CN110059772A - Remote sensing images semantic segmentation method based on migration VGG network - Google Patents
Remote sensing images semantic segmentation method based on migration VGG network Download PDFInfo
- Publication number
- CN110059772A CN110059772A CN201910397121.7A CN201910397121A CN110059772A CN 110059772 A CN110059772 A CN 110059772A CN 201910397121 A CN201910397121 A CN 201910397121A CN 110059772 A CN110059772 A CN 110059772A
- Authority
- CN
- China
- Prior art keywords
- network
- remote sensing
- sensing images
- label
- convolution
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Image Analysis (AREA)
Abstract
The remote sensing images semantic segmentation technology based on VGG network that the invention discloses a kind of, the following steps are included: the high-resolution remote sensing images label figure corresponding with its for being used to training is cut into small image at random, network structure is divided into coding and decoding two parts, pass through anti-pond path, the expansion of encoded information resolution ratio is twice by deconvolution path, the result of itself and empty convolution is subjected to channel connection, it is up-sampled by deconvolution and characteristic image is restored to original size, output label figure input PPB module is subjected to multiple dimensioned polymerization processing again, finally using cross entropy as loss function, network parameter is updated by way of stochastic gradient descent；The small image that picture is sequentially cut will be tested and be input to its corresponding label figure of neural network prediction, then label figure is spliced into original size.Above-mentioned technical proposal reduces the complexity of network while improving the segmentation precision of model, saves time consumption for training.
Description
Technical field
The present invention relates to technical field of machine vision, and in particular to a kind of remote sensing images based on migration VGG network are semantic
Dividing method.
Background technique
Semantic segmentation is the major issue of the fields common concern such as unmanned, medical imaging analysis, GIS-Geographic Information System.
Semantic segmentation is that computer is allowed to be split according to the content of image, and segmentation is the difference from the level of pixel segmentation picture
Object is labeled each pixel in original image, is classified into different labels, and the precision divided then includes pair
The understanding of information in image.Remote sensing images have the characteristics that imaging is complicated, picture pixels are high, contain much information, therefore how to utilize
Artificial intelligence technology rapidly and accurately extracts the research hotspot that useful information is field of machine vision from remote sensing images.
Semantic segmentation neural network based has more research.FCN(Fully convolutional network)
It is the canonical frame of image, semantic segmentation, it is trained in end-to-end mode, and trained sorter network is used and language
Justice segmentation；In order to restore the resolution ratio of image, FCN is also up-sampled using deconvolution.Unlike FCN, SegNet is adopted
It is up-sampled with the method in anti-pond, so that the parameter of network is far fewer than FCN.Compared with FCN and SegNet, U-Net has
More symmetrical coding and decoding structure, and the recovery for facilitating location information is connected from the jump for being encoded to decoded portion, but
But also network structure becomes complicated, more training times are needed.Above-mentioned network structure is often used Chi Hualai and increases receptive field,
But pond will cause the reduction of space rate respectively when increasing receptive field.Although expanding receptive field by empty convolution
And resolution ratio loss is avoided, and the information of different scale can be captured using the convolution of different voidages, but it is empty
Convolution meeting local message missing by the way of sparse sampling, so that poor information correlation at a distance.In semantic segmentation
Big receptive field is capable of providing more global informations, but can ignore local message.It is considered that how to balance the big of receptive field
Small is the key that improve one of semantic segmentation precision, however under the premise of guaranteeing segmentation precision, reduce model complexity,
Training time is also that consider the problems of.
Summary of the invention
It can be in the segmentation essence for improving model in view of the deficiencies of the prior art, the present invention intends to provide one kind
The complexity of network is reduced while spending, and can save the remote sensing images semanteme point based on migration VGG network of time consumption for training
Segmentation method.
To achieve the above object, the present invention provides the following technical scheme that a kind of remote sensing images based on VGG network are semantic
Cutting techniques, comprising the following steps:
(1) the high-resolution remote sensing images label figure corresponding with its for being used to training is cut into 256 × 256 at random
The small image of pixel, the picture of cutting are divided into two parts, and training set of a part as network, another part is as verifying collection；
(2) network structure is divided into coding and decoding two parts, using first 16 layers of the VGG16 of sorter network as coding net
Network, decoding network are made of three anti-pond path, deconvolution path, empty Convolution path paths, by anti-pond path, instead
The expansion of encoded information resolution ratio is twice by Convolution path, and the result of itself and empty convolution is carried out channel connection, passes through deconvolution
Characteristic image is restored to original size by up-sampling, then output label figure input PPB module is carried out multiple dimensioned polymerization processing, most
Afterwards using cross entropy as loss function, network parameter is updated by way of stochastic gradient descent；
(3) the small image for testing picture and sequentially cutting into 256 × 256 pixels is input to neural network prediction its is corresponding
Label figure, then label figure is spliced into original size.
Preferably, step (2) includes following sub-step:
(1.1) by the remote sensing images random cropping of high pixel at the images fragment of specified size；
(1.2) semantic feature of pretreatment image fragment is extracted as coding network using first 16 layers of VGG network.
Preferably, step (2) further includes following sub-step:
(2.1) it is commonly used to restore the size of characteristic image using deconvolution and anti-pond, deconvolution and anti-pondization is mutually tied
It closes to up-sample, obtains the below plus anti-pond, then with 3 × 3 and 1 × 1 convolution in the 5th pondization of VGG network
One characteristic pattern；
(2.2) 3 × 3 and 1 × 1 convolution is connected below in the 5th pondization of VGG network, then again with 4 that step-length is 2
The size of × 4 deconvolution enlarging property figure cuts it to obtain second spy further according to the size of first characteristic pattern
Sign figure generates third characteristic pattern with 3 × 3 convolution that 3 voidages are 2 in the 4th Chi Huahou of VGG network；
(2.3) the third position degree of characteristic pattern that 3 paths generate is connected, integrates the information of different scale, allow net
Network oneself selects optimal combination；Characteristic pattern is restored to original size with 32 × 32 convolution of step-length 16 later, is passed through
Softmax layers of output prediction label mapping.
Preferably, the production method of prediction label the following steps are included:
(3.1) convolution that 3 × 3 are done to decoded label is averaged obtained characteristic pattern pond by the overall situation of different scale
Change and carries out down-sampling；
(3.2) down-sampling result is up-sampled, aggregates into a characteristic tensor with the mode that depth connects；
(3.3) prediction label is obtained with 1 × 1 convolution dimensionality reduction.
Preferably, the present invention is using a meter with Intel Core-i5 central processing unit and 4G byte of memory
Calculation machine simultaneously builds the algorithm frame for migrating the remote sensing images semantic segmentation of VGG network with Matlab language.
Preferably, being joined using cross entropy as loss function by the stochastic gradient descent that momentum is 0.9 to update
Number.
The invention has the advantages that compared with prior art,
1, the semantic segmentation model of the remote sensing images proposed by the invention based on VGG has high-resolution remote sensing images
Preferable segmentation effect.
2, remote sensing images semantic segmentation model proposed by the present invention greatly reduces under the premise of guaranteeing segmentation precision
The time consumption for training of network.
3, the decoding process that three kinds of paths proposed by the present invention combine also provides for the semantic segmentation of image a kind of new
Thinking.
The invention will be further described with specific embodiment with reference to the accompanying drawings of the specification.
Detailed description of the invention
Fig. 1 is the remote sensing images semantic segmentation model schematic that the embodiment of the present invention migrates VGG network；
Fig. 2 is network architecture of embodiment of the present invention schematic diagram；
Fig. 3 is PPB of embodiment of the present invention module diagram；
Fig. 4 is the curve synoptic diagram of loss function of the embodiment of the present invention；
Fig. 5 is the curve synoptic diagram that the embodiment of the present invention verifies precision.
Specific embodiment
In the description of the present embodiment, it should be noted that such as occur term " center ", "upper", "lower", "left", "right",
"vertical", "horizontal", "inner", "outside", "front", "rear" etc., indicated by orientation or positional relationship be it is based on the figure
Orientation or positional relationship is merely for convenience of description of the present invention and simplification of the description, rather than the device of indication or suggestion meaning or
Element must have a particular orientation, be constructed and operated in a specific orientation, it is thus impossible to be interpreted as limitation of the present invention.
In addition, such as there is term " first ", " second ", " third " are used for description purposes only, be not understood to indicate or imply opposite
Importance.
Referring to Fig. 1, Fig. 2, Fig. 3, Fig. 4 and Fig. 5, a kind of remote sensing images semanteme based on VGG network point disclosed by the invention
Cut technology, comprising the following steps:
(1) the high-resolution remote sensing images label figure corresponding with its for being used to training is cut into 256 × 256 at random
The small image of pixel, the picture of cutting are divided into two parts, and training set of a part as network, another part is as verifying collection；
(2) network structure is divided into coding and decoding two parts, using first 16 layers of the VGG16 of sorter network as coding net
Network, decoding network are made of three anti-pond path, deconvolution path, empty Convolution path paths, by anti-pond path, instead
The expansion of encoded information resolution ratio is twice by Convolution path, and the result of itself and empty convolution is carried out channel connection, passes through deconvolution
Characteristic image is restored to original size by up-sampling, then output label figure input PPB module is carried out multiple dimensioned polymerization processing, most
Afterwards using cross entropy as loss function, network parameter is updated by way of stochastic gradient descent；
(3) the small image for testing picture and sequentially cutting into 256 × 256 pixels is input to neural network prediction its is corresponding
Label figure, then label figure is spliced into original size.
It is used to be quantitatively evaluated the quality of image segmentation using following 4 indexs
Global precision: ∑inii/∑iti
Mean accuracy: (1/ncl)∑inii/ti
Averagely overlapping rate: (1/ncl)∑inii/(ti+∑jnji-nii)
Weight overlapping rate: (∑ktk)-1∑itinii/(ti+∑jnij-nii)
Wherein, nijIt is the number that i class pixel is predicted to be j class, the total number of i class pixel is
Preferably, step (2) includes following sub-step:
(1.1) by the remote sensing images random cropping of high pixel at the images fragment of specified size；
(1.2) semantic feature of pretreatment image fragment is extracted as coding network using first 16 layers of VGG network.
Preferably, step (2) further includes following sub-step:
(2.1) it is commonly used to restore the size of characteristic image using deconvolution and anti-pond, deconvolution and anti-pondization is mutually tied
It closes to up-sample, obtains the below plus anti-pond, then with 3 × 3 and 1 × 1 convolution in the 5th pondization of VGG network
One characteristic pattern；
(2.2) 3 × 3 and 1 × 1 convolution is connected below in the 5th pondization of VGG network, then again with 4 that step-length is 2
The size of × 4 deconvolution enlarging property figure cuts it to obtain second spy further according to the size of first characteristic pattern
Sign figure generates third characteristic pattern with 3 × 3 convolution that 3 voidages are 2 in the 4th Chi Huahou of VGG network；
(2.3) the third position degree of characteristic pattern that 3 paths generate is connected, integrates the information of different scale, allow net
Network oneself selects optimal combination；Characteristic pattern is restored to original size with 32 × 32 convolution of step-length 16 later, is passed through
Softmax layers of output prediction label mapping.
Preferably, the production method of prediction label the following steps are included:
(3.1) convolution that 3 × 3 are done to decoded label is averaged obtained characteristic pattern pond by the overall situation of different scale
Change and carries out down-sampling；
(3.2) down-sampling result is up-sampled, aggregates into a characteristic tensor with the mode that depth connects；
(3.3) prediction label is obtained with 1 × 1 convolution dimensionality reduction.
Preferably, the present invention is using a meter with Intel Core-i5 central processing unit and 4G byte of memory
Calculation machine simultaneously builds the algorithm frame for migrating the remote sensing images semantic segmentation of VGG network with Matlab language.
Preferably, being joined using cross entropy as loss function by the stochastic gradient descent that momentum is 0.9 to update
Number.
The present embodiment is further described below:
Conv: convolution algorithm (convolution)
A kind of Pooling: operation similar to down-sampling；
ReLu: activation primitive, mathematical form are max (0, x)；
Softmax: assuming that V is an array, ViIt is i-th of element of V, mathematical notation are as follows:
Deconv: transposition convolution algorithm (deconvolution) can be used for up-sampling.
Unpooling: anti-pond can be used for up-sampling.
Dilated Conv: empty convolution improves the receptive field of convolution results using voidage without reducing resolution ratio.
Described in the detail of network structure:
Pond layer is commonly used in extraction abstract characteristics and filters out noisy activation but will cause input feature vector rate respectively
Contraction and loss of learning.Deconvolution and anti-pond are commonly used to restore the size of characteristic image, and deconvolution is combined with anti-pond
It up-samples, adds anti-pond below in the 5th pondization of VGG network, then obtain first with 3 × 3 and 1 × 1 convolution
A characteristic pattern；In addition, connecting 3 × 3 and 1 × 1 convolution behind the 5th pondization of VGG network, then using step-length again is 2
4 × 4 deconvolution expands the size of characteristic pattern, is cut to obtain second spy to it further according to the size of first characteristic pattern
Sign figure.
Third characteristic pattern is generated with 3 × 3 convolution that 3 voidages are 2 behind the 4th pond of VGG network.
Finally the third position degree for the characteristic pattern that this 3 paths generate is connected, the information of different scale is integrated, allows
Network oneself selects optimal combination.Characteristic pattern is restored to original size with 32 × 32 convolution of step-length 16 later.
As shown in figure 3, feature will be exported by carrying out 4 times after 3 × 3 process of convolution respectively, 8 back, 16 times, 32 times complete
The average pond of office, constructs 4 pond pyramids, finally with 1 × 1 convolution dimensionality reduction, is reflected by softmax layers of output prediction label
It penetrates.
The remote sensing images semantic segmentation model of migration VGG network of the invention, as loss function, is passed through using cross entropy
The stochastic gradient descent that momentum is 0.9 carrys out undated parameter.The loss function and verifying precision of network training process are as shown in Figure 3.
The hardware and programming language of method carrying out practically of the invention are not intended to limit, and being write with any language can be complete
At other operating modes repeat no more thus.
The semantic segmentation model of remote sensing images based on VGG proposed by the invention to high-resolution remote sensing images have compared with
Good segmentation effect.
Remote sensing images semantic segmentation model proposed by the present invention greatly reduces net under the premise of guaranteeing segmentation precision
The time consumption for training of network.
The decoding process that three kinds of paths proposed by the present invention combine also provides a kind of new think of for the semantic segmentation of image
Road.
Above-described embodiment is served only for that invention is further explained to specific descriptions of the invention, should not be understood as
Limiting the scope of the present invention, the technician of this field make the present invention according to the content of foregoing invention some non-
The modifications and adaptations of essence are fallen within the scope of protection of the present invention.
Claims (6)
1. a kind of remote sensing images semantic segmentation technology based on VGG network, it is characterised in that: the following steps are included:
(1) the high-resolution remote sensing images label figure corresponding with its for being used to training is cut into 256 × 256 pixels at random
Small image, the picture of cutting are divided into two parts, and training set of a part as network, another part is as verifying collection；
(2) network structure is divided into coding and decoding two parts, using first 16 layers of the VGG16 of sorter network as coding network, solves
Code network is made of three anti-pond path, deconvolution path, empty Convolution path paths, passes through anti-pond path, deconvolution
The expansion of encoded information resolution ratio is twice by path, the result of itself and empty convolution is carried out channel connection, by adopting in deconvolution
Characteristic image is restored to original size by sample, then output label figure input PPB module is carried out multiple dimensioned polymerization processing, finally with
Cross entropy is loss function, updates network parameter by way of stochastic gradient descent；
(3) the small image that test picture sequentially cuts into 256 × 256 pixels is input to its corresponding label of neural network prediction
Figure, then label figure is spliced into original size.
2. a kind of remote sensing images semantic segmentation technology based on VGG network according to claim 1, it is characterised in that: step
Suddenly (2) include following sub-step:
(1.1) by the remote sensing images random cropping of high pixel at the images fragment of specified size；
(1.2) semantic feature of pretreatment image fragment is extracted as coding network using first 16 layers of VGG network.
3. a kind of remote sensing images semantic segmentation technology based on VGG network according to claim 1, it is characterised in that: step
Suddenly (2) further include following sub-step:
(2.1) it is commonly used to restore the size of characteristic image using deconvolution and anti-pond, deconvolution is combined to come with anti-pond
Up-sampling obtains first spy below plus anti-pond, then with 3 × 3 and 1 × 1 convolution in the 5th pondization of VGG network
Sign figure；
(2.2) 3 × 3 and 1 × 1 convolution is connected below in the 5th pondization of VGG network, then again with 4 × 4 that step-length is 2
Deconvolution enlarging property figure size, it is cut to obtain second feature further according to the size of first characteristic pattern
Figure generates third characteristic pattern with 3 × 3 convolution that 3 voidages are 2 in the 4th Chi Huahou of VGG network；
(2.3) the third position degree of characteristic pattern that 3 paths generate is connected, integrates the information of different scale, allow network from
Oneself selects optimal combination；Characteristic pattern is restored to original size with 32 × 32 convolution of step-length 16 later, passes through softmax
Layer output prediction label mapping.
4. a kind of remote sensing images semantic segmentation technology based on VGG network according to claim 3, it is characterised in that: pre-
The production methods of mark label the following steps are included:
(3.1) convolution that 3 × 3 are done to decoded label is averaged obtained characteristic pattern Chi Huajin by the overall situation of different scale
Row down-sampling；
(3.2) down-sampling result is up-sampled, aggregates into a characteristic tensor with the mode that depth connects；
(3.3) prediction label is obtained with 1 × 1 convolution dimensionality reduction.
5. a kind of remote sensing images semantic segmentation technology based on VGG network according to claim 1, it is characterised in that: adopt
With a computer with Intel Core-i5 central processing unit and 4G byte of memory and migration is built with Matlab language
The algorithm frame of the remote sensing images semantic segmentation of VGG network.
6. a kind of remote sensing images semantic segmentation technology based on VGG network according to claim 1, it is characterised in that: adopt
Use cross entropy as loss function, the stochastic gradient descent for being 0.9 by momentum is come undated parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910397121.7A CN110059772B (en) | 2019-05-14 | 2019-05-14 | Remote sensing image semantic segmentation method based on multi-scale decoding network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910397121.7A CN110059772B (en) | 2019-05-14 | 2019-05-14 | Remote sensing image semantic segmentation method based on multi-scale decoding network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110059772A true CN110059772A (en) | 2019-07-26 |
| CN110059772B CN110059772B (en) | 2021-04-30 |
Family
ID=67323037
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910397121.7A Active CN110059772B (en) | 2019-05-14 | 2019-05-14 | Remote sensing image semantic segmentation method based on multi-scale decoding network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110059772B (en) |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503052A (en) * | 2019-08-27 | 2019-11-26 | 西南交通大学 | A kind of image, semantic dividing method based on improvement U-NET network |
| CN110569790A (en) * | 2019-09-05 | 2019-12-13 | 中国人民解放军61540部队 | Residential area element extraction method based on texture enhancement convolutional network |
| CN110674742A (en) * | 2019-09-24 | 2020-01-10 | 电子科技大学 | Remote sensing image road extraction method based on DLinkNet |
| CN110781895A (en) * | 2019-10-10 | 2020-02-11 | 湖北工业大学 | Image semantic segmentation method based on convolutional neural network |
| CN110781923A (en) * | 2019-09-27 | 2020-02-11 | 重庆特斯联智慧科技股份有限公司 | Feature extraction method and device |
| CN111080652A (en) * | 2019-12-23 | 2020-04-28 | 西安电子科技大学 | Optical remote sensing image segmentation method based on multi-scale lightweight cavity convolution |
| CN111079649A (en) * | 2019-12-17 | 2020-04-28 | 西安电子科技大学 | Remote sensing image ground feature classification method based on lightweight semantic segmentation network |
| CN111126451A (en) * | 2019-12-01 | 2020-05-08 | 复旦大学 | Method for dual semantic segmentation |
| CN111161277A (en) * | 2019-12-12 | 2020-05-15 | 中山大学 | Natural image matting method based on deep learning |
| CN111209813A (en) * | 2019-12-27 | 2020-05-29 | 南京航空航天大学 | Remote sensing image semantic segmentation method based on transfer learning |
| CN111259905A (en) * | 2020-01-17 | 2020-06-09 | 山西大学 | Feature fusion remote sensing image semantic segmentation method based on downsampling |
| CN111368843A (en) * | 2020-03-06 | 2020-07-03 | 电子科技大学 | Method for extracting lake on ice based on semantic segmentation |
| CN111401379A (en) * | 2020-03-24 | 2020-07-10 | 北方民族大学 | Deep L abv3plus-IRCNet image semantic segmentation algorithm based on coding and decoding structure |
| CN111460936A (en) * | 2020-03-18 | 2020-07-28 | 中国地质大学（武汉） | Remote sensing image building extraction method, system and electronic equipment based on U-Net network |
| CN111553921A (en) * | 2020-02-19 | 2020-08-18 | 中山大学 | Real-time semantic segmentation method based on channel information sharing residual error module |
| CN111738268A (en) * | 2020-07-22 | 2020-10-02 | 浙江大学 | Semantic segmentation method and system for high-resolution remote sensing image based on random block |
| CN111784682A (en) * | 2020-07-10 | 2020-10-16 | 北京医智影科技有限公司 | Network architecture, program carrier and workstation for automatic processing of images |
| CN111798460A (en) * | 2020-06-17 | 2020-10-20 | 南京信息工程大学 | Satellite image segmentation method |
| CN111931779A (en) * | 2020-08-10 | 2020-11-13 | 韶鼎人工智能科技有限公司 | Image information extraction and generation method based on condition predictable parameters |
| CN111951292A (en) * | 2020-09-01 | 2020-11-17 | 吉林大学 | Object surface reflection attribute extraction method, device, equipment and storage medium |
| CN112070779A (en) * | 2020-08-04 | 2020-12-11 | 武汉大学 | Remote sensing image road segmentation method based on convolutional neural network weak supervised learning |
| CN112084923A (en) * | 2020-09-01 | 2020-12-15 | 西安电子科技大学 | Semantic segmentation method for remote sensing image, storage medium and computing device |
| CN112419322A (en) * | 2020-10-16 | 2021-02-26 | 北京工业大学 | Temporal bone external semicircular canal segmentation method based on 3D multi-scale multi-pooling feature fusion network |
| CN112508031A (en) * | 2020-12-22 | 2021-03-16 | 北京航空航天大学 | Unsupervised remote sensing image semantic segmentation method and model from virtual to reality |
| CN112861829A (en) * | 2021-04-13 | 2021-05-28 | 山东大学 | Water body extraction method and system based on deep convolutional neural network |
| CN113592885A (en) * | 2021-06-29 | 2021-11-02 | 中南大学 | SegNet-RS network-based large obstacle contour segmentation method |
| CN113706695A (en) * | 2021-09-01 | 2021-11-26 | 杭州柳叶刀机器人有限公司 | System and method for performing 3D femoral head modeling through deep learning and storage medium |
| CN115205300A (en) * | 2022-09-19 | 2022-10-18 | 华东交通大学 | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion |
| US11514694B2 (en) | 2019-09-20 | 2022-11-29 | Samsung Electronics Co., Ltd. | Teaching GAN (generative adversarial networks) to generate per-pixel annotation |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7096272B1 (en) * | 2001-11-20 | 2006-08-22 | Cisco Technology, Inc. | Methods and apparatus for pooling and depooling the transmission of stream data |
| CN106372577A (en) * | 2016-08-23 | 2017-02-01 | 北京航空航天大学 | Deep learning-based traffic sign automatic identifying and marking method |
| CN106650690A (en) * | 2016-12-30 | 2017-05-10 | 东华大学 | Night vision image scene identification method based on deep convolution-deconvolution neural network |
| CN106682730A (en) * | 2017-01-10 | 2017-05-17 | 西安电子科技大学 | Network performance assessment method based on VGG16 image deconvolution |
| CN106981080A (en) * | 2017-02-24 | 2017-07-25 | 东华大学 | Night unmanned vehicle scene depth method of estimation based on infrared image and radar data |
| US20170243053A1 (en) * | 2016-02-18 | 2017-08-24 | Pinscreen, Inc. | Real-time facial segmentation and performance capture from rgb input |
| CN108062753A (en) * | 2017-12-29 | 2018-05-22 | 重庆理工大学 | The adaptive brain tumor semantic segmentation method in unsupervised domain based on depth confrontation study |
| AU2018101336A4 (en) * | 2018-09-12 | 2018-10-11 | Hu, Yuan Miss | Building extraction application based on machine learning in Urban-Suburban-Integration Area |
| CN108830913A (en) * | 2018-05-25 | 2018-11-16 | 大连理工大学 | Semantic level line original text painting methods based on User Colors guidance |
| CN109145769A (en) * | 2018-08-01 | 2019-01-04 | 辽宁工业大学 | The target detection network design method of blending image segmentation feature |
| CN109190626A (en) * | 2018-07-27 | 2019-01-11 | 国家新闻出版广电总局广播科学研究院 | A kind of semantic segmentation method of the multipath Fusion Features based on deep learning |
| CN109614973A (en) * | 2018-11-22 | 2019-04-12 | 华南农业大学 | Rice seedling and Weeds at seedling image, semantic dividing method, system, equipment and medium |
| CN109636802A (en) * | 2019-01-18 | 2019-04-16 | 天津工业大学 | Pulmonary parenchyma based on depth convolutional neural networks is through CT image partition method |
| CN109636905A (en) * | 2018-12-07 | 2019-04-16 | 东北大学 | Environment semanteme based on depth convolutional neural networks builds drawing method |
-
2019
- 2019-05-14 CN CN201910397121.7A patent/CN110059772B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7096272B1 (en) * | 2001-11-20 | 2006-08-22 | Cisco Technology, Inc. | Methods and apparatus for pooling and depooling the transmission of stream data |
| US20170243053A1 (en) * | 2016-02-18 | 2017-08-24 | Pinscreen, Inc. | Real-time facial segmentation and performance capture from rgb input |
| CN106372577A (en) * | 2016-08-23 | 2017-02-01 | 北京航空航天大学 | Deep learning-based traffic sign automatic identifying and marking method |
| CN106650690A (en) * | 2016-12-30 | 2017-05-10 | 东华大学 | Night vision image scene identification method based on deep convolution-deconvolution neural network |
| CN106682730A (en) * | 2017-01-10 | 2017-05-17 | 西安电子科技大学 | Network performance assessment method based on VGG16 image deconvolution |
| CN106981080A (en) * | 2017-02-24 | 2017-07-25 | 东华大学 | Night unmanned vehicle scene depth method of estimation based on infrared image and radar data |
| CN108062753A (en) * | 2017-12-29 | 2018-05-22 | 重庆理工大学 | The adaptive brain tumor semantic segmentation method in unsupervised domain based on depth confrontation study |
| CN108830913A (en) * | 2018-05-25 | 2018-11-16 | 大连理工大学 | Semantic level line original text painting methods based on User Colors guidance |
| CN109190626A (en) * | 2018-07-27 | 2019-01-11 | 国家新闻出版广电总局广播科学研究院 | A kind of semantic segmentation method of the multipath Fusion Features based on deep learning |
| CN109145769A (en) * | 2018-08-01 | 2019-01-04 | 辽宁工业大学 | The target detection network design method of blending image segmentation feature |
| AU2018101336A4 (en) * | 2018-09-12 | 2018-10-11 | Hu, Yuan Miss | Building extraction application based on machine learning in Urban-Suburban-Integration Area |
| CN109614973A (en) * | 2018-11-22 | 2019-04-12 | 华南农业大学 | Rice seedling and Weeds at seedling image, semantic dividing method, system, equipment and medium |
| CN109636905A (en) * | 2018-12-07 | 2019-04-16 | 东北大学 | Environment semanteme based on depth convolutional neural networks builds drawing method |
| CN109636802A (en) * | 2019-01-18 | 2019-04-16 | 天津工业大学 | Pulmonary parenchyma based on depth convolutional neural networks is through CT image partition method |
Non-Patent Citations (2)
| Title |
|---|
| MOHSEN FAYYAZ等: "STFCN: Spatio-Temporal Fully Convolutional Neural Network for Semantic Segmentation of Street Scenes", 《SEMANTIC SEGMENTATION VGG16 HOLE CONVOLUTION》 * |
| 侯津京: "基于GAN的城市环境图像语义分割", 《中国优秀硕士学位论文全文数据库（信息科技辑）》 * |
Cited By (49)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503052A (en) * | 2019-08-27 | 2019-11-26 | 西南交通大学 | A kind of image, semantic dividing method based on improvement U-NET network |
| CN110569790A (en) * | 2019-09-05 | 2019-12-13 | 中国人民解放军61540部队 | Residential area element extraction method based on texture enhancement convolutional network |
| CN110569790B (en) * | 2019-09-05 | 2023-03-28 | 中国人民解放军61540部队 | Residential area element extraction method based on texture enhancement convolutional network |
| US11514694B2 (en) | 2019-09-20 | 2022-11-29 | Samsung Electronics Co., Ltd. | Teaching GAN (generative adversarial networks) to generate per-pixel annotation |
| CN110674742B (en) * | 2019-09-24 | 2023-04-07 | 电子科技大学 | Remote sensing image road extraction method based on DLinkNet |
| CN110674742A (en) * | 2019-09-24 | 2020-01-10 | 电子科技大学 | Remote sensing image road extraction method based on DLinkNet |
| CN110781923B (en) * | 2019-09-27 | 2023-02-07 | 重庆特斯联智慧科技股份有限公司 | Feature extraction method and device |
| CN110781923A (en) * | 2019-09-27 | 2020-02-11 | 重庆特斯联智慧科技股份有限公司 | Feature extraction method and device |
| CN110781895B (en) * | 2019-10-10 | 2023-06-20 | 湖北工业大学 | Image semantic segmentation method based on convolutional neural network |
| CN110781895A (en) * | 2019-10-10 | 2020-02-11 | 湖北工业大学 | Image semantic segmentation method based on convolutional neural network |
| CN111126451A (en) * | 2019-12-01 | 2020-05-08 | 复旦大学 | Method for dual semantic segmentation |
| CN111161277A (en) * | 2019-12-12 | 2020-05-15 | 中山大学 | Natural image matting method based on deep learning |
| CN111161277B (en) * | 2019-12-12 | 2023-04-18 | 中山大学 | Natural image matting method based on deep learning |
| CN111079649A (en) * | 2019-12-17 | 2020-04-28 | 西安电子科技大学 | Remote sensing image ground feature classification method based on lightweight semantic segmentation network |
| CN111080652A (en) * | 2019-12-23 | 2020-04-28 | 西安电子科技大学 | Optical remote sensing image segmentation method based on multi-scale lightweight cavity convolution |
| CN111209813A (en) * | 2019-12-27 | 2020-05-29 | 南京航空航天大学 | Remote sensing image semantic segmentation method based on transfer learning |
| CN111209813B (en) * | 2019-12-27 | 2021-01-22 | 南京航空航天大学 | Remote sensing image semantic segmentation method based on transfer learning |
| CN111259905B (en) * | 2020-01-17 | 2022-05-31 | 山西大学 | Feature fusion remote sensing image semantic segmentation method based on downsampling |
| CN111259905A (en) * | 2020-01-17 | 2020-06-09 | 山西大学 | Feature fusion remote sensing image semantic segmentation method based on downsampling |
| CN111553921B (en) * | 2020-02-19 | 2023-04-25 | 中山大学 | Real-time semantic segmentation method based on channel information sharing residual error module |
| CN111553921A (en) * | 2020-02-19 | 2020-08-18 | 中山大学 | Real-time semantic segmentation method based on channel information sharing residual error module |
| 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 |
| CN111460936A (en) * | 2020-03-18 | 2020-07-28 | 中国地质大学（武汉） | Remote sensing image building extraction method, system and electronic equipment based on U-Net network |
| CN111401379A (en) * | 2020-03-24 | 2020-07-10 | 北方民族大学 | Deep L abv3plus-IRCNet image semantic segmentation algorithm based on coding and decoding structure |
| CN111798460B (en) * | 2020-06-17 | 2023-08-01 | 南京信息工程大学 | Satellite image segmentation method |
| CN111798460A (en) * | 2020-06-17 | 2020-10-20 | 南京信息工程大学 | Satellite image segmentation method |
| CN111784682B (en) * | 2020-07-10 | 2024-05-28 | 北京医智影科技有限公司 | Network architecture system, program carrier and workstation for automatic processing of images |
| CN111784682A (en) * | 2020-07-10 | 2020-10-16 | 北京医智影科技有限公司 | Network architecture, program carrier and workstation for automatic processing of images |
| CN111738268B (en) * | 2020-07-22 | 2023-11-14 | 浙江大学 | Semantic segmentation method and system for high-resolution remote sensing image based on random block |
| CN111738268A (en) * | 2020-07-22 | 2020-10-02 | 浙江大学 | Semantic segmentation method and system for high-resolution remote sensing image based on random block |
| CN112070779A (en) * | 2020-08-04 | 2020-12-11 | 武汉大学 | Remote sensing image road segmentation method based on convolutional neural network weak supervised learning |
| CN112070779B (en) * | 2020-08-04 | 2022-03-15 | 武汉大学 | Remote sensing image road segmentation method based on convolutional neural network weak supervised learning |
| CN111931779A (en) * | 2020-08-10 | 2020-11-13 | 韶鼎人工智能科技有限公司 | Image information extraction and generation method based on condition predictable parameters |
| CN111931779B (en) * | 2020-08-10 | 2024-06-04 | 韶鼎人工智能科技有限公司 | Image information extraction and generation method based on condition predictable parameters |
| CN112084923A (en) * | 2020-09-01 | 2020-12-15 | 西安电子科技大学 | Semantic segmentation method for remote sensing image, storage medium and computing device |
| CN111951292A (en) * | 2020-09-01 | 2020-11-17 | 吉林大学 | Object surface reflection attribute extraction method, device, equipment and storage medium |
| CN112084923B (en) * | 2020-09-01 | 2023-12-22 | 西安电子科技大学 | Remote sensing image semantic segmentation method, storage medium and computing device |
| CN112419322A (en) * | 2020-10-16 | 2021-02-26 | 北京工业大学 | Temporal bone external semicircular canal segmentation method based on 3D multi-scale multi-pooling feature fusion network |
| CN112508031A (en) * | 2020-12-22 | 2021-03-16 | 北京航空航天大学 | Unsupervised remote sensing image semantic segmentation method and model from virtual to reality |
| CN112508031B (en) * | 2020-12-22 | 2022-09-02 | 北京航空航天大学 | Unsupervised remote sensing image semantic segmentation method and model from virtual to reality |
| CN112861829B (en) * | 2021-04-13 | 2023-06-30 | 山东大学 | Water body extraction method and system based on deep convolutional neural network |
| CN112861829A (en) * | 2021-04-13 | 2021-05-28 | 山东大学 | Water body extraction method and system based on deep convolutional neural network |
| CN113592885B (en) * | 2021-06-29 | 2024-03-12 | 中南大学 | SegNet-RS network-based large obstacle contour segmentation method |
| CN113592885A (en) * | 2021-06-29 | 2021-11-02 | 中南大学 | SegNet-RS network-based large obstacle contour segmentation method |
| CN113706695B (en) * | 2021-09-01 | 2023-06-23 | 杭州柳叶刀机器人有限公司 | System and method for deep learning 3D femoral head modeling and storage medium |
| CN113706695A (en) * | 2021-09-01 | 2021-11-26 | 杭州柳叶刀机器人有限公司 | System and method for performing 3D femoral head modeling through deep learning and storage medium |
| CN115205300B (en) * | 2022-09-19 | 2022-12-09 | 华东交通大学 | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion |
| CN115205300A (en) * | 2022-09-19 | 2022-10-18 | 华东交通大学 | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110059772B (en) | 2021-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059772A (en) | Remote sensing images semantic segmentation method based on migration VGG network | |
| CN111259905B (en) | Feature fusion remote sensing image semantic segmentation method based on downsampling | |
| CN112200161A (en) | Face recognition detection method based on mixed attention mechanism | |
| CN113780149B (en) | Remote sensing image building target efficient extraction method based on attention mechanism | |
| CN110287960A (en) | The detection recognition method of curve text in natural scene image | |
| CN113780211A (en) | Lightweight aircraft detection method based on improved yolk 4-tiny | |
| CN109241972A (en) | Image, semantic dividing method based on deep learning | |
| CN109886066A (en) | Fast target detection method based on the fusion of multiple dimensioned and multilayer feature | |
| CN110853057B (en) | Aerial image segmentation method based on global and multi-scale full-convolution network | |
| Wang et al. | FE-YOLOv5: Feature enhancement network based on YOLOv5 for small object detection | |
| CN109559297A (en) | A method of generating the Lung neoplasm detection of network based on 3D region | |
| CN109711409A (en) | A kind of hand-written music score spectral line delet method of combination U-net and ResNet | |
| CN115035361A (en) | Target detection method and system based on attention mechanism and feature cross fusion | |
| CN112287983B (en) | Remote sensing image target extraction system and method based on deep learning | |
| CN116645592B (en) | Crack detection method based on image processing and storage medium | |
| CN116580241B (en) | Image processing method and system based on double-branch multi-scale semantic segmentation network | |
| CN112580453A (en) | Land use classification method and system based on remote sensing image and deep learning | |
| CN114267025A (en) | Traffic sign detection method based on high-resolution network and light-weight attention mechanism | |
| CN115457043A (en) | Image segmentation network based on overlapped self-attention deformer framework U-shaped network | |
| CN116310916A (en) | Semantic segmentation method and system for high-resolution remote sensing city image | |
| CN116912708A (en) | Remote sensing image building extraction method based on deep learning | |
| CN116309536A (en) | Pavement crack detection method and storage medium | |
| CN113361528B (en) | Multi-scale target detection method and system | |
| Wang et al. | YOLO V4 with hybrid dilated convolution attention module for object detection in the aerial dataset | |
| CN117351372A (en) | Remote sensing image road segmentation method based on improved deep V & lt3+ & gt |
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 |