CN109472747A - A deep learning method for decoherent speckle noise in microwave remote sensing images - Google Patents
A deep learning method for decoherent speckle noise in microwave remote sensing images Download PDFInfo
- Publication number
- CN109472747A CN109472747A CN201811213563.3A CN201811213563A CN109472747A CN 109472747 A CN109472747 A CN 109472747A CN 201811213563 A CN201811213563 A CN 201811213563A CN 109472747 A CN109472747 A CN 109472747A
- Authority
- CN
- China
- Prior art keywords
- network
- dscnn
- remote sensing
- speckle noise
- microwave remote
- 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 43
- 238000013135 deep learning Methods 0.000 title claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 16
- 230000001427 coherent effect Effects 0.000 claims abstract description 15
- 230000009467 reduction Effects 0.000 claims description 32
- 238000012360 testing method Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 6
- 230000008901 benefit Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000004088 simulation Methods 0.000 claims description 2
- 238000001914 filtration Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000000052 comparative effect Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000031068 symbiosis, encompassing mutualism through parasitism Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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]
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
本发明提出了一种微波遥感图像去相干斑噪声的深度学习方法；本发明提出的DsCNN网络采用残差网络模块串联方式构建，相干斑噪声分级处理降低了相干斑噪声去除难度，克服了DsCNN网络过深导致训练难度增加的问题；本发明提出的DsCNN网络样本数据集源自开源软件，可以免费获取充足样本数据集，克服现有微波遥感开源图像数据集少，获取成本高的缺点；本发明提出的DsCNN网络在去除相干斑噪声同时，能够保持目标纹理特征；本发明提出的DsCNN网络基于样本驱动，降低了相干斑噪声建模难度，提高了相干斑噪声建模精度；本发明提出的DsCNN网络处理速度快，满足实时处理大尺度微波遥感图像去相干斑噪声任务的需求。
The present invention proposes a deep learning method for removing coherent speckle noise from microwave remote sensing images; the DsCNN network proposed by the present invention is constructed by using residual network modules in series. Excessive depth leads to the problem of increased training difficulty; the DsCNN network sample data set proposed by the present invention is derived from open source software, and sufficient sample data sets can be obtained free of charge, overcoming the shortcomings of existing microwave remote sensing open source image data sets and high acquisition costs; the present invention The proposed DsCNN network can maintain the target texture feature while removing speckle noise; the DsCNN network proposed by the present invention is driven by samples, which reduces the difficulty of speckle noise modeling and improves the modeling accuracy of speckle noise; the DsCNN proposed by the present invention The network processing speed is fast, which meets the needs of real-time processing of large-scale microwave remote sensing image decoherence speckle noise tasks.
Description
Technical field
The present invention relates to microwave remote sensing technical field of image processing, and in particular to a kind of micro- using deep learning network implementations
The method of wave remote sensing images speckle noise reduction.
Background technique
Microwave remote sensing technique has many advantages, such as round-the-clock, round-the-clock, multi-angle of view, multi-frequency, multipolarization, and electromagnetism
Wave signal has stronger penetration power, it can be found that hiding target with tracking, thus is widely used in various military and civilians
Scene.Compared with natural image, microwave remote sensing image is often accompanied by a large amount of coherent speckle noises, significantly reduces microwave remote sensing
The resolution ratio of image causes certain difficulty to further realize the application such as edge detection, image segmentation, target identification.Microwave
Remote sensing images, especially SAR (Synthetic Aperture Radar) image, coherent speckle noise are typical multiplicative noise,
It generates along with SAR image and is existed.Compared with additive noise, multiplicative noise is more difficult to handle.
In the past few decades, researcher proposes a large amount of microwave remote sensing image speckle noise reduction methods.Often
The method of microwave remote sensing image speckle noise reduction includes: Lee filtering method, Frost filtering method, the filtering side Kuan
Method, Gamma MAP filtering method, FANS filtering method, BM-3D filtering method etc..These filtering methods pass through to sliding window
Interior microwave remote sensing image data carries out local shape factor, inhibits coherent speckle noise, improves microwave remote sensing picture quality.But
These methods face two big challenges: whether the parameter (1) counted in sliding window is reliable；(2) whether coherent speckle noise model
Reliably.And these methods successively extract characteristics of image using sliding window, reduce image processing speed, are not possible to meet real-time
Handle large scale microwave remote sensing image speckle noise reduction task.
The forming process of microwave remote sensing image coherent speckle noise is extremely complex, it is difficult to carry out essence to it with simple mathematical model
Really modeling, thus speckle noise reduction method treatment effect is undesirable at present.The deep learning method of sample driving then has
Powerful non-linear expression's ability, image processor GPU (Graphics Processing Unit) accelerate deep learning side
Method processing speed provides new approaches to explore microwave remote sensing image speckle noise reduction.
Summary of the invention
In order to solve above-mentioned key technology difficulty, the invention proposes a kind of depths of microwave remote sensing image speckle noise reduction
Spend learning method；The present invention passes through building DsCNN (Despeckle using Convolutional Neural Networks)
Network realizes the Accurate Model to multiplicative noise, and then realizes processing large scale microwave remote sensing image speckle noise reduction in real time
Task.
The deep learning method of microwave remote sensing image speckle noise reduction of the present invention the following steps are included:
1) DsCNN network is constructed；
2) DsCNN network training and test sample data set are generated；
3) training DsCNN network；
4) DsCNN network implementations microwave remote sensing image speckle noise reduction is utilized；
DsCNN network uses residual error network ResNet (Residual Network) block coupled in series form structure in step 1)
It builds, realizes the classification processing of coherent speckle noise；
DsCNN network activation function is all made of Tanh in step 1), and microwave remote sensing data value field is compressed to [0,1],
Accelerate DsCNN network convergence；
DsCNN network convolutional layer fill pattern uses " same " in step 1).
DsCNN network does not contain pond layer, full articulamentum in step 1)；
DsCNN network uses the end-to-end processing of pixel to pixel formula in step 1)；
Loss function is using mean square error MSE (Mean Squared Error) and total variation TV (Total in step 1)
Variation the form) combined, as shown in formula (1):
Ltotal=Lmse+ρLTV (1)
Wherein ρ indicates weight factor, is used for balanced mean square error loss function LmseWith total variation loss function LTVBetween
Significance level；Wherein LmseAnd LTVIt can be expressed as formula 2 and formula 3:
Wherein W indicates that image lateral coordinates element number, H indicate image longitudinal coordinate element number；W indicates that image is lateral
W-th of element, h indicate longitudinal h-th of the element of image；YW, hAnd XW, hRespectively indicate outputting and inputting for DsCNN network；Table
Show the denoising image of DsCNN network output；Indicate the DsCNN network of Complete Convergence；LmseImage is made an uproar with nothing after guaranteeing denoising
Error is minimum between acoustic image；LTVImage after guaranteeing denoising has better texture features；
Training sample is originated from the colour optics remote sensing images over the ground of Google Earth in step 2), by by colourama
It learns remote sensing images and is converted to the gray scale remote sensing images with multiplicative noise, realize the realistic simulation to true microwave remote sensing image, it is low
Cost generates a large amount of DsCNN network sample data sets.
Advantages of the present invention:
The invention proposes a kind of deep learning methods of microwave remote sensing image speckle noise reduction；The present invention has as follows
Advantage:
1, DsCNN network is constructed using residual error network module series system, and coherent speckle noise classification processing can reduce phase
Dry spot noise remove difficulty, and can overcome the problems, such as to train difficulty to increase caused by the DsCNN network number of plies is too deep；
2, DsCNN network sample data set is originated from open source Google Earth software, can be with Free Acquisition adequate sample number
According to collection, overcome existing microwave remote sensing image open source data set few, the high disadvantage of procurement cost；
3, DsCNN network is in removal coherent speckle noise meanwhile, it is capable to keep target texture feature；
4, DsCNN network processes speed is fast, meets processing large scale microwave remote sensing image speckle noise reduction task in real time
Demand；
5, DsCNN network is driven based on sample, is reduced coherent speckle noise modeling difficulty, is improved coherent speckle noise modeling
Precision；
Detailed description of the invention
Fig. 1 is DsCNN schematic network structure；
Fig. 2 is DsCNN network sample data set product process schematic diagram；
Fig. 3 is the speckle noise reduction result schematic diagram that DsCNN network is directed to training sample data collection；
Fig. 4 is the speckle noise reduction result schematic diagram that DsCNN network is directed to test sample data set；
Fig. 5 is that different speckle noise reduction methods are directed to identical SAR image speckle noise reduction comparative result figure；
Fig. 6 is the speckle noise reduction Comparative result that different speckle noise reduction methods are directed to three groups of zonule SAR images
Figure.
Specific embodiment
With reference to the accompanying drawing, by specific embodiment, the present invention is further explained.
The deep learning method of the present embodiment microwave remote sensing image speckle noise reduction the following steps are included:
1) DsCNN network is constructed:
As shown in Figure 1, DsCNN network is made of two residual error network modules；Each residual error network module includes four layers of volume
Product neural net layer and quick articulamentum (shortcut) are constituted；Convolutional neural networks layer includes matrix convolution Conv, batch standard
Change BN and activation primitive Tanh to operate；Shortcut layers of realization identical transformation are added by matrix and are realized；Input picture is 256
The image containing coherent speckle noise of × 256 pixels；Export the noise-free picture that image is corresponding 256 × 256 pixel；L1-L7
The convolution kernel size of layer convolutional layer is 3 × 3, and convolution kernel number is 64；L8 layers of convolution kernel size is 3 × 3, convolution kernel number
It is 1.
2) DsCNN network training and test sample data set are generated:
As shown in Fig. 2, the crawl of colour optics remote sensing images is in open source software Google Earth；Gray processing is carried out to image
Processing generates greyscale optical remote sensing images；Multiplicative noise is added on greyscale optical remote sensing images, then corresponding contain can be generated
There is the emulation microwave remote sensing image of multiplicative noise, multiplicative noise variance is set as 0.2 in this example；It will be imitative containing multiplicative noise
True microwave remote sensing image random division is the image of 256 × 256 pixels, then it is grey with the noiseless of corresponding 256 × 256 pixel
It spends remote sensing image and constitutes one group of sample data；The symbiosis of this example is at 3400 groups of sample data sets, including 3000 groups of instructions
Practice sample data set, 200 groups of verifying sample data sets and 200 groups of test sample data sets.
3) training DsCNN network:
DsCNN network uses adaptability moments estimation optimization method in this example；DsCNN network training every batch of inputs 16 groups
Sample data；DsCNN network training amounts to iteration 100 times；DsCNN e-learning rate Initialize installation is 0.0002, decaying
The factor is set as 0.1；Weight factor ρ is set as 0.000002 in loss function；Fig. 3 is to export after DsCNN network completes training
Training sample data collection part denoising result；Fig. 4 is that the part of the test sample data set of DsCNN network output denoises knot
Fruit；A indicates that network inputs image, b indicate noise-free picture in Fig. 3 and Fig. 4；C indicates that network exports image, that is, schemes after denoising
Picture.As can be seen from figs. 3 and 4 the coherent speckle noise of DsCNN network output image is had compared with DsCNN network inputs image
Effect inhibits, and target texture is clear, and picture quality is significantly promoted；
Introduce Y-PSNR PSNR (Peak Signal to Noise Ratio) and structuring index of similarity
(structural similarity index, be abbreviated as SSIM) Lai Hengliang picture quality.Wherein structuring index of similarity can
To indicate are as follows:
Wherein a and b respectively indicates image and noise-free picture to be assessed.For image a,Wherein P
Indicate pictorial element number, p indicates p-th of element.Above-mentioned definition is for image b
It is applicable in.C1And C2It is a small constant, such as 0.01.SSIM value is bigger, table
Show that the similitude of two images is bigger.SSIM maximum value is 1, is indicated identical.
As shown in Table 1, DsCNN network training and test sample collection output image are better than input picture, and PSNR and SSIM refer to
Number is greatly improved.
PSNR the and SSIM index of table 1:DsCNN network training and test sample collection output image
4) DsCNN network implementations microwave remote sensing image speckle noise reduction is utilized
As shown in Fig. 5 (a), two width SAR images are originated from the shooting of TerraSAR-X and TanDEM-X spaceborne radar not
Island image near Singapore in the same time.Two width SAR images are 9984 × 5888 pixels.Introduce equivalent more view indexes
(ENL:Equivalent Number of Looks) measures SAR image quality, as shown in formula 5:
Wherein K indicates the total number of sliding window, μkWithRespectively indicate the mean value and variance of pixel in k-th of sliding window.
ENL index is bigger, then shows that picture quality is higher.
B indicates FANS method speckle noise reduction result in Fig. 5 and Fig. 6；C indicates BM-3D method speckle noise reduction knot
Fruit；D indicates DsCNN network speckle noise reduction result；SAR image compares figure b in figure d and figure c image texture details is more clear
Clear, coherent speckle noise is more preferably inhibited；As shown in Table 2, the ENL index for scheming d is maximum, i.e., picture quality is best.It can by table 3
Know, handle the SAR image of same scale, FANS method needs 1400 seconds, and BM-3D method needs 600 seconds, and DsCNN method is only
Need 10 seconds or so；Therefore DsCNN method meets handles large scale microwave remote sensing image speckle noise reduction task in real time
Demand.
SAR image ENL index contrast table in table 2: Fig. 6
Table 3: SAR image distinct methods operation time contrast table in reconstruct image 5
It is finally noted that the purpose for publicizing and implementing example is to help to further understand the present invention, but this field
Technical staff be understood that without departing from the spirit and scope of the invention and the appended claims, it is various replacement and repair
It is all possible for changing.Therefore, the present invention should not be limited to embodiment disclosure of that, and the scope of protection of present invention is to weigh
Subject to the range that sharp claim defines.
Claims (6)
1. it is characterized by comprising following steps for a kind of deep learning method of microwave remote sensing image speckle noise reduction:
1) DsCNN network is constructed；
2) DsCNN network training and test sample data set are generated；
3) training DsCNN network；
4) DsCNN network implementations microwave remote sensing image speckle noise reduction is utilized.
2. the deep learning method of microwave remote sensing image speckle noise reduction according to claim 1 it is characterized in that,
DsCNN network realizes coherent speckle noise classification processing using the series connection building of residual error network module.
3. the deep learning method of microwave remote sensing image speckle noise reduction according to claim 1 it is characterized in that,
DsCNN network activation function is all made of Tanh, and microwave remote sensing value data is compressed to [0,1], accelerates DsCNN network convergence.
4. the deep learning method of microwave remote sensing image speckle noise reduction according to claim 1 it is characterized in that,
DsCNN network does not contain pond layer, and convolutional layer fill pattern uses " same ".
5. the deep learning method of microwave remote sensing image speckle noise reduction according to claim 1 it is characterized in that,
DsCNN network uses the end-to-end training of pixel to pixel formula.
6. the deep learning method of microwave remote sensing image speckle noise reduction according to claim 1 is it is characterized in that, benefit
Generate simulation sample data set with open source software: colour optics remote sensing images are grabbed in open source software, to colour optics remote sensing figure
As carrying out gray processing processing, and multiplicative noise is added, generates the emulation microwave remote sensing image containing multiplicative noise.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811213563.3A CN109472747A (en) | 2018-10-18 | 2018-10-18 | A deep learning method for decoherent speckle noise in microwave remote sensing images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811213563.3A CN109472747A (en) | 2018-10-18 | 2018-10-18 | A deep learning method for decoherent speckle noise in microwave remote sensing images |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109472747A true CN109472747A (en) | 2019-03-15 |
Family
ID=65664162
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811213563.3A Pending CN109472747A (en) | 2018-10-18 | 2018-10-18 | A deep learning method for decoherent speckle noise in microwave remote sensing images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109472747A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111266A (en) * | 2019-04-08 | 2019-08-09 | 西安交通大学 | A kind of approximate information pass-algorithm improved method based on deep learning denoising |
| CN111932467A (en) * | 2020-07-13 | 2020-11-13 | 东软医疗系统股份有限公司 | Image processing method and device |
| CN112926448A (en) * | 2021-02-24 | 2021-06-08 | 重庆交通大学 | SAR image classification method with stable fluctuation of speckle pattern |
| CN116228609A (en) * | 2023-05-10 | 2023-06-06 | 中国人民解放军国防科技大学 | Radar image speckle filtering method and device based on zero sample learning |
| CN116430347A (en) * | 2023-06-13 | 2023-07-14 | 成都实时技术股份有限公司 | Radar data acquisition and storage method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846253A (en) * | 2017-02-14 | 2017-06-13 | 深圳市唯特视科技有限公司 | A kind of image super-resolution rebuilding method based on reverse transmittance nerve network |
| CN107644401A (en) * | 2017-08-11 | 2018-01-30 | 西安电子科技大学 | Multiplicative noise minimizing technology based on deep neural network |
| CN108629746A (en) * | 2018-04-24 | 2018-10-09 | 华中科技大学 | A kind of radar map image patch based on associated loss convolutional neural networks is made an uproar suppressing method |
-
2018
- 2018-10-18 CN CN201811213563.3A patent/CN109472747A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846253A (en) * | 2017-02-14 | 2017-06-13 | 深圳市唯特视科技有限公司 | A kind of image super-resolution rebuilding method based on reverse transmittance nerve network |
| CN107644401A (en) * | 2017-08-11 | 2018-01-30 | 西安电子科技大学 | Multiplicative noise minimizing technology based on deep neural network |
| CN108629746A (en) * | 2018-04-24 | 2018-10-09 | 华中科技大学 | A kind of radar map image patch based on associated loss convolutional neural networks is made an uproar suppressing method |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111266A (en) * | 2019-04-08 | 2019-08-09 | 西安交通大学 | A kind of approximate information pass-algorithm improved method based on deep learning denoising |
| CN110111266B (en) * | 2019-04-08 | 2021-07-13 | 西安交通大学 | An Improved Approximate Information Transfer Algorithm Based on Deep Learning Denoising |
| CN111932467A (en) * | 2020-07-13 | 2020-11-13 | 东软医疗系统股份有限公司 | Image processing method and device |
| CN112926448A (en) * | 2021-02-24 | 2021-06-08 | 重庆交通大学 | SAR image classification method with stable fluctuation of speckle pattern |
| CN112926448B (en) * | 2021-02-24 | 2022-06-14 | 重庆交通大学 | SAR image classification method with stable fluctuation of speckle pattern |
| CN116228609A (en) * | 2023-05-10 | 2023-06-06 | 中国人民解放军国防科技大学 | Radar image speckle filtering method and device based on zero sample learning |
| CN116430347A (en) * | 2023-06-13 | 2023-07-14 | 成都实时技术股份有限公司 | Radar data acquisition and storage method |
| CN116430347B (en) * | 2023-06-13 | 2023-08-22 | 成都实时技术股份有限公司 | Radar data acquisition and storage method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109472747A (en) | A deep learning method for decoherent speckle noise in microwave remote sensing images | |
| Zhang et al. | Hybrid noise removal in hyperspectral imagery with a spatial–spectral gradient network | |
| CN110120020A (en) | A kind of SAR image denoising method based on multiple dimensioned empty residual error attention network | |
| CN107525588B (en) | Rapid reconstruction method of dual-camera spectral imaging system based on GPU | |
| CN111062880A (en) | Underwater image real-time enhancement method based on condition generation countermeasure network | |
| CN103873743A (en) | Video de-noising method based on structure tensor and Kalman filtering | |
| CN107909109A (en) | SAR image sorting technique based on conspicuousness and multiple dimensioned depth network model | |
| CN109658351A (en) | The high spectrum image denoising method that a kind of combination L0 gradient constraint and local low-rank matrix are restored | |
| CN109410149A (en) | A kind of CNN denoising method extracted based on Concurrent Feature | |
| CN108447113B (en) | A pulsed intensity-correlated 3D imaging method based on a deep learning model | |
| CN103295241A (en) | Frequency domain significance target detection method based on Gabor wavelet | |
| CN102509263A (en) | K-SVD (K-means singular value decomposition) speckle inhibiting method based on SAR (synthetic aperture radar) image local statistic characteristic | |
| Padthe et al. | Plug-and-Play with POA based Maximum a Posteriori Denoisers for Image | |
| CN118195901A (en) | A blind super-resolution method and device for sonar images integrating multi-dimensional self-attention | |
| Wen et al. | Learning flipping and rotation invariant sparsifying transforms | |
| CN106023109A (en) | Region similar example learning-based sparse denoising method | |
| Oral et al. | Plug-and-play regularization on magnitude with deep priors for 3D near-field MIMO imaging | |
| Niu et al. | Underwater Waste Recognition and Localization Based on Improved YOLOv5. | |
| CN111539434B (en) | Similarity-based infrared weak and small target detection method | |
| Gu et al. | A gan-based method for sar image despeckling | |
| Chen et al. | Image haze removal by adaptive CycleGAN | |
| CN109948571B (en) | Optical remote sensing image ship detection method | |
| CN107341798B (en) | High Resolution SAR image change detection method based on the overall situation-part SPP Net | |
| CN113435366B (en) | A Bayesian unmixing method for multi-temporal hyperspectral images in wavelet domain | |
| Rouhbakhshmeghrazi et al. | Super-Resolution Reconstruction of UAV Images with GANs: Achievements and Challenges |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190315 |