 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM》 * |
| 史蕾 等: "高分辨率遥感影像的自优化迭代分类方法", 《国防科技大学学报》 * |
| 康红宴: "基于SLIC的SAR图像超像素分割方法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114898216A (en) * | 2022-06-14 | 2022-08-12 | 湖南省农林工业勘察设计研究总院 | Remote sensing image deep learning segmentation method and system based on superpixels and watershed |
| CN116797598A (en) * | 2023-08-22 | 2023-09-22 | 山东万牧农业科技有限公司郯城分公司 | Image feature-based cultivation feed quality refinement detection method |
| CN116797598B (en) * | 2023-08-22 | 2023-11-17 | 山东万牧农业科技有限公司郯城分公司 | Image feature-based cultivation feed quality refinement detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111553222B (en) | 2021-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107194872B (en) | Remote sensed image super-resolution reconstruction method based on perception of content deep learning network | |
| CN111325165B (en) | Urban remote sensing image scene classification method considering spatial relationship information | |
| CN107067405B (en) | Remote sensing image segmentation method based on scale optimization | |
| CN111553222B (en) | Remote sensing ground feature classification post-processing method based on iteration superpixel segmentation | |
| CN113313180B (en) | Remote sensing image semantic segmentation method based on deep confrontation learning | |
| CN113888547A (en) | Non-supervision domain self-adaptive remote sensing road semantic segmentation method based on GAN network | |
| CN109961105B (en) | High-resolution remote sensing image classification method based on multitask deep learning | |
| CN109886159B (en) | Face detection method under non-limited condition | |
| CN111369442A (en) | Remote sensing image super-resolution reconstruction method based on fuzzy kernel classification and attention mechanism | |
| CN112819794B (en) | Small celestial body meteorite crater detection method based on deep learning | |
| CN112560719B (en) | High-resolution image water body extraction method based on multi-scale convolution-multi-core pooling | |
| CN114067019A (en) | Urban waterlogging risk map rapid prefabricating method coupling deep learning and numerical simulation | |
| CN115661655B (en) | Southwest mountain area cultivated land extraction method with hyperspectral and hyperspectral image depth feature fusion | |
| CN114037891A (en) | High-resolution remote sensing image building extraction method and device based on U-shaped attention control network | |
| CN117114984A (en) | Remote sensing image super-resolution reconstruction method based on generation countermeasure network | |
| CN111461976A (en) | Image super-resolution method based on efficient lightweight coordinate neural network | |
| CN116823627A (en) | Image complexity evaluation-based oversized image rapid denoising method | |
| CN113920421B (en) | Full convolution neural network model capable of achieving rapid classification | |
| CN118154971A (en) | SAR image classification method based on full convolution-dynamic graph network | |
| CN112907441B (en) | Space downscaling method based on super-resolution of ground water satellite image | |
| Li et al. | FA-GAN: A feature attention GAN with fusion discriminator for non-homogeneous dehazing | |
| CN113515969B (en) | Multi-temporal remote sensing image earth surface coverage classification method based on convolutional neural network | |
| CN112801021A (en) | Method and system for detecting lane line based on multilevel semantic information | |
| CN116310832A (en) | Remote sensing image processing method, device, equipment, medium and product | |
| CN114494284B (en) | Scene analysis model and method based on explicit supervision area relation |
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 |