CN105550709B - A kind of remote sensing image power transmission line corridor wood land extracting method - Google Patents
A kind of remote sensing image power transmission line corridor wood land extracting method Download PDFInfo
- Publication number
- CN105550709B CN105550709B CN201510930084.3A CN201510930084A CN105550709B CN 105550709 B CN105550709 B CN 105550709B CN 201510930084 A CN201510930084 A CN 201510930084A CN 105550709 B CN105550709 B CN 105550709B
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- scene
- sensing image
- scene unit
- training sample
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of remote sensing image power transmission line corridor wood land extracting methods, it include: step 1, the training of SVM classifier: extracting scene unit training sample from remote sensing image training sample, defines the scene type of each scene unit training sample using artificial visual mode and using labelled notation；Extract the global characteristics of each scene unit training sample；Using the scene type label and global characteristics combined training SVM classifier of scene unit training sample；Step 2, the SVM classifier that use has been trained classifies to remote sensing image test data: extracting scene unit from remote sensing image test data；Extract the global characteristics of each scene unit；Global characteristics combination based on scene unit, the SVM classifier that use has been trained extract wood land.The present invention can reinforce to the completeness of goal description is extracted, to improve the accuracy rate of wood land extraction.
Description
Technical field
The invention belongs to remote sensing image intelligent analysis technical field, in particular to a kind of remote sensing image power transmission line corridor
Wood land extracting method.
Background technique
Transmission line of electricity is the lifeblood of operation of power networks, is " lifeline " to involve the interests of the state and the people.With Chinese national economy
Fast development, the building cause of China's power grid have stepped into the brand-new epoch.Extra high voltage network has become the major network of China's power grid
Frame, UHV transmission line also build up in succession and put into effect.It is worth noting that, mostly by wood land along transmission line of electricity
When, since forest weeds are numerous in power transmission line corridor, easily cause mountain fire when suffering from a severe drought, ploughing in the spring reclamation of wasteland etc., cause
Transmission line of electricity tripping.Hubei power transmission line corridor region has the characteristics that weather is complicated, landform is changeable, vegetation is luxuriant, especially
In recent years, inside the province transmission line of electricity because of brushfire, burn the grass on waste land etc. caused by the power failure fault of stop that trips on a large scale it is more and more.From
2005 so far, and the power transmission and transformation line trip accident caused in Hubei Province by mountain fire shares more than 10 times.The especially winter in 2013
Season is dry, and only nearby generation more than 30 plays mountain fire to 500kV route daily within the border in Hubei, so that each O&M department investment of power grid is a large amount of
Manpower and material resources are taken precautions against.Therefore, the forest of the multiple sensitizing range of power transmission line corridor region, especially mountain fire is carried out effective
It identification and extracts, to instructing power transmission line corridor vegetation management meaning important.
Resource three are the civilian stereo mapping satellites of first of China high-precision, provisioned in three line scanner mapping camera and more
Three line scanner imaging mode can be used in the payload such as spectrum camera, by facing, forward sight, backsight and multispectral camera acquisition seen
Survey the high-resolution remote sensing image in region.In image other than the spectral information containing common remote sensing image, also comprising in large quantities
Shape and structure, texture information, contextual information of table object etc..For comparatively low resolution multispectral image, high-resolution is distant
Feeling image has spatial information abundant and atural object geometry, and texture information is also more obvious, easily facilitates cognition atural object
Attributive character.Therefore more passes are also received to the fast understanding and automatic marking problem of " resource three " satellite image scene
Note.
With the continuous development of power grid construction, high pressure, UHV transmission line will cover more multi-environment complex region, especially
It is the emphasis paid high attention to extensive forest region.Understand forest-covered area and growth characteristics, can be power transmission line corridor
Design and external insulation cooperation provide reliable basis；Meanwhile to reasonable selection transmission line of electricity power transmission path trend and vegetation management
It is worth with important references.Due to remote sensing image in imaging process by such environmental effects so that same category scene exists
Illumination, direction show larger difference on scale.On the other hand, different classes of scene may include same target classification, only
Since difference of the target category on spatial position and density degree results in the difference of scene type[1].Therefore, because remote sensing
The complexity and variability of scene, so that being therefrom extracted into the forest atural object intelligence in power transmission line corridor region for one
Challenging and significant work.
With the universal and development of computer technology, intelligent, mechanization has become the hot spot being concerned.Machine learning
It is to copy human brain working method, computer learning is allowed to be similar to the working method of brain.It is completed in mankind's daily life
Various simple actions need if computer is allowed to complete using highly complex neural network[2].Hinton et al.[3]It mentions
Deep learning out has started a tide of neural network research.Deep learning is to facilitate to understand picture, sound in order to obtain
The data such as sound, text institute's elocutionary meaning and carry out it is multi-level expression and extraction study[4].And with simple machine learning, such as
With the neural network for containing 2 or 3 hidden layers, it is unable to reach the differentiation decision similar with the mankind.This just needs the study of multilayer
Device, Level by level learning and the Knowledge delivery that study is arrived give next layer, so that lower layer can obtain the expression form of higher level, phase
Hope the available conclusion similar with the mankind[5]。
Following bibliography involved in text:
[1] Xu Kan, Yang Wen, the such as Chen Lijun utilize remote sensing images scene classification [J] Wuhan University Journal of topic model:
Information science version, 2011, (5): 540-543.
[2]HAYKIN S.Neural Networks:A comprehensive foundation[M].2nd ed.New
York:Prentice-Hall,1999.
[3]HINTON G E,MCCLELLAND J L,RUMELHART D E.Distributed
Representations[M].Cambridge:MIT Press,1986.
[4]BENGIO Y.Learning deep architectures for AI[J].Foundations and
Trends in Machine Learning,2009,2(1):1-127.
[5] Li Haifeng, Li Chun fruit deep learning structure and algorithm comparison analyze [J] University Of Hebei journal (natural science
Version), 2012,05:538-544.
[6]A.Krizhevsky,I.Sutskever,and G.E.Hinton.Imagenet classification
with deep convolutional neural networks[C].In NIPS,2012.
[7]T.-H.Chan*,K.Jia,S.Gao,J.Lu,Z.Zeng,and Y.Ma,"PCANet:A simple deep
Learning baseline for image classification? " [C], CoRR, 2014.
Summary of the invention
The object of the present invention is to provide a kind of remote sensing image power transmission line corridor forests based on the combination of deep learning feature
Method for extracting region.
The present invention is based on " resource three " high-resolution remote sensing images to carry out the extraction of power transmission line corridor wood land.It is first
First, three kinds of deep learning feature extractions are carried out respectively to remote sensing image, i.e. CNN feature, PCANet feature and RandomNet is special
Sign；Then tandem compound is carried out, the combination of deep learning feature is obtained；It is finally combined, is utilized according to the feature of the form
Support vector machines (SVM) classifier completes the classification to the forest in remote sensing image power transmission line corridor region.
In order to achieve the above objectives, the present invention adopts the following technical scheme:
A kind of remote sensing image power transmission line corridor wood land extracting method, comprising:
Step 1, the training of SVM classifier, specifically:
1.1 extract scene unit training sample from remote sensing image training sample, define each scene using artificial visual mode
The scene type of module training sample simultaneously uses labelled notation, and the scene type includes forest class and non-forest class；
1.2 extract CNN feature, PCANet feature and the RandomNet feature of each scene unit training sample, then will extract
Three kinds of features carry out tandem compound constitute scene unit training sample global characteristics；
1.3 train SVM classifier using the scene type label and global characteristics of scene unit training sample；
Step 2, the SVM classifier that use has been trained classifies to remote sensing image test data, specifically:
2.1 extract scene unit from remote sensing image test data；
2.2 extract CNN feature, PCANet feature and the RandomNet feature of each scene units, then by three kinds of spies of extraction
Sign carries out the global characteristics that tandem compound constitutes scene unit；
2.3 global characteristics based on scene unit, the SVM classifier that use has been trained is to remote sensing image test data midfield
Scape unit is classified, wherein the scene unit for being divided into forest class constitutes wood land.
Scene unit is extracted in sub-step 1.1 and sub-step 2.1, specifically:
Remote sensing image is divided using uniform grid, a grid represents a scene unit, nothing between adjacent scene unit
Overlapping, the remote sensing image are remote sensing image training sample or remote sensing image test data.
Compared to the prior art, the invention has the advantages that and the utility model has the advantages that
Compared with tradition is based on the classification method of hand-designed (hand-crafted) feature, the present invention utilizes deep learning
The unsupervised learning feature of feature carries out adaptive-filtering processing；Meanwhile being combined a variety of deep learning features, it merges
The advantage of each deep learning feature.The global characteristics combined by a variety of deep learning features introducing remote sensing image transmission line of electricity is walked
The forest classified in corridor region can be reinforced to the completeness for extracting goal description, to improve forest classified accuracy rate.
Detailed description of the invention
Fig. 1 is the specific flow diagram of the method for the present invention.
Specific embodiment
Below in conjunction with attached drawing, detailed description of the preferred embodiments.Technical solution of the present invention can be used
Computer software automatic running.
Step 1, remote sensing image pre-processes.
This step includes the pretreatment of remote sensing image training sample and remote sensing image test data.
Include extraction scene unit to the pretreatment of remote sensing image training sample and defines scene class in a manner of artificial visual
Not.Firstly, substantially the identical image subblock of several sizes will be divided into remote sensing image training sample, scene unit training sample is obtained
This；Then the scene type of each scene unit training sample is defined using artificial visual mode, and is marked using label, the present invention
Middle scene type only includes forest class and non-forest class.The pretreatment of remote sensing image test data is to extract its scene unit.
When it is implemented, remote sensing image is " resource three " panchromatic image, remote sensing image, grid are divided using uniform grid
Just represent a scene unit, it is non-overlapping between adjacent scene unit.Final goal of the present invention is to give remote sensing image test data
All scene units assign a scene type label, and are distinguished using different colours.In the present embodiment, scene unit size is
100 pixel *, 100 pixel.
Step 2, the deep learning feature extraction of scene unit.
Feature representation based on deep learning can be automatically performed the gradually transmitting by low-level feature to high-level characteristic, relatively
For manual feature, deep learning feature more meets human visual system.The present invention is based on scene units, extract scene respectively
Tri- kinds of deep learning features of CNN, PCANet and RandomNet of unit describe scene unit, and by three kinds of depth of extraction
It is used to classify after practising feature combination.
The present invention needs to extract the deep learning of the remote sensing image test data scene unit of scene unit training sample sum
Feature below will be described in detail deep learning characteristic extraction procedure.
(1) CNN feature (convolutional neural networks feature) is extracted.
Convolutional neural networks include convolutional layer, pond layer (pooling layer) and full articulamentum.Complete depth convolution
Neural network is made of multiple convolution, pond and full articulamentum tandem compound.In convolutional layer, input picture or input feature vector figure with
Several filter groups (also referred to as convolution kernel) carry out convolutional filtering and obtain characteristic pattern；Then to being sent into pond after its nonlinear processing
Change layer, above-mentioned processing be based on nonlinear function calculated, common nonlinear function have Sigmoid function, Tanh function,
ReLU (Rectified Linear Unit) function etc..
Pondization operation is carried out to the characteristic pattern of input in the layer of pond.It is identical that characteristic pattern is divided into size by pondization operation
Square region, to each region Counting statistics amount, as in characteristic pattern in fixed size window the average value of all pixels response or
Maximum value.By pond layer, it is equivalent to and down-sampling is carried out to characteristic pattern, obtain the lesser characteristic pattern of size.Pond layer
Output can reconnect a convolutional layer, the output of the convolutional layer reconnects another pond layer, before the reasonable network number of plies
It puts, according to convolutional layer-pond layer sequential iteration.The statistic of the last layer pond layer output is sent into full articulamentum.
Full articulamentum is made of several full connection hidden layers and Softmax Regression decision-making level.Convolutional neural networks
Training utilize back-propagation algorithm (Back Propagation) complete.Training is completed after obtaining each layer of network parameter,
To any one width input picture can be calculated by feedforward (Feed-forward) mode the input picture feature, i.e. CNN is special
Sign.CNN feature has good translation and scale invariance.
(2) PCANet feature (PCA network characterization) is extracted.
PCANet feature is that the convolution kernel of PCANet feature is by unsupervised PCA with the maximum difference of CNN feature
Training obtains, rather than uses back-propagation algorithm.In the present embodiment, scene list is extracted using existing PCANet feature tools packet
The PCANet feature of member, key step are as follows:
(a) the small image block of stochastical sampling in scene module training sample carries out PCA transformation to the small image block of sampling,
L1 principal component vector does the total L1 characteristic pattern that convolution obtains first layer as convolution kernel, convolution kernel and training image before extracting.
(b) characteristic pattern is sampled and is trained to obtain L2 convolution kernel using PCA, the L1 characteristic pattern convolution with first layer
The total L1 group characteristic pattern of the second layer is obtained afterwards, and every group includes L2 characteristic pattern.
(c) Hash coding, code length 2 are carried out respectively to each group characteristic pattern of the second layerL2, L1 coding is obtained
Figure.
(d) piecemeal processing is carried out to each code pattern, the statistic histogram of each sub-block frequency of occurrences is obtained, by L1 code pattern
The histogram of all sub-blocks is connected, i.e. the feature of input picture.
(3) RandomNet feature extraction.
RandomNet feature and PCANet feature the difference is that, the convolution kernel of first layer and the second layer is obstructed
It crosses non-supervisory PCA training to obtain, but the random number of one group of Gaussian distributed is randomly generated.In the present embodiment, use is existing
RandomNet feature tools packet extract scene unit RandomNet feature.
Step 3, deep learning feature combines.
Three feature vectors of the scene unit that step 2 is extracted are sequentially connected with cascade, obtain the depth of scene unit
Spend learning characteristic combination, the i.e. global characteristics of scene unit.Using the global characteristics of scene unit training sample to SVM classifier
It is trained, the SVM classifier that use has been trained classifies to remote sensing image test data scene unit.
Step 4, classified using SVM classifier to remote sensing image test data.
By above step, the complete of all scene unit training samples and remote sensing image test data scene unit can get
Office's feature.Based on this, the present invention completes the forest classified in power transmission line corridor region, assorting process master using SVM classifier
It to include training SVM classifier and the classification of remote sensing image test data scene unit.
(1) using the scene type label of scene unit training sample and global characteristics training SVM classifier, SVM points are obtained
The model parameter of class device.
Used SVM classifier is using maximum class interval as criterion, by finding optimal separating hyper plane to training sample
The division of this progress space.The training process of SVM classifier, which is eventually converted into, solves following optimization problem:
s.t,yi(wTφ(xi)+b)≥1-ξi, i=1,2 ... n (1)
ξi>=0, i=1,2 ... n
In formula (1), w and b are defining classification hyperplane wTφ(xi)+b=0 parameter vector；C is for controlling target letter
The constant of two weights in number；ξiIndicate slack variable；φ(xi) indicate training sample xiNonlinear Mapping；yiIndicate training
Sample xiCategory label；N is number of training.
Using method of Lagrange multipliers, the optimization problem of belt restraining shown in formula (1) is converted to unconstrained optimization problem,
Cost function L are as follows:
In formula (2), αi、riIt is training sample xiVariable to be optimized.
On the basis of meeting Karush-Kuhn-Tucker (KKT) condition, unconstrained problem shown in formula (2) can be turned
It is melted into following optimization problem:
s.t.,0≤αi≤ C, i=1 ..., n (3)
In formula (3), meet between w and α
K(x,xi) it is preparatory defined kernel function, for inner product operation of the training sample in feature space is converted
At kernel function mapping of the training sample in luv space.The present embodiment is as follows as kernel function using radial basis function:
Wherein, x is supporting vector, xiFor training sample, σ is the variance of training sample.
Variable { α is obtained by solution optimization problem (3)i}I=1 ..., nOptimal value, complete SVM classifier model parameter training
Process.
Pair (2) classified using the SVM classifier trained to all scene units in remote sensing image test data, i.e.,
Each scene unit carries out the prediction of scene type label, can be obtained the class of each scene unit by the scene type label of prediction
Not, different scene types is distinguished using different colours.
When carrying out scene classification using the SVM classifier trained, it is only necessary to by remote sensing image test data scene unit
Feature vector input SVM classifier classification function curved surface, corresponding scene type label, classification function f (H can be obtained(t)) it is as follows:
In formula (5), HiFor training sample xiGlobal characteristics, H(t)Indicate the global characteristics of scene unit to be sorted.
Specific embodiment described herein is only an example for the spirit of the invention.The neck of technology belonging to the present invention
The technical staff in domain can make various modifications or additions to the described embodiments or replace by a similar method
In generation, however, it does not deviate from the spirit of the invention or beyond the scope of the appended claims.
Claims (2)
1. a kind of remote sensing image power transmission line corridor wood land extracting method, characterized in that include:
Step 1, the training of SVM classifier, specifically:
1.1 extract scene unit training sample from remote sensing image training sample, define each scene unit using artificial visual mode
The scene type of training sample simultaneously uses labelled notation, and the scene type includes forest class and non-forest class；
1.2 extract CNN feature, PCANet feature and the RandomNet feature of each scene unit training samples, then by the three of extraction
Kind feature carries out the global characteristics that tandem compound constitutes scene unit training sample；
1.3 train SVM classifier using the scene type label and global characteristics of scene unit training sample；
Step 2, the SVM classifier that use has been trained classifies to remote sensing image test data, specifically:
2.1 extract scene unit from remote sensing image test data；
2.2 extract CNN feature, PCANet feature and the RandomNet feature of each scene units, then by three kinds of features of extraction into
The global characteristics of row tandem compound composition scene unit；
2.3 global characteristics based on scene unit, the SVM classifier that use has been trained is to scene in remote sensing image test data
Unit is classified, wherein the scene unit for being divided into forest class constitutes wood land.
2. remote sensing image power transmission line corridor as described in claim 1 wood land extracting method, it is characterized in that:
Scene unit is extracted in sub-step 1.1 and sub-step 2.1, specifically:
Remote sensing image is divided using uniform grid, a grid is to represent a scene unit, it is non-overlapping between adjacent scene unit,
The remote sensing image is remote sensing image training sample or remote sensing image test data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510930084.3A CN105550709B (en) | 2015-12-14 | 2015-12-14 | A kind of remote sensing image power transmission line corridor wood land extracting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510930084.3A CN105550709B (en) | 2015-12-14 | 2015-12-14 | A kind of remote sensing image power transmission line corridor wood land extracting method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105550709A CN105550709A (en) | 2016-05-04 |
| CN105550709B true CN105550709B (en) | 2019-01-29 |
Family
ID=55829892
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510930084.3A Expired - Fee Related CN105550709B (en) | 2015-12-14 | 2015-12-14 | A kind of remote sensing image power transmission line corridor wood land extracting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105550709B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778472A (en) * | 2016-11-17 | 2017-05-31 | 成都通甲优博科技有限责任公司 | The common invader object detection and recognition method in transmission of electricity corridor based on deep learning |
| CN106599865A (en) * | 2016-12-21 | 2017-04-26 | 四川华雁信息产业股份有限公司 | Disconnecting link state recognition device and method |
| CN106909924B (en) * | 2017-02-18 | 2020-08-28 | 北京工业大学 | Remote sensing image rapid retrieval method based on depth significance |
| CN107230205A (en) * | 2017-05-27 | 2017-10-03 | 国网上海市电力公司 | A kind of transmission line of electricity bolt detection method based on convolutional neural networks |
| CN107239802B (en) * | 2017-06-28 | 2021-06-01 | 广东工业大学 | Image classification method and device |
| CN108052966B (en) * | 2017-12-08 | 2021-02-09 | 重庆邮电大学 | Remote sensing image scene automatic extraction and classification method based on convolutional neural network |
| CN108830296B (en) * | 2018-05-18 | 2021-08-10 | 河海大学 | Improved high-resolution remote sensing image classification method based on deep learning |
| CN109726718B (en) * | 2019-01-03 | 2022-09-16 | 电子科技大学 | Visual scene graph generation system and method based on relation regularization |
| CN109816098B (en) * | 2019-01-25 | 2021-09-07 | 京东方科技集团股份有限公司 | Processing method and evaluation method of neural network, and data analysis method and device |
| CN109815914A (en) * | 2019-01-28 | 2019-05-28 | 成都蝉远科技有限公司 | A kind of convolutional neural networks model training method and system based on vegetation area identification |
| CN110263635B (en) * | 2019-05-14 | 2022-09-09 | 中国人民解放军火箭军工程大学 | Marker detection and identification method based on structural forest and PCANet |
| CN110427914B (en) * | 2019-08-14 | 2021-09-28 | 西南交通大学 | Power transmission corridor vegetation threat early warning method based on satellite remote sensing |
| CN111461258B (en) * | 2020-04-26 | 2023-04-18 | 武汉大学 | Remote sensing image scene classification method of coupling convolution neural network and graph convolution network |
| CN112347970B (en) * | 2020-11-18 | 2024-04-05 | 江苏海洋大学 | Remote sensing image ground object identification method based on graph convolution neural network |
| CN113033386B (en) * | 2021-03-23 | 2022-12-16 | 广东电网有限责任公司广州供电局 | High-resolution remote sensing image-based transmission line channel hidden danger identification method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544502A (en) * | 2013-10-29 | 2014-01-29 | 上海市城市建设设计研究总院 | High-resolution remote-sensing image ship extraction method based on SVM |
| CN103942564A (en) * | 2014-04-08 | 2014-07-23 | 武汉大学 | High-resolution remote sensing image scene classifying method based on unsupervised feature learning |
| CN104036293A (en) * | 2014-06-13 | 2014-09-10 | 武汉大学 | Rapid binary encoding based high resolution remote sensing image scene classification method |
| CN104732215A (en) * | 2015-03-25 | 2015-06-24 | 广西大学 | Remote-sensing image coastline extracting method based on information vector machine |
-
2015
- 2015-12-14 CN CN201510930084.3A patent/CN105550709B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544502A (en) * | 2013-10-29 | 2014-01-29 | 上海市城市建设设计研究总院 | High-resolution remote-sensing image ship extraction method based on SVM |
| CN103942564A (en) * | 2014-04-08 | 2014-07-23 | 武汉大学 | High-resolution remote sensing image scene classifying method based on unsupervised feature learning |
| CN104036293A (en) * | 2014-06-13 | 2014-09-10 | 武汉大学 | Rapid binary encoding based high resolution remote sensing image scene classification method |
| CN104732215A (en) * | 2015-03-25 | 2015-06-24 | 广西大学 | Remote-sensing image coastline extracting method based on information vector machine |
Non-Patent Citations (2)
| Title |
|---|
| Transferring Deep Convolutional Neural Networks for the Scene Classification of High-Resolution Remote Sensing Imagery;Fan Hu等;《Directory of Open Access Journals》;20151105;第7卷(第11期);第14680-14707页 |
| 基于高分辨率遥感影像的输电线走廊场景分类;龚浩 等;《武汉大学学报（工学版）》;20141031;第47卷(第5期);第712-716页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105550709A (en) | 2016-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105550709B (en) | A kind of remote sensing image power transmission line corridor wood land extracting method | |
| Jia et al. | A semisupervised Siamese network for hyperspectral image classification | |
| Liao et al. | A review of graph neural networks and their applications in power systems | |
| Gandhi et al. | Plant disease detection using CNNs and GANs as an augmentative approach | |
| CN111639544B (en) | Expression recognition method based on multi-branch cross-connection convolutional neural network | |
| CN105787501B (en) | Power transmission line corridor region automatically selects the vegetation classification method of feature | |
| Deng et al. | Multiple diseases and pests detection based on federated learning and improved faster R-CNN | |
| CN106503687A (en) | The monitor video system for identifying figures of fusion face multi-angle feature and its method | |
| CN109558806A (en) | The detection method and system of high score Remote Sensing Imagery Change | |
| CN109034184A (en) | A kind of grading ring detection recognition method based on deep learning | |
| CN109255469A (en) | Merge the Forecasting Flood method of stack self-encoding encoder and support vector regression | |
| CN107092935A (en) | A kind of assets alteration detection method | |
| Yu et al. | Convolutional neural network with feature reconstruction for monitoring mismatched photovoltaic systems | |
| Ye et al. | Vegetable pest image recognition method based on improved VGG convolution neural network | |
| Dhanalakshmi et al. | [Retracted] Application of Machine Learning in Multi‐Directional Model to Follow Solar Energy Using Photo Sensor Matrix | |
| Bennani et al. | WITS 2020: Proceedings of the 6th International Conference on Wireless Technologies, Embedded, and Intelligent Systems | |
| Yadikar et al. | A Review of Knowledge Distillation in Object Detection | |
| CN114359578A (en) | Application method and system of pest and disease damage identification intelligent terminal | |
| Guo et al. | Varied channels region proposal and classification network for wildlife image classification under complex environment | |
| Sharma et al. | Self-attention vision transformer with transfer learning for efficient crops and weeds classification | |
| Hu et al. | Learning salient features for flower classification using convolutional neural network | |
| CN117392547A (en) | Remote sensing image scene classification method integrating Swin transducer and visual map neural network | |
| Yang et al. | Extraction of land covers from remote sensing images based on a deep learning model of NDVI-RSU-Net | |
| Jiang et al. | Multi-feature deep learning for face gender recognition | |
| Xiong et al. | MLP-Pose: Human pose estimation by MLP-mixer |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190129 Termination date: 20211214 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |