CN114998242A - Method for detecting power transmission line pole tower in satellite image - Google Patents
Method for detecting power transmission line pole tower in satellite image Download PDFInfo
- Publication number
- CN114998242A CN114998242A CN202210588429.1A CN202210588429A CN114998242A CN 114998242 A CN114998242 A CN 114998242A CN 202210588429 A CN202210588429 A CN 202210588429A CN 114998242 A CN114998242 A CN 114998242A
- Authority
- CN
- China
- Prior art keywords
- tower
- transmission line
- power transmission
- image
- target
- 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
- 230000005540 biological transmission Effects 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000001514 detection method Methods 0.000 claims abstract description 45
- 238000012549 training Methods 0.000 claims description 13
- 230000004927 fusion Effects 0.000 claims description 12
- 238000005520 cutting process Methods 0.000 claims description 11
- 230000009467 reduction Effects 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 9
- 238000013434 data augmentation Methods 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 3
- 230000003321 amplification Effects 0.000 claims description 2
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 2
- 230000007306 turnover Effects 0.000 claims description 2
- 238000000605 extraction Methods 0.000 abstract description 2
- 230000000007 visual effect Effects 0.000 abstract 1
- 238000012360 testing method Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000007689 inspection Methods 0.000 description 3
- 238000012805 post-processing Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 238000012800 visualization Methods 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
- G06T7/0004—Industrial image inspection
-
- 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
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- 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
- G06T2207/10036—Multispectral image; Hyperspectral 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
- G06T2207/20081—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
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
-
- 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
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for detecting a power transmission line tower in a satellite image, and belongs to the field of target detection. The method provided by the invention can be used for optimally extracting the small target characteristics of the power transmission line tower of the multi-source remote sensing satellite image, so that the extracted characteristics can better reflect the characteristics of the power transmission line tower, and the attached coordinate information is finally output to the target detection of the tower, thus the method has a stronger application prospect. In order to improve the learning capacity of a power transmission line tower detection network on a tower small target, a characteristic pyramid extraction module is further built, and the purpose of learning multi-scale tower characteristics is achieved; in order to better utilize each characteristic layer to learn tower characteristics with different scales, the size of a tower target is further estimated in advance, and the size of target anchor frames of different characteristic layers is designed, so that tower targets with different scales are predicted more accurately; meanwhile, the tower target is cut, input and output are combined, and the geographical coordinate information is attached, so that the visual output of the power grid application scene is facilitated, and the readability is realized.
Description
Technical Field
The invention belongs to the field of satellite image identification, and particularly relates to identification of a power transmission line tower.
Background
In order to reduce the potential safety hazard of the power transmission line caused by the fault of the power transmission line tower, a power grid 'sky-ground cooperation' inspection key technology research is carried out, and a power grid department accumulates a large amount of remote sensing satellite image data, unmanned aerial vehicle inspection data and power transmission line tower ledger data. In the development of pole tower identification and detection research of a power transmission line, the existing method is based on unmanned aerial vehicle routing inspection data and is combined with a deep learning method to detect the pole tower, and due to the fact that the pole tower has a large target, part of research can obtain good detection accuracy. However, in the remote sensing satellite image, there is no study on the problem of the target size and resolution, and the transmission line tower can be detected on the remote sensing satellite image with high resolution with high precision.
Disclosure of Invention
The invention aims to provide a method for detecting a power transmission line tower in a satellite image, which is used for solving the problems of identification and detection of the power transmission line tower based on a remote sensing satellite image at present. The method solves the problem that the characteristics of the power transmission line tower based on the remote sensing satellite image are difficult to extract, solves the problem of high-precision identification and detection of the power transmission line tower based on the remote sensing satellite image, and is applied to the target detection of the power transmission line tower based on the remote sensing satellite image in the Yunnan power grid.
In order to realize the purpose, the technical scheme of the invention is as follows: a method for detecting a power transmission line tower in a satellite image comprises the following steps:
step 1: acquiring a satellite image containing a power transmission line;
and 2, step: performing orthorectification on the satellite image acquired in the step 1;
and step 3: fusing the multispectral image and the panchromatic band image of the satellite by adopting a Gram-Schmidt orthogonalization method to obtain a red band, green band, blue band and near infrared band 4-channel fused image;
and 4, step 4: normalizing the fused image in a 0-255 pixel range;
and 5: performing 4 times of super-resolution amplification on the fusion image by using an expandable deep super-resolution network;
step 6: cutting the fused image by 1024 x 1024 pixels, marking a tower in the cut image, and recording a label as the tower to obtain a training data set;
and 7: carrying out random cutting, random noise addition and horizontal turnover on the training data to obtain data after data augmentation;
and step 8: building a tower detection network;
step 8.1: after the fused image is input, the fused image firstly passes through a depth residual error network to obtain the characteristic { C 2 ,C 3 ,C 4 ,C 5 }，{C 2 ,C 3 ,C 4 ,C 5 Each element in the graph corresponds to a feature map with different scales;
step 8.2: will { C 2 ,C 3 ,C 4 ,C 5 In C 2 ,C 3 ,C 4 ,C 5 Dimension reduction by 1 × 1 convolution kernel, C 3 ,C 4 ,C 5 2 times of upsampling is carried out, and then the C after dimensionality reduction is carried out 2 And C after 2 times up-sampling 3 Performing feature fusion to obtain a feature P 2 (ii) a Reducing dimension C 3 And C after 2 times up-sampling 4 Performing feature fusion to obtain a feature P 3 (ii) a Reducing dimension C 4 And C after 2 times up-sampling 5 Performing feature fusion to obtain a feature P 4 ；C 5 Directly obtaining the characteristic P after dimension reduction through 1 multiplied by 1 convolution kernel 5 (ii) a Finally, the characteristic { P is obtained 2 ,P 3 ,P 4 ,P 5 }；
Step 8.3: feature { P } 2 ,P 3 ,P 4 ,P 5 Inputting the area suggestion network to obtain target anchor frames of towers with different sizes, then pooling, and dividing the frames into two branches, wherein one branch passes through a full connection layer and a classification layer in sequence to obtain whether a target in the target anchor frame is a tower; the other branch passes through the full connection layer and the boundary regression layer in sequence to adjust the position of the target anchor frame;
and step 9: taking the data set obtained in the step 6 as input, and training the tower detection network obtained in the step 8;
step 10: when the detection application of the power transmission line tower is carried out, specifically, cutting 1024 × 1024 pixels of the image normalized in the step 4, detecting one by one, and drawing a detection frame; and finally merging the cut detection images into the size of the original image.
Further, the area in the step 8.2 suggests that the size of a target anchor frame corresponding to the tower detection network in the network is {30,60,80,120 };
further, in step 9, the number of training iterations is 100000, the initial learning rate is 0.001, attenuation is performed at the number of iterations of 50000 and 70000, the attenuation factor is 0.1, 0.001, and the Batch size is 4.
The method for detecting the power transmission line tower by the multisource remote sensing satellite image based on the deep learning has the advantages of high detection precision, high positioning accuracy and high speed; according to the method, the characteristics of the small target of the power transmission line tower in the remote sensing image are extracted by building the tower detection network, so that the extracted characteristics can better reflect the shape and size of the tower, the tower detection precision is higher, and the small target omission ratio is lower. In order to improve the detection precision of the small target transmission line tower, the size of a target anchor frame is further calculated according to the size of the tower label in the data set, and then the corresponding size of the target anchor frame is designed according to each characteristic layer, so that the problem that the tower characteristics of the small target are difficult to learn is avoided. In order to enable the invention to fall into an application which is convenient for visualization, a cutting input and output combining mode is introduced, and the corresponding geographic coordinate information of the transmission line tower is output, so that the invention has a complete transmission line tower detection application function. The invention provides a multi-source remote sensing satellite image power transmission line tower detection method based on deep learning. The method starts from the small target feature learning direction, multi-level learning is carried out on the multi-source remote sensing satellite image power transmission line tower features, model parameter learning is carried out according to the extracted tower features, a set of multi-source remote sensing satellite image power transmission line tower detection processes is researched and formulated, and a high-precision multi-source remote sensing satellite image power transmission line tower detection model is constructed. The method can be used for power grid transmission line investigation, transmission line operation and maintenance and the like.
Drawings
FIG. 1 is a flow chart of a multi-source remote sensing satellite image transmission line tower detection method based on deep learning;
FIG. 2 is a network architecture diagram of a tower detection network according to the present invention;
FIG. 3 is a diagram of the results of the WorldView-3 transmission line tower detection part of the present invention.
Detailed Description
The present invention will be described in detail with reference to the following embodiments.
The technical scheme of the invention is shown in figure 1, which takes a certain power transmission line in Kunming City of Yunnan province as an example for explanation, and comprises the following steps:
step 1: data pre-processing
Opening the acquired multispectral images and panchromatic band images of SuperView-1 and WorldView-3 in the same area by using Envi software, selecting an orthorectification tool, performing orthorectification by using elevation (DEM) data of 90m multiplied by 90m, and then performing remote sensing image fusion on the orthorectified multispectral and panchromatic band images by using Gram-Schmidt algorithm, wherein the spectral advantages of the multispectral and the resolution advantages of the panchromatic band are combined. Finally, obtaining red, green, blue and near-infrared 4-channel images, and carrying out pixel normalization of 0-255, wherein the resolution of SuperView-1 is 0.5m, and the resolution of WorldView-3 is 0.3 m.
Step 2: remote sensing satellite image super-resolution
In order to avoid insufficient memory of a computer, pre-cutting SuperView-1 and WorldView-3 images fused by Gram-Schmidt algorithm, inputting the cut images into a super-resolution network as input images, and performing 4-time super-resolution processing, wherein the resolution of the SuperView-1 is 0.125m, and the resolution of the WorldView-3 is 0.075 m.
And step 3: hyperspectral image post-processing
Through experiments, the tower characteristics in the red, near-infrared and blue three-channel images are more obvious than the background, so that the remote sensing satellite images after being subjected to overdivision are subjected to red, near-infrared and blue three-channel false color output, meanwhile, the images are subjected to 1% linear stretching, and the tower target characteristic identification degree is improved.
And 4, step 4: transmission line tower detection data set manufacturing method
And (4) cutting the super-resolution image after the post-processing in the step (3) by 1024 pixels, then selecting a cut image containing the tower of the power transmission line, labeling the target of the tower, and marking the label as the tower, wherein the total number of the labels is 870. And carrying out data augmentation on the marked cut image, wherein the data augmentation mainly comprises random cutting, random noise addition, horizontal turning and the like, and the number of the augmented cut image is 6960.
And 5: transmission line tower detection model construction
As shown in figure 2, a power transmission line tower detection model is constructed by firstly constructing a feature extraction module which is mainly composed of a depth residual error network to obtain features { C } 2 ,C 3 ,C 4 ,C 5 }，{C 2 ,C 3 ,C 4 ,C 5 Each element in the graph corresponds to a feature map with different scales; secondly, will { C 2 ,C 3 ,C 4 ,C 5 In C 2 ,C 3 ,C 4 ,C 5 Dimension reduction by 1 × 1 convolution kernel, C 3 ,C 4 ,C 5 2 times of upsampling is carried out, and then the C after dimensionality reduction is carried out 2 And C after 2 times up-sampling 3 Performing feature fusion to obtain a feature P 2 (ii) a Reducing dimension C 3 And C after 2 times up-sampling 4 Performing feature fusion to obtain a feature P 3 (ii) a Reducing the dimension of C 4 And C after 2 times up-sampling 5 Performing feature fusion to obtain a feature P 4 ；C 5 Directly obtaining the characteristic P after dimension reduction through 1 multiplied by 1 convolution kernel 5 (ii) a Finally, the characteristic layer { P is obtained 2 ,P 3 ,P 4 ,P 5 }; and then, constructing a regional suggestion network to obtain target anchor frames of towers with different sizes, wherein the feature layer { P } 2 ,P 3 ,P 4 ,P 5 The sizes of anchor frames corresponding to different scales are {30,60,80,120 and 160}, then a pooling layer is built and is divided into two branches, and one branch passes through a full-connection layer and a classification layer in sequence; the other branch passes through the full connection layer and the boundary regression layer in sequence.
Step 6: tower detection model training
And (4) designing a corresponding data format for the training data obtained in the step (4) and putting the training data into the power transmission line tower detection model established in the step (5), wherein the main parameters related to the characteristic pyramid model comprise a learning rate, a batch size, an initial momentum value, iteration times, a target anchor frame scaling scale and the like. The learning rate and the target anchor frame scaling are the more important parameters in the model, and the specific parameter cases are shown in table 1. After the training is completed, the optimization model is saved.
TABLE 1 partial hyper-parameters involved in tower detection model
Step 6: tower detection application and model evaluation
Selecting different areas as test data according to SuperView-1 and WorldView-3 images subjected to the over-resolution post-processing in the step 3, firstly cutting by 1024 pixels, inputting the cut images into the trained optimal model in the step 5, comparing the final cut test images with the true values, performing progress calculation, finally combining the cut test images into the original image size, and simultaneously reserving the detected geographical coordinate information of the power transmission line tower through a Gdal library. The test results are shown in Table 2.
TABLE 2 Tower testing accuracy results
And 7: results verification and interpretation
According to the example, the power transmission line tower belongs to the small target category in the remote sensing satellite image, the small target tower can be accurately positioned and detected in the characteristic pyramid power transmission line tower detection model established by the invention, and the detection precision is high. Looking at fig. 3, it can be seen that in the two types of remote sensing satellite images, the deformation of the pole target in the SuperView-1 image data is serious, the pole target can be well detected in the feature pyramid transmission line pole detection network constructed by the invention, and the deformation of the pole target in the WorldView-3 image data is small, so that the pole target can be detected with higher precision in the feature pyramid transmission line pole detection network constructed by the invention.
The method has the advantages that the tower target can be detected on the multisource remote sensing satellite image quickly and accurately, and corresponding coordinate information is output. The detection precision of the characteristic pyramid target detection model obtained by training of the invention can reach more than 80% for the transmission line tower, which indicates that the result has higher reliability. The method can be used for power grid transmission line investigation, transmission line operation and maintenance and the like.
Claims (3)
1. A method for detecting a power transmission line tower in a satellite image comprises the following steps:
step 1: acquiring a satellite image containing a power transmission line;
step 2: performing orthorectification on the satellite image acquired in the step 1;
and step 3: fusing the multispectral image and the panchromatic band image of the satellite by adopting a Gram-Schmidt orthogonalization method to obtain a red band, green band, blue band and near infrared band 4-channel fused image;
and 4, step 4: normalizing the fused image in a 0-255 pixel range;
and 5: performing 4-time super-resolution amplification on the fusion image by using an expandable deep super-resolution network;
step 6: cutting the fused image by 1024 x 1024 pixels, marking a tower in the cut image, and recording a label as the tower to obtain a training data set;
and 7: carrying out random cutting, random noise addition and horizontal turnover on training data to obtain data after data augmentation;
and 8: building a tower detection network;
step 8.1: after the fused image is input, the fused image firstly passes through a depth residual error network to obtain the characteristic { C 2 ,C 3 ,C 4 ,C 5 }，{C 2 ,C 3 ,C 4 ,C 5 Each element in the graph corresponds to a feature map with different scales;
step 8.2: will { C 2 ,C 3 ,C 4 ,C 5 In C 2 ,C 3 ,C 4 ,C 5 Dimension reduction by 1 × 1 convolution kernel, C 3 ,C 4 ,C 5 2 times of upsampling is carried out, and then the C after dimensionality reduction is carried out 2 And C after 2 times up-sampling 3 Performing feature fusion to obtain a feature P 2 (ii) a Reducing dimension C 3 And C after 2 times up-sampling 4 Performing feature fusion to obtain a feature P 3 (ii) a Reducing dimension C 4 And C after 2 times up-sampling 5 Performing feature fusion to obtain a feature P 4 ；C 5 Directly obtaining the characteristic P after dimension reduction through 1 multiplied by 1 convolution kernel 5 (ii) a Finally, the characteristic { P is obtained 2 ,P 3 ,P 4 ,P 5 }；
Step 8.3: will feature { P 2 ,P 3 ,P 4 ,P 5 Inputting the area suggestion network to obtain target anchor frames of towers with different sizes, then pooling, and dividing the frames into two branches, wherein one branch passes through a full connection layer and a classification layer in sequence to obtain whether a target in the target anchor frame is a tower; the other branch passes through the full connection layer and the boundary regression layer in sequence to adjust the position of the target anchor frame;
and step 9: taking the data set obtained in the step 6 as input, and training the tower detection network obtained in the step 8;
step 10: when the detection application of the power transmission line tower is carried out, specifically, cutting 1024 x 1024 pixels is carried out on the image normalized in the step 4, detection is carried out one by one, and a detection frame is drawn; and finally merging the cut detection images into the size of the original image.
2. The method for detecting the power transmission line tower in the satellite image according to claim 1, wherein the target anchor frame size corresponding to the tower detection network in the area recommendation network in the step 8.2 is {30,60,80,120 }.
3. The method for detecting the power transmission line tower in the satellite image according to claim 1, wherein in the step 9, the number of training iterations is 100000, the initial learning rate is 0.001, attenuation is performed at the number of iterations of 50000 and 70000, the attenuation factor is 0.1 and 0.001, and the Batch size is 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210588429.1A CN114998242A (en) | 2022-05-26 | 2022-05-26 | Method for detecting power transmission line pole tower in satellite image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210588429.1A CN114998242A (en) | 2022-05-26 | 2022-05-26 | Method for detecting power transmission line pole tower in satellite image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114998242A true CN114998242A (en) | 2022-09-02 |
Family
ID=83029886
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210588429.1A Pending CN114998242A (en) | 2022-05-26 | 2022-05-26 | Method for detecting power transmission line pole tower in satellite image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114998242A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108596103A (en) * | 2018-04-26 | 2018-09-28 | 吉林大学 | High resolution ratio satellite remote-sensing image building extracting method based on optimal spectrum Index selection |
| CN110728658A (en) * | 2019-09-16 | 2020-01-24 | 武汉大学 | High-resolution remote sensing image weak target detection method based on deep learning |
| CN113033446A (en) * | 2021-04-01 | 2021-06-25 | 辽宁工程技术大学 | Transmission tower identification and positioning method based on high-resolution remote sensing image |
| CN113673586A (en) * | 2021-08-10 | 2021-11-19 | 北京航天创智科技有限公司 | Mariculture area classification method fusing multi-source high-resolution satellite remote sensing images |
| CN113989612A (en) * | 2021-05-13 | 2022-01-28 | 中国地质大学（武汉） | Remote sensing image target detection method based on attention and generation countermeasure network |
-
2022
- 2022-05-26 CN CN202210588429.1A patent/CN114998242A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108596103A (en) * | 2018-04-26 | 2018-09-28 | 吉林大学 | High resolution ratio satellite remote-sensing image building extracting method based on optimal spectrum Index selection |
| CN110728658A (en) * | 2019-09-16 | 2020-01-24 | 武汉大学 | High-resolution remote sensing image weak target detection method based on deep learning |
| CN113033446A (en) * | 2021-04-01 | 2021-06-25 | 辽宁工程技术大学 | Transmission tower identification and positioning method based on high-resolution remote sensing image |
| CN113989612A (en) * | 2021-05-13 | 2022-01-28 | 中国地质大学（武汉） | Remote sensing image target detection method based on attention and generation countermeasure network |
| CN113673586A (en) * | 2021-08-10 | 2021-11-19 | 北京航天创智科技有限公司 | Mariculture area classification method fusing multi-source high-resolution satellite remote sensing images |
Non-Patent Citations (2)
| Title |
|---|
| SHAOQING REN等: "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks", 《COMPUTER VISION AND PATTERN RECOGNITION》, 6 January 2016 (2016-01-06) * |
| TSUNG-YI LIN等: "Feature Pyramid Networks for Object Detection", 《COMPUTER VISION AND PATTERN RECOGNITION》, 19 April 2017 (2017-04-19), pages 3 - 5 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110378196B (en) | Road visual detection method combining laser point cloud data | |
| CN111080629B (en) | Method for detecting image splicing tampering | |
| Liu et al. | Multiscale U-shaped CNN building instance extraction framework with edge constraint for high-spatial-resolution remote sensing imagery | |
| CN110929607B (en) | Remote sensing identification method and system for urban building construction progress | |
| CN112083422B (en) | Single-navigation InSAR system end-to-end classification method based on multistage deep learning network | |
| CN110991430B (en) | Ground feature identification and coverage rate calculation method and system based on remote sensing image | |
| CN113239736B (en) | Land coverage classification annotation drawing acquisition method based on multi-source remote sensing data | |
| Dibs et al. | Multi-fusion algorithms for detecting land surface pattern changes using multi-high spatial resolution images and remote sensing analysis | |
| CN111652039B (en) | Hyperspectral remote sensing ground object classification method based on residual error network and feature fusion module | |
| CN114022408A (en) | Remote sensing image cloud detection method based on multi-scale convolution neural network | |
| CN112700489B (en) | Ship-based video image sea ice thickness measurement method and system based on deep learning | |
| CN114066755A (en) | Remote sensing image thin cloud removing method and system based on full-band feature fusion | |
| CN111640116A (en) | Aerial photography graph building segmentation method and device based on deep convolutional residual error network | |
| CN114937206A (en) | Hyperspectral image target detection method based on transfer learning and semantic segmentation | |
| Khatriker et al. | Building footprint extraction from high resolution satellite imagery using segmentation | |
| CN115661655A (en) | Southwest mountain area cultivated land extraction method with hyperspectral and hyperspectral image depth feature fusion | |
| CN117152414A (en) | Target detection method and system based on scale attention auxiliary learning method | |
| Chawda et al. | Extracting building footprints from satellite images using convolutional neural networks | |
| CN111738052A (en) | Multi-feature fusion hyperspectral remote sensing ground object classification method based on deep learning | |
| Jaffe et al. | Remote sensor design for visual recognition with convolutional neural networks | |
| CN115240066A (en) | Remote sensing image mining area greening monitoring method and system based on deep learning | |
| CN114998242A (en) | Method for detecting power transmission line pole tower in satellite image | |
| CN115019174B (en) | Up-sampling remote sensing image target recognition method based on pixel recombination and attention | |
| CN113887656B (en) | Hyperspectral image classification method combining deep learning and sparse representation | |
| CN116205879A (en) | Unmanned aerial vehicle image and deep learning-based wheat lodging area estimation method |
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 |