CN112446859A - Satellite-borne thermal infrared camera image cloud detection method based on deep learning - Google Patents
Satellite-borne thermal infrared camera image cloud detection method based on deep learning Download PDFInfo
- Publication number
- CN112446859A CN112446859A CN202011292436.4A CN202011292436A CN112446859A CN 112446859 A CN112446859 A CN 112446859A CN 202011292436 A CN202011292436 A CN 202011292436A CN 112446859 A CN112446859 A CN 112446859A
- Authority
- CN
- China
- Prior art keywords
- thermal infrared
- dice
- image
- convolution
- model
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 30
- 238000013135 deep learning Methods 0.000 title claims abstract description 8
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000013528 artificial neural network Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000013139 quantization Methods 0.000 claims abstract description 8
- 238000011176 pooling Methods 0.000 claims abstract description 7
- 230000004927 fusion Effects 0.000 claims abstract description 5
- 238000013136 deep learning model Methods 0.000 claims abstract description 4
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 4
- 230000004580 weight loss Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000011800 void material Substances 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 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
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (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)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a satellite-borne thermal infrared camera image cloud detection method based on deep learning, which comprises the following steps: (1) converting an uncalibrated thermal infrared image digital quantization value into an atmospheric top layer brightness temperature value, fusing all thermal infrared bands into a multi-channel image, cutting a remote sensing image, and constructing a data set; (2) the method comprises the steps that a full convolution neural network is combined with a residual error network to serve as a basic structure, cavity convolution is introduced to replace the ordinary convolution and pooling processes, and a new fusion neural network is constructed; (3) before training the model, setting a loss function distributed according to the weight; (4) training the constructed data set by a back propagation algorithm until a new model converges to obtain a deep learning model; (5) individual pixels are determined to be cloud or non-cloud targets using this model. The model has the advantages of light weight, strong generalization capability and suitability for various scenes and multi-channel images.
Description
Technical Field
The invention belongs to the technical field of remote sensing image detection and deep learning, and particularly relates to a satellite-borne thermal infrared camera image cloud detection method based on deep learning.
Background
Estimated from global energy and water cycle experimental (GEWEX) cloud assessment databases, the global annual average cloud coverage is about 68%. When a satellite observes the earth surface, the existence of a cloud layer can block earth surface information, and extreme weather can be predicted by analyzing the movement of the cloud layer in meteorology, so remote sensing image cloud detection is an important step in various remote sensing applications. More importantly, when the thermal infrared band image is used for observation all day long, the satellite-ground transmission resources are occupied by excessive cloud layer information, so that the satellite-carried infrared camera image on-orbit cloud detection becomes a trend in the field of remote sensing.
The current cloud detection method comprises a traditional detection method and an intelligent detection method. However, the traditional cloud detection method is difficult to distinguish cloud layers with complex image backgrounds from the earth surface, and depends too much on human experience, so that many missed detections and false detections exist. In order to solve the above problem, an intelligent detection method is proposed for cloud detection. The deep learning network can eliminate the step of manually setting the characteristic parameters in the traditional cloud detection method, so that the artificial subjective influence is weakened, and the robustness of the detection algorithm is improved. The full convolution neural network and the U-type network evolve cloud detection into a classification problem at the pixel level, and end-to-end training is performed by using a remote sensing image through an encoder-decoder structure. However, the full convolution neural network has a network degradation phenomenon, and the pooling operation reduces the resolution, resulting in a reduction in the detection accuracy. The number of floating point operations required by the U-type network is high, the requirement on the computing capacity of hardware is high, and the on-orbit cloud detection is not facilitated.
In summary, according to the requirements of all-day, accurate, real-time and on-orbit cloud detection, there is an urgent need to develop a method for cloud detection of images of satellite-borne heat infrared cameras.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a satellite-borne thermal infrared camera image cloud detection method based on deep learning, and solves the problem of satellite-borne thermal infrared camera image cloud detection.
The technical problem to be solved by the invention is realized by adopting the following technical scheme:
1) converting an uncalibrated thermal infrared image digital quantization value into an atmospheric top layer brightness temperature value, fusing all thermal infrared bands into a multi-channel image, cutting a remote sensing image, and constructing a data set, wherein the method comprises the following specific steps:
1-1) converting the digital quantized value of the uncalibrated thermal infrared image into the value of the brightness temperature of the top layer of the atmosphere with physical significance, wherein the conversion formula is expressed by the following equation:
TOA_RA＝ML·DN+AL (1)
wherein TOA _ RA is the atmospheric top level radiance value, DN is the digital quantization value, MLMultiplication rescaling factor for the thermal infrared band, ALIs the addition and readjustment factor of the thermal infrared band, TOA _ BT is the value of the top brightness temperature of the atmosphere, K1And K2Thermal conversion constants for the thermal infrared band;
1-2) fusing all thermal infrared bands into a multi-channel image by using a GDAL library in software;
1-3) cutting the remote sensing image into small blocks of images n.n, wherein n is between 200 and 600;
1-4) when the data volume is less, the proportion of the verification set and the test set is consistent to be 40-20%, and the proportion of the training set is 60-80%;
2) the method is characterized in that a full convolution neural network is combined with a residual error network to serve as a basic structure, a cavity convolution is introduced to replace the common convolution and pooling process, a new fusion neural network is constructed, and a residual error module and a cavity convolution formula are expressed by the following equation:
y＝F(x)+x (3)
wherein x and y represent the input and output of the residual module, and f (x) is the residual function to be learned;
(F*lk)(s+lt)＝∑s+ltF(s)k(t) (4)
wherein F is a discrete functionlFor discrete convolution, k is a discrete filter, s is a step size, l is a hole factor, t is an argument, F(s) is a discrete convolution with a step size s, k (t) is an argument of tA dispersive filter;
3) before training the model, setting a loss function distributed according to weight, wherein the loss function L is the sum of weight distribution of Dice loss suitable for extreme unevenness of the sample and binary cross entropy loss suitable for class equalization of the sample, and is represented by the following equation:
L＝WBceLBce+WDiceLDice (5)
LBce＝-∑iTiln(Pi)+(1-Ti)ln(1-Pi) (6)
WBce+WDice＝1 (8)
wherein L is the loss function of the method, LBceIs a two-class cross entropy loss, LDiceIs the loss of Dice, P is the network prediction result, T is the true value, i is the independent variable, WDiceThe weight loss of the Dice is in the range of 0 to 0.2, WBceThe cross entropy loss weight for the binary class is in the range of 1 to 0.8;
4) training by using the constructed data set through a back propagation algorithm until a new model converges to obtain a deep learning model FCN _ ResNet _ k;
5) individual pixels are determined to be cloud or non-cloud targets using this model.
The invention has the advantages and positive effects that:
1. the invention takes the full convolution neural network and the residual error network as the basic structure, and effectively avoids the network degradation phenomenon caused by the deeper network. The cavity convolution is introduced to replace the common convolution and pooling processes, so that the influence of the reduction of the resolution due to the pooling operation in the original model is overcome, and the receptive field index is increased. And constructing a new converged neural network FCN _ ResNet _ k for cloud detection.
2. The invention has reasonable design and less floating point operation times required by the network structure, and can improve the reasoning speed and the network efficiency. The method has the advantages of light weight, strong generalization capability, suitability for various scenes and multi-channel images and the like, and can be applied and popularized in the field of satellite-borne heat infrared camera image cloud detection.
Drawings
Fig. 1 is a diagram illustrating an exemplary remote sensing image detection process according to an embodiment of the present invention;
fig. 2 is an exemplary diagram of a remote sensing image preprocessing process according to an embodiment of the present invention, where (a) and (b) are original images of bands of 10um to 11.19um and 11.5um to 12.51um, respectively, and (c) and (d) are images obtained by converting digital quantization values of corresponding bands into atmospheric top luminance temperature values, respectively, and (e) is a two-channel image obtained by fusing two thermal infrared bands.
Detailed Description
The following takes thermal infrared band cloud detection as an example, and the following detailed description is made on the embodiments of the present invention with reference to the accompanying drawings:
the invention realizes the high-precision infrared band cloud detection method by constructing an improved FCN model and introducing a voting principle to vote by using three different models. The method mainly comprises the following steps:
1) converting an uncalibrated thermal infrared image digital quantization value into an atmospheric top layer brightness temperature value, fusing all thermal infrared bands into a multi-channel image, cutting a remote sensing image, and constructing a data set, wherein the method comprises the following specific steps:
1-1) converting the digital quantized value of the uncalibrated thermal infrared image into the value of the brightness temperature of the top layer of the atmosphere with physical significance, wherein the conversion formula is expressed by the following equation:
TOA_RA＝ML·DN+AL (1)
wherein TOA _ RA is the atmospheric top level radiance value, DN is the digital quantization value, ML＝3.342×10-4Multiplication rescaling factor for the thermal infrared band, AL0.1 is the thermal infrared band addition and readjustment factor, TOA _ BT is the atmospheric top luminance temperature value, K1And K2In the thermal infrared bandThermal conversion constant of (d); selecting 774.89 for K1 with a wave band of 10um-11.19um, 1321.08 for K2, 480.89 for K1 with a wave band of 11.5um-12.51um and 1201.14 for K2;
1-2) fusing two thermal infrared bands of 10um-11.19um and 11.5um-12.51um into a two-channel image by using a GDAL library in software, and converting a 16-bit remote sensing image into an 8-bit remote sensing image;
1-3) cutting 96 two-channel images into small block images n.n, wherein n is 224, and 38400 small block images 224 are obtained after cutting;
1-4) the data volume is less, a training set is set: and (4) verification set: the ratio of test sets was 8: 1: 1, the data set comprises 32000 training sets, 3200 verification sets and 3200 test sets;
wherein F is a discrete functionlFor discrete convolution, k is a discrete filter, s is a step size, l is a hole factor, t is an argument, F(s) is discrete convolution with a step size s, and k (t) is a discrete filter with an argument t;
2) the method is characterized in that a full convolution neural network is combined with a residual error network to serve as a basic structure, void convolution is introduced to replace the common convolution and pooling process, a new model is constructed, and a residual error module and a void convolution formula are expressed by the following equation:
y＝F(x)+x (3)
wherein x and y represent the input and output of the residual module, and f (x) is the residual function to be learned;
(F*lk)(s+lt)＝∑s+ltF(s)k(t) (4)
wherein F is a discrete functionlFor discrete convolution, k is a discrete filter, s is a step size, l is a hole factor, t is an argument, F(s) is discrete convolution with a step size s, and k (t) is a discrete filter with an argument t;
3) before training the model, setting a loss function distributed according to weight, wherein the loss function L is the sum of weight distribution of Dice loss suitable for extreme unevenness of the sample and binary cross entropy loss suitable for class equalization of the sample, and is represented by the following equation:
L＝WBceLBce+WDiceLDice (5)
LBce＝-∑iTiln(Pi)+(1-Ti)ln(1-Pi) (6)
WBce+WDice＝1 (8)
wherein P is the network prediction result, T is the truth value, and LDiceFor loss of Dice, LBceIs a binary cross entropy loss, i is an independent variable, WBceTake 0.2, W for the Dice loss weightDiceTaking 0.8 as the weight range of the binary cross entropy loss, wherein L is the loss function of the method;
4) training by using the constructed data set through a back propagation algorithm until a new model converges to obtain a deep learning model FCN _ ResNet _ k;
5) individual pixels are determined to be cloud or non-cloud targets using this model.
6) And calculating the floating point operation times of the model and measuring the calculation complexity of the model. The number of floating point operations of the new fusion neural network is 7445043712, the number of floating point operations of the U-type network is 85683945472, and the calculation complexity of the U-type network is about 11.51 times that of the new fusion neural network.
Claims (1)
1. A satellite-borne thermal infrared camera image cloud detection method based on deep learning is characterized by comprising the following steps:
1) converting an uncalibrated thermal infrared image digital quantization value into an atmospheric top layer brightness temperature value, fusing all thermal infrared bands into a multi-channel image, cutting a remote sensing image, and constructing a data set, wherein the method comprises the following specific steps:
1-1) converting the digital quantized value of the uncalibrated thermal infrared image into the value of the brightness temperature of the top layer of the atmosphere with physical significance, wherein the conversion formula is expressed by the following equation:
TOA_RA＝ML·DN+AL (1)
wherein TOA _ RA is the atmospheric top level radiance value, DN is the digital quantization value, MLMultiplication rescaling factor for the thermal infrared band, ALIs the addition and readjustment factor of the thermal infrared band, TOA _ BT is the value of the top brightness temperature of the atmosphere, K1And K2Thermal conversion constants for the thermal infrared band;
1-2) fusing all thermal infrared bands into a multi-channel image by using a GDAL library in software;
1-3) cutting the remote sensing image into small blocks of images n.n, wherein n is between 200 and 600;
1-4) when the data volume is less, the proportion of the verification set and the test set is consistent to be 40-20%, and the proportion of the training set is 60-80%;
2) the method is characterized in that a full convolution neural network is combined with a residual error network to serve as a basic structure, a cavity convolution is introduced to replace the common convolution and pooling process, a new fusion neural network is constructed, and a residual error module and a cavity convolution formula are expressed by the following equation:
y＝F(x)+x (3)
wherein x and y represent the input and output of the residual module, and f (x) is the residual function to be learned;
(F*lk)(s+lt)＝∑s+ltF(s)k(t) (4)
wherein F is a discrete functionlFor discrete convolution, k is a discrete filter, s is a step size, l is a hole factor, t is an argument, F(s) is discrete convolution with a step size s, and k (t) is a discrete filter with an argument t;
3) before training the model, setting a loss function distributed according to weight, wherein the loss function L is the sum of weight distribution of Dice loss suitable for extreme unevenness of the sample and binary cross entropy loss suitable for class equalization of the sample, and is represented by the following equation:
L＝WBceLBce+WDiceLDice (5)
LBce＝-∑iTiln(Pi)+(1-Ti)ln(1-Pi) (6)
WBce+WDice＝1 (8)
wherein P is the network prediction result, T is the truth value, and LDiceFor loss of Dice, LBceIs a binary cross entropy loss, i is an independent variable, WBceThe weight loss of the Dice is in the range of 0 to 0.2, WDiceThe weight loss range of the binary cross entropy is 1-0.8, and L is the loss function of the method;
4) training the constructed data set by a back propagation algorithm until a new model converges to obtain a deep learning model;
5) individual pixels are determined to be cloud or non-cloud targets using this model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011292436.4A CN112446859A (en) | 2020-11-18 | 2020-11-18 | Satellite-borne thermal infrared camera image cloud detection method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011292436.4A CN112446859A (en) | 2020-11-18 | 2020-11-18 | Satellite-borne thermal infrared camera image cloud detection method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112446859A true CN112446859A (en) | 2021-03-05 |
Family
ID=74737473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011292436.4A Pending CN112446859A (en) | 2020-11-18 | 2020-11-18 | Satellite-borne thermal infrared camera image cloud detection method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112446859A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115564778A (en) * | 2022-12-06 | 2023-01-03 | 深圳思谋信息科技有限公司 | Defect detection method and device, electronic equipment and computer readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200058126A1 (en) * | 2018-08-17 | 2020-02-20 | 12 Sigma Technologies | Image segmentation and object detection using fully convolutional neural network |
| CN111274865A (en) * | 2019-12-14 | 2020-06-12 | 深圳先进技术研究院 | Remote sensing image cloud detection method and device based on full convolution neural network |
-
2020
- 2020-11-18 CN CN202011292436.4A patent/CN112446859A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200058126A1 (en) * | 2018-08-17 | 2020-02-20 | 12 Sigma Technologies | Image segmentation and object detection using fully convolutional neural network |
| CN111274865A (en) * | 2019-12-14 | 2020-06-12 | 深圳先进技术研究院 | Remote sensing image cloud detection method and device based on full convolution neural network |
Non-Patent Citations (3)
| Title |
|---|
| 张丹丹: "《基于深度神经网络技术的高分遥感图像处理及应用》", 31 August 2020, 中国宇航出版社 * |
| 张家强 等: "基于深度残差全卷积网络的Landsat 8遥感影像云检测方法", 《激光与光电子学进展》 * |
| 赵志刚: "《区域土地资源研究与农业规划实例 以宜春市袁州区为例》", 31 October 2017, 科学技术文献出版社 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115564778A (en) * | 2022-12-06 | 2023-01-03 | 深圳思谋信息科技有限公司 | Defect detection method and device, electronic equipment and computer readable storage medium |
| CN115564778B (en) * | 2022-12-06 | 2023-03-14 | 深圳思谋信息科技有限公司 | Defect detection method and device, electronic equipment and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112507793B (en) | Ultra-short term photovoltaic power prediction method | |
| CN108960333B (en) | Hyperspectral image lossless compression method based on deep learning | |
| CN112634292A (en) | Asphalt pavement crack image segmentation method based on deep convolutional neural network | |
| CN110210378B (en) | Embedded video image analysis method and device based on edge calculation | |
| CN111582555B (en) | Photovoltaic power prediction method based on foundation cloud picture image characteristics | |
| CN110826693A (en) | Three-dimensional atmospheric temperature profile inversion method and system based on DenseNet convolutional neural network | |
| CN113239830A (en) | Remote sensing image cloud detection method based on full-scale feature fusion | |
| CN112883887B (en) | Building instance automatic extraction method based on high spatial resolution optical remote sensing image | |
| CN112232543A (en) | Multi-site prediction method based on graph convolution network | |
| CN114444757A (en) | Combined prediction method for plateau mountain multi-model multi-scale new energy power station output | |
| CN114120036A (en) | Lightweight remote sensing image cloud detection method | |
| CN115935809A (en) | Photovoltaic power generation power interval prediction method based on cloud cover characteristic of foundation cloud cover | |
| CN115578280A (en) | Construction method of double-branch remote sensing image defogging network | |
| CN116205962A (en) | Monocular depth estimation method and system based on complete context information | |
| CN115393690A (en) | Light neural network air-to-ground observation multi-target identification method | |
| Zhou et al. | MSAR‐DefogNet: Lightweight cloud removal network for high resolution remote sensing images based on multi scale convolution | |
| CN112446859A (en) | Satellite-borne thermal infrared camera image cloud detection method based on deep learning | |
| CN113705340B (en) | Deep learning change detection method based on radar remote sensing data | |
| CN114373104A (en) | Three-dimensional point cloud semantic segmentation method and system based on dynamic aggregation | |
| CN114511787A (en) | Neural network-based remote sensing image ground feature information generation method and system | |
| CN117197632A (en) | Transformer-based electron microscope pollen image target detection method | |
| CN116720156A (en) | Weather element forecasting method based on graph neural network multi-mode weather data fusion | |
| CN116523875A (en) | Insulator defect detection method based on FPGA pretreatment and improved YOLOv5 | |
| CN115561834A (en) | Meteorological short-term and temporary forecasting all-in-one machine based on artificial intelligence | |
| CN113034445B (en) | Multi-scale connection image defogging algorithm based on UNet3+ |
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 |
Application publication date: 20210305 |
|
| WD01 | Invention patent application deemed withdrawn after publication |