CN109446992A - Remote sensing image building extracting method and system, storage medium, electronic equipment based on deep learning - Google Patents
Remote sensing image building extracting method and system, storage medium, electronic equipment based on deep learning Download PDFInfo
- Publication number
- CN109446992A CN109446992A CN201811279197.1A CN201811279197A CN109446992A CN 109446992 A CN109446992 A CN 109446992A CN 201811279197 A CN201811279197 A CN 201811279197A CN 109446992 A CN109446992 A CN 109446992A
- Authority
- CN
- China
- Prior art keywords
- building
- spot
- remote sensing
- sensing image
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention provides the remote sensing image building extracting method based on deep learning, including the production of step sample, model training, accuracy assessment, building prediction, merging and vector quantization.The invention further relates to the remote sensing image building extraction system based on deep learning, storage medium, electronic equipments；The present invention is based on improved RCF boundary constraint model to city single building profile, the isolated building profile in rural area and the intensive group's peripheral boundary of farm building extract, U-Net semantic segmentation prototype network structure is improved simultaneously, Pixel-level classification is carried out to image using U-Net is improved, finally two kinds of models are merged, by largely building sample label data training deep learning model, the network model merged with improved U-Net with RCF extracts the building on No. 2 remote sensing images of the domestic high score of sub-meter grade, realize that automatical and efficient building vector data extracts, the greatly time cost and human cost of reduction manual drawing.
Description
Technical field
The invention belongs to technical field of remote sensing image processing, are a kind of high spatials based on deep learning boundary constraint algorithm
The method that resolution remote sense image building extracts, is mainly used in the automatic building extraction of sub-meter grade remote sensing image.
Background technique
Building is the part that is easiest to increase in geographical data bank and change, and need most update.Due to building
Object is built for the importance of urban construction, generalized information system update, digitalized city and military surveillance etc., rapidly extracting is built
Build object information technology and carry out building variation detection urban development planning, electronic information, in terms of have it is important
Application.The extraction of artificial structure's information is a complicated process in remote sensing images, necessary not only for the automatic of computer
Identification, it is also desirable to which the auxiliary of people is completed, and the efficiency for causing current building to extract is lower.
In recent years, with the fast development of the significant increase of computer performance and deep learning, convolutional neural networks are answered
Field constantly expands, and is graduallyd mature using the automation extracting method that convolutional neural networks carry out atural object, and leads in remote sensing
Domain achieves larger effect.Due to the raising of sensor accuracy, the remote sensing image resolution ratio that can be utilized is also higher and higher, convolution
The feature that neural network is extracted is also more and more abundant.So, high score is extracted as unit of building using convolutional neural networks
There has also been theoretical basis for the multiple types building for distinguishing on remote sensing image.
It is only capable of extracting more rule currently based on the technical method of remote sensing image Building extraction and feature is significantly built, lead to
Poor with property, when building is more intensive, extraction effect is general, therefore how rapidly and accurately to extract all kinds of complex buildings and be
The committed step of remote sensing information process.
The architectural vector that Building extraction method obtains and the practical building that image is showed are carried out using deep learning at present
Plot gap is larger, it is difficult to be applied in land use survey.It is badly in need of a kind of novel remote sensing image building extraction side at present
Method.
Summary of the invention
For overcome the deficiencies in the prior art, a kind of remote sensing image building based on deep learning proposed by the present invention mentions
Method is taken, the network model merged using improved U-Net with RCF is not only able to quickly detect on remote sensing image various
The building of classification, additionally it is possible to operative constraint be carried out to the building boundary of extraction, guarantee that extract building boundary builds with practical image
Build the consistency on boundary.
The present invention provides the remote sensing image building extracting method based on deep learning, comprising the following steps:
S1, sample production, acquire remote sensing image, and the face arrow for cutting and obtaining comprising object of classification is split to remote sensing image
File is measured, and draws the building label in the face vector file, by the arrow for the face vector sample that label is building
Data conversion is measured into raster data, obtains the building sample of rasterizing, wherein the building specimen types include city list
Body building, the isolated building in rural area, Rural Housing Land intensive building；
S2, model training adjust the network model parameter that improved U-Net is merged with RCF, to the building sample into
The model training that is merged based on improved U-Net with RCF of row obtains corresponding to the city single building, the rural area isolates and builds
It builds, three kinds of disaggregated models of the boundary constraint of the Rural Housing Land intensive building；
Remote sensing image data to be tested is inputted the disaggregated model and predicts test data, and calculated by S3, accuracy assessment
Detection the evaluation function MIoU and MPA of this test are jumped to if MIoU value or MPA are up to standard in next step, if MIoU value
Or MPA then adjusts network model parameter return step S1 that improved U-Net is merged with RCF in the presence of situation not up to standard and modifies sample
This repetitive exercise again；
S4, building prediction, using being evaluated three kinds of disaggregated models up to standard in step S3 to remote sensing image target area
Building prediction is carried out, three kinds of building data of rasterizing in remote sensing image are obtained；
Three kinds of building data of final rasterizing are overlapped to obtain overall building by S7, merging and vector quantization
Raster data, then overall building raster data is subjected to vector quantization, the remote sensing image building data extracted.
Further, it is further comprised the steps of: between the step S4 and the step S7
S5, Imaging processing, using image processing module to three kinds of buildings of the rasterizing predicted in step S4
Data carry out edge smoothing, reduce building boundary sawtooth.
Further, it is further comprised the steps of: between the step S4 and the step S7
S6, the processing of building figure spot, obtain figure spot decision threshold according to building classification, if the figure of corresponding building classification
Spot area be more than or equal to the figure spot decision threshold, then determine figure spot be effective figure spot, and carry out figure spot burn into opening and closing operation,
Filling is handled to repair figure spot；If the figure spot area of corresponding building classification is less than the figure spot decision threshold, figure spot is determined
For invalid figure spot, and the denoising of figure spot burn into is carried out to eliminate filtering figure spot.
Further, the building label includes line label, face label.
Further, the boundary of remote sensing image is extracted by RCF algorithm in step sl, and building point is carried out according to boundary
Class, and the boundary for extracting remote sensing image by RCF algorithm as constraint condition and is inputted into the building sample.
Further, the remote sensing image is three wave bands or four Band fusion images.
A kind of electronic equipment, comprising: processor；
Memory；And program, wherein described program is stored in the memory, and is configured to by processor
It executes, described program includes for executing the remote sensing image building extracting method based on deep learning.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
Remote sensing image building extracting method of the row based on deep learning.
Remote sensing image building extraction system based on deep learning, including sample make module, model training module, essence
Spend assessment module, building prediction module, image processing module；Wherein,
Described image processing module includes combined vector unit, rasterizing unit, and the combined vector unit is used for
The rasterizing data of remote sensing image are first merged and carry out vector quantization again, the rasterizing unit is used for the vector quantization of remote sensing image
Data rasterizing；
The sample production module is split cutting to remote sensing image and obtains comprising classification pair for acquiring remote sensing image
The face vector sample of elephant, and the label of face vector sample is drawn, by the vector number for the face vector sample that label is building
According to raster data is converted into, the building sample of rasterizing is obtained, wherein the building specimen types include that city monomer is built
It builds, the isolated building in rural area, Rural Housing Land intensive building；
The model training module is for adjusting the network model parameter that improved U-Net is merged with RCF, to building sample
This carries out the model training merged based on improved U-Net with RCF, obtains to Yingcheng City single building, the isolated building in rural area, agriculture
Three kinds of disaggregated models of the boundary constraint of village's house site intensive building；
The accuracy assessment module is used for remote sensing image data input disaggregated model prediction test data to be tested, and
Detection the evaluation function MIoU and MPA for calculating this test are jumped in next step if MIoU value or MPA are up to standard, if
MIoU value or MPA then adjust the network model parameter that improved U-Net is merged with RCF in the presence of situation not up to standard and return and modify sample
This repetitive exercise again；
The building prediction module is used for using through evaluating disaggregated model up to standard to distant in the accuracy assessment module
Feel image target region and carry out building prediction, obtains three kinds of building data of rasterizing in remote sensing image；
Three kinds of building data of final rasterizing are overlapped to obtain overall building by the combined vector unit
Object raster data, then overall building raster data is subjected to vector quantization, the remote sensing image building data extracted.
Further, the remote sensing image building extraction system based on deep learning further includes building plot recognition mould
Block, the building plot recognition module are used to obtain figure spot decision threshold according to building classification, if corresponding building classification
Figure spot area be more than or equal to the figure spot decision threshold, then determine figure spot be effective figure spot；If the figure of corresponding building classification
Spot area is less than the figure spot decision threshold, then determines that figure spot is invalid figure spot；
Described image processing module further includes Imaging processing unit, invalid figure spot processing unit, effective figure spot processing list
Member；The Imaging processing unit is for three kinds of building numbers to the rasterizing predicted in the building prediction module
According to edge smoothing is carried out, building boundary sawtooth is reduced；The invalid figure spot processing unit be used for be determined as invalid figure spot into
Filtering figure spot is eliminated in row figure spot burn into denoising；Effective figure spot processing unit be used for be determined as effective figure spot into
Row figure spot burn into opening and closing operation, filling processing are to repair figure spot.
Compared with prior art, the beneficial effects of the present invention are:
The present invention provides the remote sensing image building extracting method based on deep learning, including the production of step sample, model
Training, accuracy assessment, building prediction, merging and vector quantization.The invention further relates to the remote sensing image buildings based on deep learning
Object extraction system, storage medium, electronic equipment；The present invention uses the full convolutional neural networks of depth optimization, passes through improved model
The parameter optimizations networks such as network layer, pond size, while depth edge model is added to improve building contours extract effect.This hair
It is bright that city single building profile, the isolated building profile in rural area and rural area are built based on improved RCF boundary constraint model
It builds intensive group's peripheral boundary to extract, while improving U-Net semantic segmentation prototype network structure, using improvement U-Net to shadow
As carrying out Pixel-level classification, finally two kinds of models are merged, by largely building sample label data training depth
Model is practised, the network model merged with improved U-Net with RCF is to the building on No. 2 remote sensing images of the domestic high score of sub-meter grade
Extract, realize that automatical and efficient building vector data extracts, greatly reduce manual drawing time cost and manpower at
This.Compared to traditional culture plot extracting method, the present invention can effectively improve the production efficiency that building plot is extracted, together
When guarantee the consistency of the building form shown on building concentration form and remote sensing image, promote and answer convenient for remote sensing image processing
With.
Classification and form of the building on remote sensing image are more complicated, and not only color is different, but also different scale
It is very big, there are single house of smaller Rural Housing Land, the industrial building for also having occupied area very big.For the complexity of building
Convolutional layer information all in network is all used, explicitly makes using U-Net and RCF converged network model by feature
Characteristics of objects is sufficiently extracted with Multiscale Fusion, including building external contour feature and building interior textural characteristics,
The present invention can obtain more careful building as a result, and guaranteeing to extract the reliability for building result.
The above description is only an overview of the technical scheme of the present invention, in order to better understand the technical means of the present invention,
And can be implemented in accordance with the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention and the accompanying drawings.
A specific embodiment of the invention is shown in detail by following embodiment and its attached drawing.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present invention, constitutes part of this application, this hair
Bright illustrative embodiments and their description are used to explain the present invention, and are not constituted improper limitations of the present invention.In the accompanying drawings:
Fig. 1 is the remote sensing image building extracting method flow diagram of the invention based on deep learning；
Fig. 2 is the remote sensing image building extracting method logical schematic of the invention based on deep learning；
Fig. 3 is deep learning algorithm U-Net model schematic；
Fig. 4 is RCF algorithm model schematic diagram；
Fig. 5 is improved U-Net and RCF Fusion Model schematic diagram of the invention；
Fig. 6 is Fig. 5 with the building partial schematic diagram after the present invention；
Fig. 7 is the remote sensing image building extraction system schematic diagram of the invention based on deep learning.
Specific embodiment
In the following, being described further in conjunction with attached drawing and specific embodiment to the present invention, it should be noted that not
Under the premise of conflicting, new implementation can be formed between various embodiments described below or between each technical characteristic in any combination
Example.
Remote sensing image building extracting method based on deep learning, as shown in Figure 1 and Figure 2, comprising the following steps:
S1, sample production, acquire remote sensing image, and the face arrow for cutting and obtaining comprising object of classification is split to remote sensing image
File is measured, and draws the building label in the face vector file, by the arrow for the face vector sample that label is building
Data conversion is measured into raster data, obtains the building sample of rasterizing, wherein the building specimen types include city list
Body building, the isolated building in rural area, Rural Housing Land intensive building.
In one embodiment, No. 2 0.8 meter of resolution ratio, the 4 Band fusion images of high score for compiling research area, establish image
Library, at the same according to image establish corresponding architectural vector exemplar library (including city single building, the isolated building in rural area and
Rural Housing Land intensive building)；The format conversion that sample label is carried out according to building sample database, will build according to label vector
Sample is converted to contour line sample raster data and building sides sample raster data.For example, on No. two remote sensing images of high score
Cutting automatically for image sample block is carried out by the point vector file of selection, size is obtained and is 1000*1000 and includes classification pair
The face vector frame sample of elephant draws sample label, utilizes GDAL (Geospatial Data Abstraction Library) batch
Amount processing multiband sample data is reference format, wherein specimen types include city single building, the isolated building in rural area and
Rural Housing Land intensive building, every kind of sample include two class label of line and face；It should be noted that in the present embodiment preferably
Four Band fusion images, because further including near infrared image on the basis of comprising tri- wave band image of RGB in four Band fusion images,
It is big compared with data volume for three common wave band images first in the data volume of data sample, be conducive to sample training, to mention
High accuracy；Secondly, effectively quickly being filtered after capable of effectively filtering out green interference and the combination of tri- wave band of RGB because of near infrared image
The image information of greenbelt improves building extraction efficiency.In the present embodiment, the contour line of building passes through RCF (Richer
Convolutional Features) algorithm extracts the boundary of remote sensing image, and building sides sample raster data is by improving U-
Net carries out Pixel-level classification to remote sensing image, and in the network model for being input to deep learning as constraint using boundary.
S2, model training adjust the network model parameter that improved U-Net is merged with RCF, to the building sample into
The model training that is merged based on improved U-Net with RCF of row obtains corresponding to the city single building, the rural area isolates and builds
It builds, three kinds of disaggregated models of the boundary constraint of the Rural Housing Land intensive building.
As shown in figure 3, being deep learning algorithm U-Net modular concept, U-Net is on increasing on the basis of FCN network
What sample phase obtained has multiple dimensioned network model, is suitble to the convolutional neural networks of the big figure semantic segmentation of remote sensing, network knot
Structure is broadly divided into two parts as shown in Figure 3, and input side section is to extract the convolutional network of characteristics of image as FCN, output
Side section is to up-sample the deconvolution operation of part, this stage adds many feature channels, more original texture images is allowed to believe
Breath is propagated in high-resolution layer.It illustrates, U-Net does not have full articulamentum, and whole process uses valid
To carry out convolution, it can guarantee segmentation result all in this way and be based on obtained in the contextual feature not lacked, therefore defeated
The picture size for entering output can be inconsistent, the input for very big image, can carry out nothing using overlap-strategy
The image of seam exports.The output of part is extracted additionally, due to up-sampling part meeting fusion feature, this is actually one multiple dimensioned
The feature of fusion, object is the output from first convolution block, and also from the output of up-sampling, such connection runs through
Whole network has preferable extraction effect for the building with a variety of scales.
As shown in figure 4, being RCF algorithm model principle, RCF algorithm is carried out centainly on the basis of HED algorithm structure
It improves, wherein HED algorithm carries out fine-tune as model using VGG simultaneously using cascade structure, exports in VGG 5
The output of stage is merged, but HED is only exported using the convolution feature of the last layer in each stage, and RCF is then utilized
The feature of all convolutional layers in each stage, and introduce new loss function loss/sigmoid.
As shown in figure 5, being the rear modified hydrothermal process mould in conjunction with U-Net of boundary constraint model, that is, RCF based on deep learning
Type principle；RCF algorithm principle is run through into entire U-Net network, uses 1*1's behind each convolutional layer of U-Net network
Convolution kernel carries out the information integration of liter dimension or dimensionality reduction and interchannel, and merges to the characteristic layer in each stage, to each
Fused layer calculates loss, and all edge loss are weighted and averaged, and is finally weighted again with the loss of classification flat
Obtain final loss, and final output result be edge module the output of the last layer convolution and categorization module last
Layer convolution output is fused as a result, boundary characteristic is made full use of to be constrained.
In one embodiment, the model of the boundary constraint based on deep learning is carried out according to three kinds of classification samples after conversion
Training, is adjusted the network number of plies pond size etc. in model, and carry out Fusion Features.For example, the figure that every kind is built
As raster data and corresponding two classes label input the improvement network of U-Net and RCF as shown in Figure 5 simultaneously, mould is fully considered
Type to the feature and building interior feature on boundary, realize classification while carry out boundary constraint, by adjusting parameter into
The capable training that iterates respectively obtains the disaggregated model of boundary constraint corresponding with three kinds of buildings.
Remote sensing image data to be tested is inputted the disaggregated model and predicts test data, and calculated by S3, accuracy assessment
Detection the evaluation function MIoU and MPA of this test are jumped to if MIoU value or MPA are up to standard in next step, if MIoU value
Or MPA then adjusts network model parameter return step S1 that improved U-Net is merged with RCF in the presence of situation not up to standard and modifies sample
This repetitive exercise again；For example, setting MIoU value range, between 0.80-1, setting MPA value range is regarded as reaching between 0.85-1
Mark then adjusts the network model parameter that improved U-Net is merged with RCF and returns to re -training in S1 when going beyond the scope.
S4, building prediction, using being evaluated three kinds of disaggregated models up to standard in step S3 to remote sensing image target area
Building prediction is carried out, three kinds of building data of rasterizing in remote sensing image are obtained；For example, utilizing improved U-Net and RCF
The network model of fusion carries out building extraction to No. 2 remote sensing images of remote sensing image target area high score, fully considers model pair
The feature and building interior feature on boundary realize and carry out boundary constraint while classification.
Three kinds of building data of final rasterizing are overlapped to obtain overall building by S7, merging and vector quantization
Raster data, then overall building raster data is subjected to vector quantization, the remote sensing image building data extracted, to obtain
To applicable as a result, last carry out topological inspection, marginal check isovector to the remote sensing image building data of vector quantization again
Change operation；For example, building result schematic diagram shown in Fig. 6 to extract Nanjing Qixia District urban parts, wherein black silhouette is to build
Build boundary.By the verifying of practical artificial architecture production test, technical method of the invention is greatly improved in efficiency:
Sample area about 200 sq-km in trial zone shares building (group) 16231, single to extract time-consuming about 48 hours, using this
The technical method of invention extracts the used time 1 hour, artificial refine 1.5 hours, and production efficiency improves about 20 times.Therefore the present invention can be with
The efficiency of existing building extracting method is effectively improved under the premise of guaranteeing precision.
It should be noted that vector data structure can be specifically divided into point, line, surface, may be constructed various multiple in real world
Miscellaneous entity, when problem can be described as line or boundary, especially effectively.Vector data it is compact-sized, redundancy is low, and has
The topology information of spatial entities is easy definition and operates single spatial entities, is convenient for network analysis.The output quality of vector data
Good, precision height.But the complexity of vector data structure results in operation and the complication of algorithm, is based on line and side as one kind
The coding method on boundary cannot effectively support image algebraic operation, if do not can be effectively carried out the set operation of point set (as folded
Add), operation efficiency is low and complicated.Since the storage of vector data structure is more complicated, lead to the inquiry very expense of spatial entities
When, need it is point-by-point, by-line, by face inquire.The image data that vector data and grid indicate cannot directly operation (such as combine and look into
Inquiry and spatial analysis), vector sum grid conversion must be carried out when interactive.The interaction of vector data and DEM (digital elevation model)
It is to be realized by contour, cannot directly carries out joint space analysis with DEM.Raster data structure is by spatial point
Intensive and rule arrangement indicates whole spatial phenomenon.Its data structure is simple, and positioning access performance is good, can be with image
Joint space analysis is carried out with dem data, data sharing is easy to accomplish, is easier to the operation of raster data.Raster data
Data volume and mesh spacing square be inversely proportional, the cost of higher geometric accuracy is being significantly greatly increased for data volume.Because only
It using row and column as the station location marker of spatial entities, therefore is difficult to obtain the topology information of spatial entities, it is difficult to carry out network
The operation such as analysis.Raster data structure is not entity-oriented, and various entities are often to be superimposed to reflect, thus
It is difficult to and separates.The identification of point entity is needed to use edge detection skill to the identification of line entity using matching technique
Art then needs using image classification technology the identification of face entity, these technologies are not only time-consuming, but also cannot be guaranteed completely correct.
By the above analysis as can be seen that the advantage and disadvantage of vector data structure and raster data structure be it is complementary, in order to effectively
Realize that the various functions (such as and the combination of remote sensing image data, effective spatial analysis etc.) in GIS need while using two kinds
Data structure, and in GIS realize two kinds of data structures efficient conversion.
In one embodiment, it is further comprised the steps of: between the step S4 and the step S7
S5, Imaging processing, using image processing module to three kinds of buildings of the rasterizing predicted in step S4
Data carry out edge smoothing, reduce building boundary sawtooth.In the present embodiment, the building profile and border of extraction is carried out first
Smooth and micronization processes.It should be appreciated that image procossing includes but is not limited to existing graphics algorithm such as edge-smoothing, refinement
Processing, in the present embodiment, post-processes prediction result using opencv image processing module.
In one embodiment, originally coherent building is caused to interrupt because easily occurring deviation in building extraction process,
It is further comprised the steps of: between the step S4 and the step S7
S6, the processing of building figure spot, obtain figure spot decision threshold according to building classification, if the figure of corresponding building classification
Spot area be more than or equal to the figure spot decision threshold, then determine figure spot be effective figure spot, and carry out figure spot burn into opening and closing operation,
Filling is handled to repair figure spot；If the figure spot area of corresponding building classification is less than the figure spot decision threshold, figure spot is determined
For invalid figure spot, and the denoising of figure spot burn into is carried out to eliminate filtering figure spot.In the present embodiment, what first judgement was extracted builds
Build figure spot size, building concentration and groups of building distinguished, by graphics algorithm realize groups of building hole repair with
And the denoising of building concentration；For example, figure spot is divided into effective figure spot and invalid figure spot, the standard of judgement is an empirical value, mainly
It is divided according to floor area of building size, since the mean size of three kinds of buildings is different, so the figure spot of different kinds of building object
Decision threshold is also different, when being determined as effective figure spot, will carry out a series of images operation, including but not limited to figure spot is rotten
Erosion, opening and closing operation, filling；It when being determined as invalid figure spot, is then eliminated by burn into denoising method, has filtered a part of mistake and mentioned
As a result.
A kind of electronic equipment, comprising: processor；Memory；And program, wherein described program is stored in the storage
It in device, and is configured to be executed by processor, described program includes for executing the remote sensing image building based on deep learning
Object proposes method.A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
Remote sensing image building extracting method of the row based on deep learning.
Remote sensing image building extraction system based on deep learning, as shown in fig. 7, comprises sample makes module, model
Training module, accuracy assessment module, building prediction module, image processing module；Wherein,
Described image processing module includes combined vector unit, rasterizing unit, and the combined vector unit is used for
The rasterizing data of remote sensing image are first merged and carry out vector quantization again, the rasterizing unit is used for the vector quantization of remote sensing image
Data rasterizing；
The sample production module is split cutting to remote sensing image and obtains comprising classification pair for acquiring remote sensing image
The face vector sample of elephant, and the label of face vector sample is drawn, by the vector number for the face vector sample that label is building
According to raster data is converted into, the building sample of rasterizing is obtained, wherein the building specimen types include that city monomer is built
It builds, the isolated building in rural area, Rural Housing Land intensive building；
The model training module is for adjusting the network model parameter that improved U-Net is merged with RCF, to building sample
This carries out the model training merged based on improved U-Net with RCF, obtains to Yingcheng City single building, the isolated building in rural area, agriculture
Three kinds of disaggregated models of the boundary constraint of village's house site intensive building；
The accuracy assessment module is used for remote sensing image data input disaggregated model prediction test data to be tested, and
Detection the evaluation function MIoU and MPA for calculating this test are jumped in next step if MIoU value or MPA are up to standard, if
MIoU value or MPA then adjust the network model parameter that improved U-Net is merged with RCF in the presence of situation not up to standard and return and modify sample
This repetitive exercise again；
The building prediction module is used for using through evaluating disaggregated model up to standard to distant in the accuracy assessment module
Feel image target region and carry out building prediction, obtains three kinds of building data of rasterizing in remote sensing image；
Three kinds of building data of final rasterizing are overlapped to obtain overall building by the combined vector unit
Object raster data, then overall building raster data is subjected to vector quantization, the remote sensing image building data extracted.
Preferably, as shown in fig. 7, the remote sensing image building extraction system based on deep learning further includes building figure spot
Identification module, the building plot recognition module are used to obtain figure spot decision threshold according to building classification, if corresponding building
The other figure spot area of species is more than or equal to the figure spot decision threshold, then determines that figure spot is effective figure spot；If corresponding building species
Other figure spot area is less than the figure spot decision threshold, then determines that figure spot is invalid figure spot；
Described image processing module further includes Imaging processing unit, invalid figure spot processing unit, effective figure spot processing list
Member；The Imaging processing unit is for three kinds of building numbers to the rasterizing predicted in the building prediction module
According to edge smoothing is carried out, building boundary sawtooth is reduced；The invalid figure spot processing unit be used for be determined as invalid figure spot into
Filtering figure spot is eliminated in row figure spot burn into denoising；Effective figure spot processing unit be used for be determined as effective figure spot into
Row figure spot burn into opening and closing operation, filling processing are to repair figure spot.
The present invention provides a kind of remote sensing image building extracting method based on deep learning, using the full convolution of depth optimization
Neural network by parameter optimizations networks such as the network layer of improved model, pond sizes, while adding depth edge model to mention
High building contours extract effect.The present invention is based on improved RCF boundary constraint model to city single building profile, rural area
Isolated building profile and the intensive group's peripheral boundary of farm building extract, while improving U-Net semantic segmentation prototype network
Structure carries out Pixel-level classification to image using U-Net is improved, finally merges two kinds of models, by largely building
Sample label data train deep learning model, and the network model merged with improved U-Net with RCF is to the domestic high score of sub-meter grade
Building on No. 2 remote sensing images extracts, and realizes that automatical and efficient building vector data extracts, and greatly reduction is drawn by hand
The time cost and human cost of system.Compared to traditional culture plot extracting method, the present invention can effectively improve building
The production efficiency that plot is extracted, while guaranteeing the consistency of the building form shown on building concentration form and remote sensing image.
Classification and form of the building on remote sensing image are more complicated, and not only color is different, but also different scale
It is very big, there are single house of smaller Rural Housing Land, the industrial building for also having occupied area very big.For the complexity of building
Convolutional layer information all in network is all used, explicitly makes using U-Net and RCF converged network model by feature
Characteristics of objects is sufficiently extracted with Multiscale Fusion, including building external contour feature and building interior textural characteristics,
The present invention can obtain more careful building as a result, and guaranteeing to extract the reliability for building result.
More than, only presently preferred embodiments of the present invention is not intended to limit the present invention in any form；All current rows
The those of ordinary skill of industry can be shown in by specification attached drawing and above and swimmingly implement the present invention；But all to be familiar with sheet special
The technical staff of industry without departing from the scope of the present invention, is made a little using disclosed above technology contents
The equivalent variations of variation, modification and evolution is equivalent embodiment of the invention；Meanwhile all substantial technologicals according to the present invention
The variation, modification and evolution etc. of any equivalent variations to the above embodiments, still fall within technical solution of the present invention
Within protection scope.
Claims (10)
1. the remote sensing image building extracting method based on deep learning, which comprises the following steps:
S1, sample production, acquire remote sensing image, and the face vector text for cutting and obtaining comprising object of classification is split to remote sensing image
Part, and the building label in the face vector file is drawn, by the vector number for the face vector sample that label is building
According to raster data is converted into, the building sample of rasterizing is obtained, wherein the building specimen types include that city monomer is built
It builds, the isolated building in rural area, Rural Housing Land intensive building；
S2, model training adjust the network model parameter that improved U-Net is merged with RCF, carry out base to the building sample
In the model training that improved U-Net is merged with RCF, obtain corresponding to the city single building, the isolated building in the rural area, institute
State three kinds of disaggregated models of the boundary constraint of Rural Housing Land intensive building；
Remote sensing image data to be tested is inputted the disaggregated model and predicts test data, and calculates this by S3, accuracy assessment
Detection the evaluation function MIoU and MPA of test are jumped to if MIoU value or MPA are up to standard in next step, if MIoU value or MPA
Network model parameter return step S1 that improved U-Net is merged with RCF is then adjusted in the presence of situation not up to standard and modifies sample weight
New repetitive exercise；
S4, building prediction, carry out remote sensing image target area using three kinds of disaggregated models up to standard are evaluated in step S3
Building prediction, obtains three kinds of building data of rasterizing in remote sensing image；
Three kinds of building data of final rasterizing are overlapped to obtain overall building grid by S7, merging and vector quantization
Data, then overall building raster data is subjected to vector quantization, the remote sensing image building data extracted.
2. the remote sensing image building extracting method based on deep learning as described in claim 1, which is characterized in that step S4
It is further comprised the steps of: between step S7
S5, Imaging processing, using image processing module to three kinds of building data of the rasterizing predicted in step S4
Edge smoothing is carried out, building boundary sawtooth is reduced.
3. the remote sensing image building extracting method based on deep learning as claimed in claim 1 or 2, which is characterized in that step
It is further comprised the steps of: between rapid S4 and step S7
S6, the processing of building figure spot, obtain figure spot decision threshold according to building classification, if the figure spot face of corresponding building classification
Product is more than or equal to the figure spot decision threshold, then determines that figure spot is effective figure spot, and carry out figure spot burn into opening and closing operation, filling
Processing is to repair figure spot；If the figure spot area of corresponding building classification is less than the figure spot decision threshold, determine figure spot for nothing
Figure spot is imitated, and carries out the denoising of figure spot burn into eliminate filtering figure spot.
4. the remote sensing image building extracting method based on deep learning as claimed in claim 3, it is characterised in that: described to build
Building object label includes line label, face label.
5. the remote sensing image building extracting method based on deep learning as claimed in claim 3, it is characterised in that: in step
The boundary of remote sensing image is extracted in S1 by RCF algorithm, and building classifications are carried out according to boundary, and will extract by RCF algorithm
The boundary of remote sensing image is as constraint condition and inputs the building sample.
6. the remote sensing image building extracting method based on deep learning as described in claim 4 or 5, it is characterised in that: institute
Stating remote sensing image is three wave bands or four Band fusion images.
7. a kind of electronic equipment, characterized by comprising: processor；
Memory；And program, wherein described program is stored in the memory, and is configured to be held by processor
Row, described program includes for executing the method as described in claim 1.
8. a kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that: the computer program quilt
Processor executes the method as described in claim 1.
9. the remote sensing image building extraction system based on deep learning, it is characterised in that: including sample production module, model instruction
Practice module, accuracy assessment module, building prediction module, image processing module；Wherein,
Described image processing module includes combined vector unit, rasterizing unit, and the combined vector unit is used for will be distant
The rasterizing data of sense image first merge carries out vector quantization again, and the rasterizing unit is used for the vector quantization data of remote sensing image
Rasterizing；
The sample production module is split cutting to remote sensing image and obtains comprising object of classification for acquiring remote sensing image
Face vector sample, and the label of face vector sample is drawn, the vector data for the face vector sample that label is building is turned
Change raster data into, obtain the building sample of rasterizing, wherein the building specimen types include city single building,
The isolated building in rural area, Rural Housing Land intensive building；
The model training module for adjusting the network model parameter that improved U-Net is merged with RCF, to building sample into
The model training that row is merged based on improved U-Net with RCF is obtained to Yingcheng City single building, the isolated building in rural area, rural area residence
Three kinds of disaggregated models of the boundary constraint of base intensive building；
The accuracy assessment module is used to remote sensing image data input disaggregated model to be tested predicting test data, and calculates
Detection the evaluation function MIoU and MPA of this test are jumped to if MIoU value or MPA are up to standard in next step, if MIoU value
Or MPA then adjusts the network model parameter that improved U-Net is merged with RCF in the presence of situation not up to standard and returns and modify sample again
Repetitive exercise；
The building prediction module is used for using through evaluating disaggregated model up to standard to remote sensing shadow in the accuracy assessment module
As target area progress building prediction, three kinds of building data of rasterizing in remote sensing image are obtained；
Three kinds of building data of final rasterizing are overlapped to obtain overall building grid by the combined vector unit
Lattice data, then overall building raster data is subjected to vector quantization, the remote sensing image building data extracted.
10. the remote sensing image building extraction system based on deep learning as claimed in claim 9, it is characterised in that: also wrap
Building plot recognition module is included, the building plot recognition module is used to obtain figure spot decision threshold according to building classification
Value determines that figure spot is effective figure spot if the figure spot area of corresponding building classification is more than or equal to the figure spot decision threshold；If
The figure spot area of corresponding building classification is less than the figure spot decision threshold, then determines that figure spot is invalid figure spot；
Described image processing module further includes Imaging processing unit, invalid figure spot processing unit, effective figure spot processing unit；Institute
State Imaging processing unit for three kinds of building data of the rasterizing predicted in the building prediction module into
Row bound is smooth, reduces building boundary sawtooth；The invalid figure spot processing unit is used for being determined as that invalid figure spot carries out figure
Filtering figure spot is eliminated in the denoising of spot burn into；Effective figure spot processing unit is used for being determined as that effective figure spot carries out figure
Spot burn into opening and closing operation, filling processing are to repair figure spot.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811279197.1A CN109446992B (en) | 2018-10-30 | 2018-10-30 | Remote sensing image building extraction method and system based on deep learning, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811279197.1A CN109446992B (en) | 2018-10-30 | 2018-10-30 | Remote sensing image building extraction method and system based on deep learning, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109446992A true CN109446992A (en) | 2019-03-08 |
| CN109446992B CN109446992B (en) | 2022-06-17 |
Family
ID=65549932
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811279197.1A Active CN109446992B (en) | 2018-10-30 | 2018-10-30 | Remote sensing image building extraction method and system based on deep learning, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446992B (en) |
Cited By (49)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110175548A (en) * | 2019-05-20 | 2019-08-27 | 中国科学院光电技术研究所 | Remote sensing images building extracting method based on attention mechanism and channel information |
| CN110197147A (en) * | 2019-05-23 | 2019-09-03 | 星际空间（天津）科技发展有限公司 | Building Cass collection method, apparatus, storage medium and the equipment of remote sensing image |
| CN110298348A (en) * | 2019-06-12 | 2019-10-01 | 苏州中科天启遥感科技有限公司 | Remote sensing image building sample areas extracting method and system, storage medium, equipment |
| CN110516539A (en) * | 2019-07-17 | 2019-11-29 | 苏州中科天启遥感科技有限公司 | Remote sensing image building extracting method, system, storage medium and equipment based on confrontation network |
| CN110532894A (en) * | 2019-08-05 | 2019-12-03 | 西安电子科技大学 | Remote sensing target detection method based on boundary constraint CenterNet |
| CN110580443A (en) * | 2019-06-19 | 2019-12-17 | 深圳大学 | Low-altitude near-real-time building earthquake damage assessment method |
| CN110675408A (en) * | 2019-09-19 | 2020-01-10 | 成都数之联科技有限公司 | High-resolution image building extraction method and system based on deep learning |
| CN110674735A (en) * | 2019-09-23 | 2020-01-10 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
| CN110717420A (en) * | 2019-09-25 | 2020-01-21 | 中国科学院深圳先进技术研究院 | Cultivated land extraction method and system based on remote sensing image and electronic equipment |
| CN110738133A (en) * | 2019-09-23 | 2020-01-31 | 中科禾信遥感科技（苏州）有限公司 | Method and device for recognizing contour boundaries of images of different agricultural facilities |
| CN110956196A (en) * | 2019-10-11 | 2020-04-03 | 东南大学 | Automatic recognition method for window-wall ratio of urban building |
| CN110992257A (en) * | 2019-12-20 | 2020-04-10 | 北京航天泰坦科技股份有限公司 | Remote sensing image sensitive information automatic shielding method and device based on deep learning |
| CN111144335A (en) * | 2019-12-30 | 2020-05-12 | 自然资源部国土卫星遥感应用中心 | Method and device for building deep learning model |
| CN111191500A (en) * | 2019-11-12 | 2020-05-22 | 广东融合通信股份有限公司 | Photovoltaic roof resource identification method based on deep learning image segmentation |
| CN111242028A (en) * | 2020-01-13 | 2020-06-05 | 北京工业大学 | Remote sensing image ground object segmentation method based on U-Net |
| CN111339947A (en) * | 2020-02-26 | 2020-06-26 | 苏州中科天启遥感科技有限公司 | Method and system for extracting remote sensing image fuzzy boundary ground object, storage medium and equipment |
| CN111368843A (en) * | 2020-03-06 | 2020-07-03 | 电子科技大学 | Method for extracting lake on ice based on semantic segmentation |
| CN111414878A (en) * | 2020-03-26 | 2020-07-14 | 京东城市(北京)数字科技有限公司 | Method and device for social attribute analysis and image processing of land parcel |
| CN111476308A (en) * | 2020-04-10 | 2020-07-31 | 高崟 | Remote sensing image classification method and device based on prior geometric constraint and electronic equipment |
| CN111489387A (en) * | 2020-04-09 | 2020-08-04 | 湖南盛鼎科技发展有限责任公司 | Remote sensing image building area calculation method |
| CN111553272A (en) * | 2020-04-28 | 2020-08-18 | 上海市测绘院 | High-resolution satellite optical remote sensing image building change detection method based on deep learning |
| CN111652250A (en) * | 2020-06-09 | 2020-09-11 | 星际空间（天津）科技发展有限公司 | Remote sensing image building extraction method and device based on polygon and storage medium |
| CN111652251A (en) * | 2020-06-09 | 2020-09-11 | 星际空间（天津）科技发展有限公司 | Method and device for building remote sensing image building feature extraction model and storage medium |
| CN111709346A (en) * | 2020-06-10 | 2020-09-25 | 嘉应学院 | Historical building identification and detection method based on deep learning and high-resolution images |
| CN111832502A (en) * | 2020-07-20 | 2020-10-27 | 中国人民解放军战略支援部队航天工程大学 | Remote sensing image visual salient region intelligent search method for satellite in-orbit application |
| CN111914611A (en) * | 2020-05-09 | 2020-11-10 | 中国科学院空天信息创新研究院 | Urban green space high-resolution remote sensing monitoring method and system |
| CN111914328A (en) * | 2020-07-31 | 2020-11-10 | 同济大学 | Building plane aided design method and system based on artificial intelligence |
| CN112052783A (en) * | 2020-09-02 | 2020-12-08 | 中南大学 | High-resolution image weak supervision building extraction method combining pixel semantic association and boundary attention |
| CN112149594A (en) * | 2020-09-29 | 2020-12-29 | 同济大学 | Urban construction assessment method based on deep learning and high-resolution satellite images |
| CN112182125A (en) * | 2020-09-14 | 2021-01-05 | 中国科学院重庆绿色智能技术研究院 | Business gathering area boundary identification system |
| CN112232102A (en) * | 2019-07-15 | 2021-01-15 | 中国司法大数据研究院有限公司 | Building target identification method and system based on deep neural network and multitask learning |
| CN112288331A (en) * | 2020-11-24 | 2021-01-29 | 哈尔滨工业大学 | Building reconstruction evaluation system based on semantic recognition |
| CN112580510A (en) * | 2020-12-18 | 2021-03-30 | 广州市城市规划设计所 | Permeable ground area rate estimation method, permeable ground area rate estimation device, permeable ground area rate estimation equipment and storage medium |
| CN112990001A (en) * | 2021-03-11 | 2021-06-18 | 中国电建集团华东勘测设计研究院有限公司 | Remote sensing image monomer building rapid and accurate extraction method applied to digital collection |
| CN113486135A (en) * | 2021-07-27 | 2021-10-08 | 中南大学 | Building comprehensive method based on deep learning network |
| CN113505842A (en) * | 2021-07-21 | 2021-10-15 | 中国科学院空天信息创新研究院 | Automatic extraction method suitable for large-scale regional remote sensing image urban building |
| CN113516135A (en) * | 2021-06-23 | 2021-10-19 | 江苏师范大学 | Remote sensing image building extraction and contour optimization method based on deep learning |
| CN113610013A (en) * | 2021-08-10 | 2021-11-05 | 四川易方智慧科技有限公司 | Method for extracting building outline based on RGB (Red Green blue) wave bands of high-definition remote sensing image |
| CN113807301A (en) * | 2021-09-26 | 2021-12-17 | 武汉汉达瑞科技有限公司 | Automatic extraction method and automatic extraction system for newly-added construction land |
| CN114049569A (en) * | 2022-01-13 | 2022-02-15 | 自然资源部第三地理信息制图院 | Deep learning model performance evaluation method and system |
| CN114092813A (en) * | 2021-11-25 | 2022-02-25 | 中国科学院空天信息创新研究院 | Industrial park image extraction method, model, electronic equipment and storage medium |
| CN114331031A (en) * | 2021-12-08 | 2022-04-12 | 北京华清安地建筑设计有限公司 | Building traditional feature recognition and evaluation method and system |
| CN114638769A (en) * | 2020-12-15 | 2022-06-17 | 武汉市探测者科技有限公司 | 2.5-dimensional map building hot zone automatic extraction method based on image processing algorithm |
| CN114677589A (en) * | 2022-03-18 | 2022-06-28 | 广州市城市规划勘测设计研究院 | City functional area identification method, device, equipment and medium based on remote sensing interpretation |
| CN114677522A (en) * | 2022-03-17 | 2022-06-28 | 北京科技大学 | Building structure type prediction method and system based on building pattern spots and deep learning |
| CN115841625A (en) * | 2023-02-23 | 2023-03-24 | 杭州电子科技大学 | Remote sensing building image extraction method based on improved U-Net model |
| CN115994987A (en) * | 2023-03-21 | 2023-04-21 | 天津市勘察设计院集团有限公司 | Rural building extraction and vectorization method based on inclined three-dimensional model |
| CN116052019A (en) * | 2023-03-31 | 2023-05-02 | 深圳市规划和自然资源数据管理中心(深圳市空间地理信息中心) | High-quality detection method suitable for built-up area of large-area high-resolution satellite image |
| CN117218485A (en) * | 2023-09-05 | 2023-12-12 | 安徽省第二测绘院 | Deep learning model-based multi-source remote sensing image interpretation sample library creation method |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004922A (en) * | 2010-12-01 | 2011-04-06 | 南京大学 | High-resolution remote sensing image plane extraction method based on skeleton characteristic |
| CN104504718A (en) * | 2015-01-06 | 2015-04-08 | 南京大学 | High-definition aerial remote sensing data automatic road extraction method |
| US20160171682A1 (en) * | 2014-12-14 | 2016-06-16 | International Business Machines Corporation | Cloud-based infrastructure for feedback-driven training and image recognition |
| US20170076438A1 (en) * | 2015-08-31 | 2017-03-16 | Cape Analytics, Inc. | Systems and methods for analyzing remote sensing imagery |
| CN107092877A (en) * | 2017-04-12 | 2017-08-25 | 武汉大学 | Remote sensing image roof contour extracting method based on basement bottom of the building vector |
| CN107247938A (en) * | 2017-06-08 | 2017-10-13 | 中国科学院遥感与数字地球研究所 | A kind of method of high-resolution remote sensing image City Building function classification |
| CN107767382A (en) * | 2017-09-26 | 2018-03-06 | 武汉市国土资源和规划信息中心 | The extraction method and system of static three-dimensional map contour of building line |
| CN107967454A (en) * | 2017-11-24 | 2018-04-27 | 武汉理工大学 | Take the two-way convolutional neural networks Classification in Remote Sensing Image method of spatial neighborhood relation into account |
| CN108198181A (en) * | 2018-01-23 | 2018-06-22 | 电子科技大学 | A kind of thermal-induced imagery processing method based on region segmentation and image co-registration |
| CN108537192A (en) * | 2018-04-17 | 2018-09-14 | 福州大学 | A kind of remote sensing image ground mulching sorting technique based on full convolutional network |
-
2018
- 2018-10-30 CN CN201811279197.1A patent/CN109446992B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004922A (en) * | 2010-12-01 | 2011-04-06 | 南京大学 | High-resolution remote sensing image plane extraction method based on skeleton characteristic |
| US20160171682A1 (en) * | 2014-12-14 | 2016-06-16 | International Business Machines Corporation | Cloud-based infrastructure for feedback-driven training and image recognition |
| CN104504718A (en) * | 2015-01-06 | 2015-04-08 | 南京大学 | High-definition aerial remote sensing data automatic road extraction method |
| US20170076438A1 (en) * | 2015-08-31 | 2017-03-16 | Cape Analytics, Inc. | Systems and methods for analyzing remote sensing imagery |
| CN107092877A (en) * | 2017-04-12 | 2017-08-25 | 武汉大学 | Remote sensing image roof contour extracting method based on basement bottom of the building vector |
| CN107247938A (en) * | 2017-06-08 | 2017-10-13 | 中国科学院遥感与数字地球研究所 | A kind of method of high-resolution remote sensing image City Building function classification |
| CN107767382A (en) * | 2017-09-26 | 2018-03-06 | 武汉市国土资源和规划信息中心 | The extraction method and system of static three-dimensional map contour of building line |
| CN107967454A (en) * | 2017-11-24 | 2018-04-27 | 武汉理工大学 | Take the two-way convolutional neural networks Classification in Remote Sensing Image method of spatial neighborhood relation into account |
| CN108198181A (en) * | 2018-01-23 | 2018-06-22 | 电子科技大学 | A kind of thermal-induced imagery processing method based on region segmentation and image co-registration |
| CN108537192A (en) * | 2018-04-17 | 2018-09-14 | 福州大学 | A kind of remote sensing image ground mulching sorting technique based on full convolutional network |
Non-Patent Citations (4)
| Title |
|---|
| GUANGMING WU等: "A Boundary Regulated Network for Accurate Roof Segmentation and Outline Extraction", 《REMOTE SENING》 * |
| ZHAOLI HONG等: "Road Extraction From a High Spatial Resolution Remote Sensing Image Based on Richer Convolutional Features", 《IEEE ACCESS》 * |
| 伍光明等: "基于Ｕ型卷积神经网络的航空影像建筑物检测", 《测绘学报》 * |
| 许慧敏: "基于深度学习U-net模型的高分辨率遥感影像分类方法研究", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
Cited By (75)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110175548A (en) * | 2019-05-20 | 2019-08-27 | 中国科学院光电技术研究所 | Remote sensing images building extracting method based on attention mechanism and channel information |
| CN110175548B (en) * | 2019-05-20 | 2022-08-23 | 中国科学院光电技术研究所 | Remote sensing image building extraction method based on attention mechanism and channel information |
| CN110197147A (en) * | 2019-05-23 | 2019-09-03 | 星际空间（天津）科技发展有限公司 | Building Cass collection method, apparatus, storage medium and the equipment of remote sensing image |
| CN110197147B (en) * | 2019-05-23 | 2022-12-02 | 星际空间（天津）科技发展有限公司 | Building example extraction method, device, storage medium and equipment of remote sensing image |
| CN110298348B (en) * | 2019-06-12 | 2020-04-28 | 苏州中科天启遥感科技有限公司 | Method and system for extracting remote sensing image building sample region, storage medium and equipment |
| CN110298348A (en) * | 2019-06-12 | 2019-10-01 | 苏州中科天启遥感科技有限公司 | Remote sensing image building sample areas extracting method and system, storage medium, equipment |
| CN110580443B (en) * | 2019-06-19 | 2022-03-08 | 深圳大学 | Low-altitude near-real-time building earthquake damage assessment method |
| CN110580443A (en) * | 2019-06-19 | 2019-12-17 | 深圳大学 | Low-altitude near-real-time building earthquake damage assessment method |
| CN112232102A (en) * | 2019-07-15 | 2021-01-15 | 中国司法大数据研究院有限公司 | Building target identification method and system based on deep neural network and multitask learning |
| CN110516539A (en) * | 2019-07-17 | 2019-11-29 | 苏州中科天启遥感科技有限公司 | Remote sensing image building extracting method, system, storage medium and equipment based on confrontation network |
| CN110532894A (en) * | 2019-08-05 | 2019-12-03 | 西安电子科技大学 | Remote sensing target detection method based on boundary constraint CenterNet |
| CN110532894B (en) * | 2019-08-05 | 2021-09-03 | 西安电子科技大学 | Remote sensing target detection method based on boundary constraint CenterNet |
| CN110675408A (en) * | 2019-09-19 | 2020-01-10 | 成都数之联科技有限公司 | High-resolution image building extraction method and system based on deep learning |
| CN110674735B (en) * | 2019-09-23 | 2022-05-24 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
| CN110738133A (en) * | 2019-09-23 | 2020-01-31 | 中科禾信遥感科技（苏州）有限公司 | Method and device for recognizing contour boundaries of images of different agricultural facilities |
| CN110674735A (en) * | 2019-09-23 | 2020-01-10 | 中国科学院地理科学与资源研究所 | Agricultural facility remote sensing extraction method and device based on fine classification |
| CN110738133B (en) * | 2019-09-23 | 2023-10-27 | 中科禾信遥感科技（苏州）有限公司 | Method and device for identifying image contour boundaries of different agricultural facilities |
| CN110717420A (en) * | 2019-09-25 | 2020-01-21 | 中国科学院深圳先进技术研究院 | Cultivated land extraction method and system based on remote sensing image and electronic equipment |
| CN110956196A (en) * | 2019-10-11 | 2020-04-03 | 东南大学 | Automatic recognition method for window-wall ratio of urban building |
| CN110956196B (en) * | 2019-10-11 | 2024-03-08 | 东南大学 | Automatic recognition method for window wall ratio of urban building |
| CN111191500A (en) * | 2019-11-12 | 2020-05-22 | 广东融合通信股份有限公司 | Photovoltaic roof resource identification method based on deep learning image segmentation |
| CN110992257A (en) * | 2019-12-20 | 2020-04-10 | 北京航天泰坦科技股份有限公司 | Remote sensing image sensitive information automatic shielding method and device based on deep learning |
| CN111144335A (en) * | 2019-12-30 | 2020-05-12 | 自然资源部国土卫星遥感应用中心 | Method and device for building deep learning model |
| CN111242028A (en) * | 2020-01-13 | 2020-06-05 | 北京工业大学 | Remote sensing image ground object segmentation method based on U-Net |
| CN111339947A (en) * | 2020-02-26 | 2020-06-26 | 苏州中科天启遥感科技有限公司 | Method and system for extracting remote sensing image fuzzy boundary ground object, storage medium and equipment |
| CN111368843A (en) * | 2020-03-06 | 2020-07-03 | 电子科技大学 | Method for extracting lake on ice based on semantic segmentation |
| CN111368843B (en) * | 2020-03-06 | 2022-06-10 | 电子科技大学 | Method for extracting lake on ice based on semantic segmentation |
| CN111414878A (en) * | 2020-03-26 | 2020-07-14 | 京东城市(北京)数字科技有限公司 | Method and device for social attribute analysis and image processing of land parcel |
| CN111414878B (en) * | 2020-03-26 | 2023-11-03 | 北京京东智能城市大数据研究院 | Social attribute analysis and image processing method and device for land parcels |
| CN111489387A (en) * | 2020-04-09 | 2020-08-04 | 湖南盛鼎科技发展有限责任公司 | Remote sensing image building area calculation method |
| CN111489387B (en) * | 2020-04-09 | 2023-10-20 | 湖南盛鼎科技发展有限责任公司 | Remote sensing image building area calculation method |
| CN111476308A (en) * | 2020-04-10 | 2020-07-31 | 高崟 | Remote sensing image classification method and device based on prior geometric constraint and electronic equipment |
| CN111553272B (en) * | 2020-04-28 | 2022-05-06 | 上海市测绘院 | High-resolution satellite optical remote sensing image building change detection method based on deep learning |
| CN111553272A (en) * | 2020-04-28 | 2020-08-18 | 上海市测绘院 | High-resolution satellite optical remote sensing image building change detection method based on deep learning |
| CN111914611A (en) * | 2020-05-09 | 2020-11-10 | 中国科学院空天信息创新研究院 | Urban green space high-resolution remote sensing monitoring method and system |
| CN111652251B (en) * | 2020-06-09 | 2023-06-27 | 星际空间（天津）科技发展有限公司 | Remote sensing image building feature extraction model construction method, device and storage medium |
| CN111652250A (en) * | 2020-06-09 | 2020-09-11 | 星际空间（天津）科技发展有限公司 | Remote sensing image building extraction method and device based on polygon and storage medium |
| CN111652251A (en) * | 2020-06-09 | 2020-09-11 | 星际空间（天津）科技发展有限公司 | Method and device for building remote sensing image building feature extraction model and storage medium |
| CN111652250B (en) * | 2020-06-09 | 2023-05-26 | 星际空间（天津）科技发展有限公司 | Remote sensing image building extraction method and device based on polygons and storage medium |
| CN111709346A (en) * | 2020-06-10 | 2020-09-25 | 嘉应学院 | Historical building identification and detection method based on deep learning and high-resolution images |
| CN111832502A (en) * | 2020-07-20 | 2020-10-27 | 中国人民解放军战略支援部队航天工程大学 | Remote sensing image visual salient region intelligent search method for satellite in-orbit application |
| CN111914328B (en) * | 2020-07-31 | 2023-11-17 | 同济大学 | Building plane aided design method and system based on artificial intelligence |
| CN111914328A (en) * | 2020-07-31 | 2020-11-10 | 同济大学 | Building plane aided design method and system based on artificial intelligence |
| CN112052783B (en) * | 2020-09-02 | 2024-04-09 | 中南大学 | High-resolution image weak supervision building extraction method combining pixel semantic association and boundary attention |
| CN112052783A (en) * | 2020-09-02 | 2020-12-08 | 中南大学 | High-resolution image weak supervision building extraction method combining pixel semantic association and boundary attention |
| CN112182125A (en) * | 2020-09-14 | 2021-01-05 | 中国科学院重庆绿色智能技术研究院 | Business gathering area boundary identification system |
| CN112149594A (en) * | 2020-09-29 | 2020-12-29 | 同济大学 | Urban construction assessment method based on deep learning and high-resolution satellite images |
| CN112288331A (en) * | 2020-11-24 | 2021-01-29 | 哈尔滨工业大学 | Building reconstruction evaluation system based on semantic recognition |
| CN112288331B (en) * | 2020-11-24 | 2022-04-12 | 哈尔滨工业大学 | Building reconstruction evaluation system based on semantic recognition |
| CN114638769A (en) * | 2020-12-15 | 2022-06-17 | 武汉市探测者科技有限公司 | 2.5-dimensional map building hot zone automatic extraction method based on image processing algorithm |
| CN112580510B (en) * | 2020-12-18 | 2024-05-21 | 广州市城市规划设计有限公司 | Method, device, equipment and storage medium for estimating permeable ground area rate |
| CN112580510A (en) * | 2020-12-18 | 2021-03-30 | 广州市城市规划设计所 | Permeable ground area rate estimation method, permeable ground area rate estimation device, permeable ground area rate estimation equipment and storage medium |
| CN112990001A (en) * | 2021-03-11 | 2021-06-18 | 中国电建集团华东勘测设计研究院有限公司 | Remote sensing image monomer building rapid and accurate extraction method applied to digital collection |
| CN113516135B (en) * | 2021-06-23 | 2023-10-31 | 江苏师范大学 | Remote sensing image building extraction and contour optimization method based on deep learning |
| CN113516135A (en) * | 2021-06-23 | 2021-10-19 | 江苏师范大学 | Remote sensing image building extraction and contour optimization method based on deep learning |
| CN113505842B (en) * | 2021-07-21 | 2024-01-05 | 中国科学院空天信息创新研究院 | Automatic urban building extraction method suitable for large-scale regional remote sensing image |
| CN113505842A (en) * | 2021-07-21 | 2021-10-15 | 中国科学院空天信息创新研究院 | Automatic extraction method suitable for large-scale regional remote sensing image urban building |
| CN113486135B (en) * | 2021-07-27 | 2024-04-12 | 中南大学 | Building comprehensive method based on deep learning network |
| CN113486135A (en) * | 2021-07-27 | 2021-10-08 | 中南大学 | Building comprehensive method based on deep learning network |
| CN113610013A (en) * | 2021-08-10 | 2021-11-05 | 四川易方智慧科技有限公司 | Method for extracting building outline based on RGB (Red Green blue) wave bands of high-definition remote sensing image |
| CN113807301A (en) * | 2021-09-26 | 2021-12-17 | 武汉汉达瑞科技有限公司 | Automatic extraction method and automatic extraction system for newly-added construction land |
| CN114092813A (en) * | 2021-11-25 | 2022-02-25 | 中国科学院空天信息创新研究院 | Industrial park image extraction method, model, electronic equipment and storage medium |
| CN114092813B (en) * | 2021-11-25 | 2022-08-05 | 中国科学院空天信息创新研究院 | Industrial park image extraction method and system, electronic equipment and storage medium |
| CN114331031B (en) * | 2021-12-08 | 2022-12-09 | 北京华清安地建筑设计有限公司 | Building traditional feature recognition and evaluation method and system |
| CN114331031A (en) * | 2021-12-08 | 2022-04-12 | 北京华清安地建筑设计有限公司 | Building traditional feature recognition and evaluation method and system |
| CN114049569A (en) * | 2022-01-13 | 2022-02-15 | 自然资源部第三地理信息制图院 | Deep learning model performance evaluation method and system |
| CN114049569B (en) * | 2022-01-13 | 2022-03-18 | 自然资源部第三地理信息制图院 | Deep learning model performance evaluation method and system |
| CN114677522A (en) * | 2022-03-17 | 2022-06-28 | 北京科技大学 | Building structure type prediction method and system based on building pattern spots and deep learning |
| CN114677589B (en) * | 2022-03-18 | 2022-12-06 | 广州市城市规划勘测设计研究院 | City functional area identification method, device, equipment and medium based on remote sensing interpretation |
| CN114677589A (en) * | 2022-03-18 | 2022-06-28 | 广州市城市规划勘测设计研究院 | City functional area identification method, device, equipment and medium based on remote sensing interpretation |
| CN115841625A (en) * | 2023-02-23 | 2023-03-24 | 杭州电子科技大学 | Remote sensing building image extraction method based on improved U-Net model |
| CN115994987A (en) * | 2023-03-21 | 2023-04-21 | 天津市勘察设计院集团有限公司 | Rural building extraction and vectorization method based on inclined three-dimensional model |
| CN116052019A (en) * | 2023-03-31 | 2023-05-02 | 深圳市规划和自然资源数据管理中心(深圳市空间地理信息中心) | High-quality detection method suitable for built-up area of large-area high-resolution satellite image |
| CN116052019B (en) * | 2023-03-31 | 2023-07-25 | 深圳市规划和自然资源数据管理中心(深圳市空间地理信息中心) | High-quality detection method suitable for built-up area of large-area high-resolution satellite image |
| CN117218485A (en) * | 2023-09-05 | 2023-12-12 | 安徽省第二测绘院 | Deep learning model-based multi-source remote sensing image interpretation sample library creation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109446992B (en) | 2022-06-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109446992A (en) | Remote sensing image building extracting method and system, storage medium, electronic equipment based on deep learning | |
| CN110992311B (en) | Convolutional neural network flaw detection method based on feature fusion | |
| CN108009509A (en) | Vehicle target detection method | |
| CN109543630A (en) | Remote sensing image forest land extracting method and system, storage medium, electronic equipment based on deep learning | |
| CN105260738B (en) | High-resolution remote sensing image change detecting method and system based on Active Learning | |
| CN108416266B (en) | Method for rapidly identifying video behaviors by extracting moving object through optical flow | |
| CN110516539A (en) | Remote sensing image building extracting method, system, storage medium and equipment based on confrontation network | |
| CN106127204A (en) | A kind of multi-direction meter reading Region detection algorithms of full convolutional neural networks | |
| CN109800628A (en) | A kind of network structure and detection method for reinforcing SSD Small object pedestrian detection performance | |
| CN110490212A (en) | Molybdenum target image processing arrangement, method and apparatus | |
| CN107451602A (en) | A kind of fruits and vegetables detection method based on deep learning | |
| CN108416307A (en) | A kind of Aerial Images road surface crack detection method, device and equipment | |
| CN106920243A (en) | The ceramic material part method for sequence image segmentation of improved full convolutional neural networks | |
| CN110060237A (en) | A kind of fault detection method, device, equipment and system | |
| CN106296638A (en) | Significance information acquisition device and significance information acquisition method | |
| CN104103033B (en) | View synthesis method | |
| CN109493320A (en) | Method for extracting remote sensing image road and system, storage medium, electronic equipment based on deep learning | |
| CN107154048A (en) | The remote sensing image segmentation method and device of a kind of Pulse-coupled Neural Network Model | |
| CN109543632A (en) | A kind of deep layer network pedestrian detection method based on the guidance of shallow-layer Fusion Features | |
| CN107590515A (en) | The hyperspectral image classification method of self-encoding encoder based on entropy rate super-pixel segmentation | |
| CN109508675A (en) | A kind of pedestrian detection method for complex scene | |
| CN105405138A (en) | Water surface target tracking method based on saliency detection | |
| CN108229551A (en) | A kind of Classification of hyperspectral remote sensing image method based on compact dictionary rarefaction representation | |
| CN108664994A (en) | A kind of remote sensing image processing model construction system and method | |
| CN107807986A (en) | A kind of method for the remote sensing image intelligent Understanding for describing atural object spatial relation semantics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |