CN109255334A - Remote sensing image terrain classification method based on deep learning semantic segmentation network - Google Patents

Remote sensing image terrain classification method based on deep learning semantic segmentation network Download PDF

Info

Publication number
CN109255334A
CN109255334A CN201811130333.0A CN201811130333A CN109255334A CN 109255334 A CN109255334 A CN 109255334A CN 201811130333 A CN201811130333 A CN 201811130333A CN 109255334 A CN109255334 A CN 109255334A
Authority
CN
China
Prior art keywords
remote sensing
layer
image
deep learning
semantic segmentation
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
Application number
CN201811130333.0A
Other languages
Chinese (zh)
Other versions
CN109255334B (en
Inventor
楚博策
帅通
高峰
王士成
陈金勇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CETC 54 Research Institute
Original Assignee
CETC 54 Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by CETC 54 Research Institute filed Critical CETC 54 Research Institute
Priority to CN201811130333.0A priority Critical patent/CN109255334B/en
Publication of CN109255334A publication Critical patent/CN109255334A/en
Application granted granted Critical
Publication of CN109255334B publication Critical patent/CN109255334B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Astronomy & Astrophysics (AREA)
  • Remote Sensing (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a kind of remote sensing image terrain classification methods based on deep learning semantic segmentation network, are marked pixel-by-pixel to atural objects all kinds of in remote sensing image first, and building remote sensing atural object mark Image Database is as training label.The subsequent present invention designs a kind of using the method for constructing Analysis On Multi-scale Features figure group based on texture, structure feature, characteristic pattern group and raw video are incorporated as to the input of deep learning network, in addition to this present invention devises a kind of improvement network structure of full convolutional network according to deeplab algorithm, parameter training is carried out by convolution and deconvolution, overlapping cutting finally is carried out to wide cut remote sensing images, merges after classification and obtains final wide cut remote sensing image terrain classification result.It efficiently can promptly realize the various atural object Pixel-level classification of high-resolution remote sensing image, simplify the Complicated Flow of traditional classification method, and realize segmentation and classifying quality well.

Description

Remote sensing image terrain classification method based on deep learning semantic segmentation network
Technical field
The invention belongs to remote sensing images intelligent classification technical fields, more specifically, are related under atural object interpretation demand A kind of remote sensing terrain classification method based on full convolution semantic segmentation network.
Background technique
Remote sensing image terrain classification be now widely used for ground investigation, defend piece law enforcement, region investigation etc. all kinds of the army and the people answer With in the middle, achieving preferable application effect and possess biggish market development potential.With satellite load and data volume by Step increases, traditional when especially facing big region (whole nation or Global Scale) Surface classification in remote sensing high-precision classification research The method manually demarcated is difficult to support the task of explosive growth and required workload, therefore how studies using artificial intelligence Method realizes that the intelligent automatic processing of remote sensing image is a far-reaching important process.
Current existing terrain classification method,
(1) the most of method of tradition mainly uses the dividing methods such as super-pixel that Remote Sensing Image Segmentation is become several regions, Then to the traditional characteristics such as form, texture are extracted in region, finally to territorial classification and closed using classifiers such as SVM according to feature And form classification results.
(2) research of deep learning recent years in terms of terrain classification is focused primarily upon to raw video using super-pixel The methods of carry out initial segmentation, and classified to the block image after segmentation using the neural networks such as CNN, DBN, to reach picture The purpose of plain grade terrain classification.
(3) present invention innovatively proposes the Analysis On Multi-scale Features description figure of a kind of image texture and structure, multiple dimensioned spy Sign description figure merges generation dimensional images with original 3 d image and realizes that the features such as image texture describe power intensification, then application It is improved in the full convolution depth network of the deep learning model-of semantic segmentation task and is applied to high-resolution remote sensing image Terrain classification, it is integrated with assorting process to realize segmentation, achieves good classifying quality.
Summary of the invention
It is an object of the invention to overcome the deficiencies in the prior art, propose the remote sensing based on deep learning semantic segmentation network Image terrain classification method reduces error in classification to further increase nicety of grading and efficiency.
The object of the present invention is achieved like this:
A kind of remote sensing image terrain classification method based on deep learning semantic segmentation network, comprising the following steps:
(1) remote sensing images for acquiring the High Resolution Visible Light of different loads, carry out by picture atural object in each image Element marks, and the data after mark form a width bianry image, original remote sensing images and corresponding mark image packing composing training Collection and test set；
(2) special to two-dimensional entropy, roughness and contrast texture using multiple dimensioned window to remote sensing images original in training set Sign extracts, and forms Analysis On Multi-scale Features figure group；And use Canny operator by atural object edge remote sensing images original in training set It extracts, forms structure feature figure；
(3) based on the thought of DeepLab, the full convolution semantic segmentation model of deep learning is constructed；
(4) the Analysis On Multi-scale Features figure group generated in step (2), structure feature figure and original remote sensing images are combined shape At input figure group, the input as the full convolution semantic segmentation model of deep learning in step (3) carries out model training, finally obtains The model of parameter stability；
(5) cutting is carried out to original remote sensing images to be sorted in test set, image after cutting is passed through into instruction in step (4) The model for the parameter stability perfected is classified, and classification results are merged together and generate wide cut testing result, work as overlapping region Testing result generate contradiction when, retain wherein be classified as non-background pixel as a result, obtaining final amalgamation result.
Wherein, full convolution semantic segmentation model in step (3) specifically:
Model is divided into downwardly and upwardly two sections, wherein downward access is according to classical strength by the original 13 layers of convolution of VGGnet Layer is changed to 6 layers, and input layer obtains 16 × 16 dimensional feature temperature figures as the 7th layer after 6 layers of convolution sum pondization；Upward access In, interpolation up-sampling is carried out to the 7th layer of warp lamination and restores extremely to merge after up-sampling the 7th layer with the 6th layer of identical size 5th layer of porous convolution generates the 8th layer；The 6th layer of porous convolution is merged after up-sampling to the 8th layer and generates the 9th layer, to the 9th layer Output carry out size change over be restored to original remote sensing images size, obtain final classification result.
It has the advantages that compared with the background technology, the present invention
1, the invention proposes Analysis On Multi-scale Features figure groups inputs instead of simple RGB image, and Enhanced feature characterizes power, and guides The feature extraction direction of neural network；
2, the present invention uses the full convolutional neural networks of deep learning, and terrain classification task end to end may be implemented, and replaces Error accumulation caused by conventional method is made of multi-step.
3, the present invention is applied to remote sensing terrain classification direction using Deeplab network structure.Porous convolution can be solved effectively The certainly problem of remote sensing large scale atural object receptive field deficiency, while effectively optimization boundary can be used using condition random field CRF, it extracts Marginal classification effect.
Detailed description of the invention
Fig. 1 is process design drawing of the invention.
Fig. 2 is textural characteristics figure of the present invention.
Fig. 3 is Analysis On Multi-scale Features figure group building schematic diagram of the present invention.
Fig. 4 is structure of the invention characteristic pattern.
Fig. 5 is the full convolutional network design drawing of foundation of the present invention.
Fig. 6 is that the present invention improves Deeplab network design figure.
Fig. 7 is depth network structure of the present invention.
Fig. 8 is that the present invention uses narrow high resolution image figure.
Fig. 9 is the accuracy comparison figure of classification method of the present invention and other methods.
Figure 10 is the method for the present invention and other methods classifying quality comparison diagram.
Figure 11 is the method for the present invention wide cut image classification effect picture.
Specific embodiment
A specific embodiment of the invention is described with reference to the accompanying drawing, preferably so as to those skilled in the art Understand the present invention.Requiring particular attention is that in the following description, when known function and the detailed description of design perhaps When can desalinate main contents of the invention, these descriptions will be ignored herein.
Fig. 1 is that the present invention is based under a kind of specific implementation of remote sensing image terrain classification method of deep semantic segmentation network Functional block diagram.
In the present embodiment, as shown in Figure 1 based on the remote sensing image terrain classification method of deep learning semantic segmentation network The following steps are included:
1, data preparation
Data preparation includes acquisition and the mark of image, wherein the remote sensing figure of the High Resolution Visible Light of acquisition different loads Then picture marks atural object in each image pixel-by-pixel, the data after mark form a width bianry image, wherein 0 ash Degree indicates background pixel, and 1-6 gray scale indicates 6 kinds of types of ground objects such as building, meadow, and remote sensing images and corresponding mark image are beaten Packet composing training collection and test set；
2. characteristic pattern group is extracted
Characteristic pattern group extracts the extraction including multi-dimension texture characteristic pattern group and structure feature figure.Consider different atural objects at certain Otherness is unobvious in a little textural characteristics (such as: directionality etc.), and certain textural characteristics (such as: closeness, complexity Degree, bright variation etc.) in otherness it is larger, present invention design using two-dimensional entropy, roughness, contrast as textural characteristics, Extraction effect figure is as shown in Figure 2.In addition to this, consider that textural characteristics by the variation of window size, are retouched global and local The ability of stating will receive influence, and the present invention extracts feature using multiple dimensioned window, and row guarantees special at Analysis On Multi-scale Features figure group Sign describes the comprehensive of power, as shown in Figure 3.In terms of structure feature, the present invention does not characterize structure, but figure Atural object edge is come out using Canny operator extraction as in, to information of the prominent structural information in network inputs image set Specific gravity draws as shown in figure 4, can be very good that structure feature is avoided to be covered in deep learning training process by redundancy Leading abstract structure feature in successive depths network characterization training extraction process can more preferably extract.
3. model training
The present invention refers to the full convolution semantic segmentation model of thought designed, designed deep learning of DeepLab, and Deeplab is complete A kind of derivative improvement network of convolutional network, network design figure are as shown in Figure 5.Inventive network model can be divided into two sections, right In DeepLab (feature extraction and gradually down-sampled while extracting semantic feature should be carried out using porous convolution) downwards and upward (gradually up-sampling characteristic recovery detailed information) two sections of accesses.Wherein downward access is improved based on VGGnet, net Network structure is as shown in Figure 6.By carrying out porous convolution operation to multichannel image and characteristic pattern group, by 6 layers of convolution sum pond Feature temperature figure (16 × 16 dimension image) is obtained later as the 7th layer, up-sampling is carried out to the 7th layer of warp lamination and is restored to the Low-resolution image is switched to high-resolution by interpolation method by 6 layers of identical size, after up-sampling to the 7th layer and more than the 5th layer Output fusion after the convolution of hole generates the 8th layer；8th layer up-sampled after fusion the 6th layer of porous convolution generate the 9th layer i.e. finally Classification results, as shown in Figure 7.The loss that classification results individual element is calculated to more classification regression model softmax classification, obtains Each target calculates penalty values, and penalty values are ranked up, and selects penalty values a target of maximum preceding B (empirical value) as hardly possible Example sample feeds back into full convolutional neural networks model then by the penalty values of these difficult example samples, uses stochastic gradient descent Method updates the parameter of full convolutional neural networks model.Image is marked for each width remote sensing terrain classification, was trained according to above-mentioned The parameter that journey constantly updates the full convolutional neural networks in region is used for obtain the full convolutional neural networks model of terrain classification In subsequent atural object classification task.
4, terrain classification
It is as shown in Figure 8 that the present invention carries out cutting to original wide cut remote sensing image first, it is assumed that resolution ratio X is arranged narrow Image size L uses overlapping cutting method using the 0.5*L of narrow image shorter edge length as superposition image prime number, will then cut Image is classified by depth model after point, and classification results are finally merged together production wide cut testing result.Work as overlapping When area detection result generates contradiction, retain the result for being wherein classified as non-background pixel as final amalgamation result such as Fig. 9 It is shown.
In order to verify effectiveness of the invention, the data set that we use make first ourselves carries out the training of model, so The remote sensing images under the complex scene based on acquisition carry out the contrast verification of human body target classifying quality afterwards.In the present embodiment, It selects Tensorflow frame to realize the Deeplab network architecture, is based on data collection quantity and terrain classification task category Simultaneously training pattern parameter is initialized, the model for terrain classification is finally obtained
The present invention realizes remote sensing terrain classification end to end, using kappa coefficient and hands over and carries out measurement index than IOU, Wherein kappa coefficient indicates that certain class atural object mistake assigns in other atural object classifications pixel accounting and how many pixel mistakes with being divided into certain class The discrete overall merit of object is handed over and indicates that the sum of all pixels for being correctly categorized into the atural object classification and mistake are divided into the ground image than IOU The ratio of the sum of prime number and the atural object total pixel number.The final kappa coefficient 83% of the present invention is handed over and is 81% than IOU, compares it His deep learning method such as VGG (kappa coefficient 75%%, hand over and than IOU be 71%), resnet50 (kappa coefficient 81%, hand over and than IOU be 78%), resnet101 (kappa coefficient 77%, hand over and than IOU be 72%), resnet152 (kappa coefficient 79% is handed over and is 75%) to have larger performance boost than IOU, as shown in Figure 10.Specific different classifications model knot The classifying quality comparison diagram of structure is as shown in figure 11.
Although the illustrative specific embodiment of the present invention is described above, in order to the technology of the art Personnel understand the present invention, it should be apparent that the present invention is not limited to the range of specific embodiment, to the common skill of the art For art personnel, if various change the attached claims limit and determine the spirit and scope of the present invention in, these Variation is it will be apparent that all utilize the innovation and creation of present inventive concept in the column of protection.

Claims (2)

1. a kind of remote sensing image terrain classification method based on deep learning semantic segmentation network, it is characterised in that including following step It is rapid:
(1) remote sensing images for acquiring the High Resolution Visible Light of different loads, mark atural object in each image pixel-by-pixel Note, the data after mark form a width bianry image, original remote sensing images and corresponding mark image be packaged composing training collection and Test set；
(2) to remote sensing images original in training set using multiple dimensioned window to two-dimensional entropy, roughness and contrast textural characteristics into Row extracts, and forms Analysis On Multi-scale Features figure group；And use Canny operator by atural object edge extracting remote sensing images original in training set Out, structure feature figure is formed；
(3) based on the thought of DeepLab, the full convolution semantic segmentation model of deep learning is constructed；
(4) the Analysis On Multi-scale Features figure group generated in step (2), structure feature figure are combined with original remote sensing images to be formed it is defeated Enter figure group, the input as the full convolution semantic segmentation model of deep learning in step (3) carries out model training, finally obtains parameter Stable model；
(5) cutting is carried out to original remote sensing images to be sorted in test set, by image after cutting by training in step (4) The model of parameter stability classify, classification results are merged together and generate wide cut testing result, when overlapping region is detected As a result generate contradiction when, retain wherein be classified as non-background pixel as a result, obtaining final amalgamation result.
2. a kind of remote sensing image terrain classification method based on deep learning semantic segmentation network according to claim 1, It is characterized in that, full convolution semantic segmentation model in step (3) specifically:
Model is divided into downwardly and upwardly two sections, wherein downward access changes the original 13 layers of convolutional layer of VGGnet according to classical strength It is 6 layers, input layer obtains 16 × 16 dimensional feature temperature figures as the 7th layer after 6 layers of convolution sum pondization；In upward access, Interpolation up-sampling is carried out to the 7th layer of warp lamination to restore extremely to merge the 5th after up-sampling the 7th layer with the 6th layer of identical size The porous convolution of layer generates the 8th layer；The 6th layer of porous convolution is merged after up-sampling to the 8th layer generates the 9th layer, it is defeated to the 9th layer Size change over is carried out out and is restored to original remote sensing images size, obtains final classification result.
CN201811130333.0A 2018-09-27 2018-09-27 Remote sensing image ground feature classification method based on deep learning semantic segmentation network Active CN109255334B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811130333.0A CN109255334B (en) 2018-09-27 2018-09-27 Remote sensing image ground feature classification method based on deep learning semantic segmentation network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811130333.0A CN109255334B (en) 2018-09-27 2018-09-27 Remote sensing image ground feature classification method based on deep learning semantic segmentation network

Publications (2)

Publication Number Publication Date
CN109255334A true CN109255334A (en) 2019-01-22
CN109255334B CN109255334B (en) 2021-12-07

Family

ID=65047834

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811130333.0A Active CN109255334B (en) 2018-09-27 2018-09-27 Remote sensing image ground feature classification method based on deep learning semantic segmentation network

Country Status (1)

Country Link
CN (1) CN109255334B (en)

Cited By (46)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109800736A (en) * 2019-02-01 2019-05-24 东北大学 A kind of method for extracting roads based on remote sensing image and deep learning
CN109872331A (en) * 2019-01-30 2019-06-11 天津大学 A kind of remote sensing image data automatic recognition classification method based on deep learning
CN109948517A (en) * 2019-03-18 2019-06-28 长沙理工大学 A kind of high-resolution remote sensing image semantic segmentation method based on intensive full convolutional network
CN110009637A (en) * 2019-04-09 2019-07-12 北京化工大学 A kind of Remote Sensing Image Segmentation network based on tree structure
CN110110599A (en) * 2019-04-03 2019-08-09 天津大学 A kind of Remote Sensing Target detection method based on multi-scale feature fusion
CN110197182A (en) * 2019-06-11 2019-09-03 中国电子科技集团公司第五十四研究所 Remote sensing image semantic segmentation method based on contextual information and attention mechanism
CN110458218A (en) * 2019-07-31 2019-11-15 北京市商汤科技开发有限公司 Image classification method and device, sorter network training method and device
CN110660023A (en) * 2019-09-12 2020-01-07 中国测绘科学研究院 Video stitching method based on image semantic segmentation
CN110706239A (en) * 2019-09-26 2020-01-17 哈尔滨工程大学 Scene segmentation method fusing full convolution neural network and improved ASPP module
CN110866494A (en) * 2019-11-14 2020-03-06 三亚中科遥感研究所 Optical remote sensing image-based town group extraction method and system
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
CN110991430A (en) * 2020-03-02 2020-04-10 中科星图股份有限公司 Ground feature identification and coverage rate calculation method and system based on remote sensing image
CN111079807A (en) * 2019-12-05 2020-04-28 二十一世纪空间技术应用股份有限公司 Ground object classification method and device
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
CN111259905A (en) * 2020-01-17 2020-06-09 山西大学 Feature fusion remote sensing image semantic segmentation method based on downsampling
CN111274865A (en) * 2019-12-14 2020-06-12 深圳先进技术研究院 Remote sensing image cloud detection method and device based on full convolution neural network
CN111368843A (en) * 2020-03-06 2020-07-03 电子科技大学 Method for extracting lake on ice based on semantic segmentation
CN111428781A (en) * 2020-03-20 2020-07-17 中国科学院深圳先进技术研究院 Remote sensing image ground object classification method and system
CN111582104A (en) * 2020-04-28 2020-08-25 中国科学院空天信息创新研究院 Semantic segmentation method and device for remote sensing image
CN111597861A (en) * 2019-02-21 2020-08-28 中科星图股份有限公司 System and method for automatically interpreting ground object of remote sensing image
CN111753834A (en) * 2019-03-29 2020-10-09 中国水利水电科学研究院 Planting land structure semantic segmentation method and device based on deep neural network
CN111797703A (en) * 2020-06-11 2020-10-20 武汉大学 Multi-source remote sensing image classification method based on robust deep semantic segmentation network
CN111860173A (en) * 2020-06-22 2020-10-30 中国科学院空天信息创新研究院 Remote sensing image ground feature element extraction method and system based on weak supervision
CN111860208A (en) * 2020-06-29 2020-10-30 中山大学 Remote sensing image ground object classification method, system, device and medium based on super pixels
CN111931689A (en) * 2020-08-26 2020-11-13 北京建筑大学 Method for extracting video satellite data identification features on line
CN111951285A (en) * 2020-08-12 2020-11-17 湖南神帆科技有限公司 Optical remote sensing image woodland classification method based on cascade deep convolutional neural network
CN112001214A (en) * 2020-05-18 2020-11-27 天津大学 Land classification method based on high-resolution remote sensing image
CN112001293A (en) * 2020-08-19 2020-11-27 四创科技有限公司 Remote sensing image ground object classification method combining multi-scale information and coding and decoding network
CN112036246A (en) * 2020-07-30 2020-12-04 长安大学 Construction method of remote sensing image classification model, remote sensing image classification method and system
CN112233062A (en) * 2020-09-10 2021-01-15 浙江大华技术股份有限公司 Surface feature change detection method, electronic device, and storage medium
CN112329852A (en) * 2020-11-05 2021-02-05 西安泽塔云科技股份有限公司 Classification method and device for earth surface coverage images and electronic equipment
CN112464732A (en) * 2020-11-04 2021-03-09 北京理工大学重庆创新中心 Optical remote sensing image ground object classification method based on dual-path sparse hierarchical network
CN112464745A (en) * 2020-11-09 2021-03-09 中国科学院计算机网络信息中心 Ground feature identification and classification method and device based on semantic segmentation
CN112507764A (en) * 2019-09-16 2021-03-16 中科星图股份有限公司 Method and system for extracting target object of remote sensing image and readable storage medium
CN112598692A (en) * 2020-12-21 2021-04-02 陕西土豆数据科技有限公司 Remote sensing image segmentation post-processing algorithm based on marked pixel matrix
CN112712087A (en) * 2020-12-29 2021-04-27 哈尔滨工业大学 Remote sensing image ground object semantic segmentation method based on deep convolutional neural network
CN112950655A (en) * 2021-03-08 2021-06-11 甘肃农业大学 Land use information automatic extraction method based on deep learning
CN113239815A (en) * 2021-05-17 2021-08-10 广东工业大学 Remote sensing image classification method, device and equipment based on real semantic full-network learning
CN113298042A (en) * 2021-06-22 2021-08-24 中国平安财产保险股份有限公司 Method and device for processing remote sensing image data, storage medium and computer equipment
CN113361373A (en) * 2021-06-02 2021-09-07 武汉理工大学 Real-time semantic segmentation method for aerial image in agricultural scene
CN113706372A (en) * 2020-06-30 2021-11-26 稿定（厦门）科技有限公司 Automatic cutout model establishing method and system
CN113723411A (en) * 2021-06-18 2021-11-30 湖北工业大学 Feature extraction method and segmentation system for semantic segmentation of remote sensing image
CN113920421A (en) * 2021-07-03 2022-01-11 桂林理工大学 Fast-classification full convolution neural network model
CN114092423A (en) * 2021-11-11 2022-02-25 中国电子科技集团公司第五十四研究所 Intelligent extraction method for remote sensing image information label
US11669990B2 (en) 2021-01-21 2023-06-06 Beijing Baidu Netcom Science And Technology Co., Ltd. Object area measurement method, electronic device and storage medium
CN118378780A (en) * 2024-04-08 2024-07-23 广西壮族自治区自然资源遥感院 Environment comprehensive evaluation method and system based on remote sensing image

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587587A (en) * 2009-07-14 2009-11-25 武汉大学 The segmentation method for synthetic aperture radar images of consideration of multi-scale Markov field
CN102521595A (en) * 2011-12-07 2012-06-27 中南大学 Method for extracting image region of interest based on eye movement data and bottom-layer features
CN106529447A (en) * 2016-11-03 2017-03-22 河北工业大学 Small-sample face recognition method
CN107527352A (en) * 2017-08-09 2017-12-29 中国电子科技集团公司第五十四研究所 Remote sensing Ship Target contours segmentation and detection method based on deep learning FCN networks
CN107564025A (en) * 2017-08-09 2018-01-09 浙江大学 A kind of power equipment infrared image semantic segmentation method based on deep neural network
CN108154192A (en) * 2018-01-12 2018-06-12 西安电子科技大学 High Resolution SAR terrain classification method based on multiple dimensioned convolution and Fusion Features
US20180253980A1 (en) * 2017-03-03 2018-09-06 Farrokh Mohamadi Drone Terrain Surveillance with Camera and Radar Sensor Fusion for Collision Avoidance

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587587A (en) * 2009-07-14 2009-11-25 武汉大学 The segmentation method for synthetic aperture radar images of consideration of multi-scale Markov field
CN102521595A (en) * 2011-12-07 2012-06-27 中南大学 Method for extracting image region of interest based on eye movement data and bottom-layer features
CN106529447A (en) * 2016-11-03 2017-03-22 河北工业大学 Small-sample face recognition method
US20180253980A1 (en) * 2017-03-03 2018-09-06 Farrokh Mohamadi Drone Terrain Surveillance with Camera and Radar Sensor Fusion for Collision Avoidance
CN107527352A (en) * 2017-08-09 2017-12-29 中国电子科技集团公司第五十四研究所 Remote sensing Ship Target contours segmentation and detection method based on deep learning FCN networks
CN107564025A (en) * 2017-08-09 2018-01-09 浙江大学 A kind of power equipment infrared image semantic segmentation method based on deep neural network
CN108154192A (en) * 2018-01-12 2018-06-12 西安电子科技大学 High Resolution SAR terrain classification method based on multiple dimensioned convolution and Fusion Features

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHANGJUNIOR: "从FCN到DeepLab", 《CSDN》 *

Cited By (64)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109872331A (en) * 2019-01-30 2019-06-11 天津大学 A kind of remote sensing image data automatic recognition classification method based on deep learning
CN109800736B (en) * 2019-02-01 2023-07-21 东北大学 Road extraction method based on remote sensing image and deep learning
CN109800736A (en) * 2019-02-01 2019-05-24 东北大学 A kind of method for extracting roads based on remote sensing image and deep learning
CN111597861A (en) * 2019-02-21 2020-08-28 中科星图股份有限公司 System and method for automatically interpreting ground object of remote sensing image
CN109948517A (en) * 2019-03-18 2019-06-28 长沙理工大学 A kind of high-resolution remote sensing image semantic segmentation method based on intensive full convolutional network
CN111753834A (en) * 2019-03-29 2020-10-09 中国水利水电科学研究院 Planting land structure semantic segmentation method and device based on deep neural network
CN111753834B (en) * 2019-03-29 2024-03-26 中国水利水电科学研究院 Planting land block structure semantic segmentation method and device based on deep neural network
CN110110599A (en) * 2019-04-03 2019-08-09 天津大学 A kind of Remote Sensing Target detection method based on multi-scale feature fusion
CN110110599B (en) * 2019-04-03 2023-05-09 天津大学 Remote sensing image target detection method based on multi-scale feature fusion
CN110009637A (en) * 2019-04-09 2019-07-12 北京化工大学 A kind of Remote Sensing Image Segmentation network based on tree structure
CN110197182A (en) * 2019-06-11 2019-09-03 中国电子科技集团公司第五十四研究所 Remote sensing image semantic segmentation method based on contextual information and attention mechanism
CN110458218B (en) * 2019-07-31 2022-09-27 北京市商汤科技开发有限公司 Image classification method and device and classification network training method and device
CN110458218A (en) * 2019-07-31 2019-11-15 北京市商汤科技开发有限公司 Image classification method and device, sorter network training method and device
CN110660023A (en) * 2019-09-12 2020-01-07 中国测绘科学研究院 Video stitching method based on image semantic segmentation
CN112507764A (en) * 2019-09-16 2021-03-16 中科星图股份有限公司 Method and system for extracting target object of remote sensing image and readable storage medium
CN110706239B (en) * 2019-09-26 2022-11-11 哈尔滨工程大学 Scene segmentation method fusing full convolution neural network and improved ASPP module
CN110706239A (en) * 2019-09-26 2020-01-17 哈尔滨工程大学 Scene segmentation method fusing full convolution neural network and improved ASPP module
CN110866494B (en) * 2019-11-14 2022-09-06 三亚中科遥感研究所 Urban group extraction method and system based on optical remote sensing image
CN110866494A (en) * 2019-11-14 2020-03-06 三亚中科遥感研究所 Optical remote sensing image-based town group extraction method and system
CN111079807B (en) * 2019-12-05 2023-07-07 二十一世纪空间技术应用股份有限公司 Ground object classification method and device
CN111079807A (en) * 2019-12-05 2020-04-28 二十一世纪空间技术应用股份有限公司 Ground object classification method and device
CN111274865B (en) * 2019-12-14 2023-09-19 深圳先进技术研究院 Remote sensing image cloud detection method and device based on full convolution neural network
CN111274865A (en) * 2019-12-14 2020-06-12 深圳先进技术研究院 Remote sensing image cloud detection method and device based on full convolution neural network
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
CN111259905A (en) * 2020-01-17 2020-06-09 山西大学 Feature fusion remote sensing image semantic segmentation method based on downsampling
CN111259905B (en) * 2020-01-17 2022-05-31 山西大学 Feature fusion remote sensing image semantic segmentation method based on downsampling
CN110991430A (en) * 2020-03-02 2020-04-10 中科星图股份有限公司 Ground feature identification and coverage rate calculation method and system based on remote sensing image
CN111368843B (en) * 2020-03-06 2022-06-10 电子科技大学 Method for extracting lake on ice based on semantic segmentation
CN111368843A (en) * 2020-03-06 2020-07-03 电子科技大学 Method for extracting lake on ice based on semantic segmentation
CN111428781A (en) * 2020-03-20 2020-07-17 中国科学院深圳先进技术研究院 Remote sensing image ground object classification method and system
CN111582104B (en) * 2020-04-28 2021-08-06 中国科学院空天信息创新研究院 Remote sensing image semantic segmentation method and device based on self-attention feature aggregation network
CN111582104A (en) * 2020-04-28 2020-08-25 中国科学院空天信息创新研究院 Semantic segmentation method and device for remote sensing image
CN112001214A (en) * 2020-05-18 2020-11-27 天津大学 Land classification method based on high-resolution remote sensing image
CN111797703A (en) * 2020-06-11 2020-10-20 武汉大学 Multi-source remote sensing image classification method based on robust deep semantic segmentation network
CN111860173A (en) * 2020-06-22 2020-10-30 中国科学院空天信息创新研究院 Remote sensing image ground feature element extraction method and system based on weak supervision
CN111860173B (en) * 2020-06-22 2021-10-15 中国科学院空天信息创新研究院 Remote sensing image ground feature element extraction method and system based on weak supervision
CN111860208B (en) * 2020-06-29 2023-10-24 中山大学 Super-pixel-based remote sensing image ground object classification method, system, device and medium
CN111860208A (en) * 2020-06-29 2020-10-30 中山大学 Remote sensing image ground object classification method, system, device and medium based on super pixels
CN113706372A (en) * 2020-06-30 2021-11-26 稿定（厦门）科技有限公司 Automatic cutout model establishing method and system
CN112036246A (en) * 2020-07-30 2020-12-04 长安大学 Construction method of remote sensing image classification model, remote sensing image classification method and system
CN112036246B (en) * 2020-07-30 2021-08-24 长安大学 Construction method of remote sensing image classification model, remote sensing image classification method and system
CN111951285A (en) * 2020-08-12 2020-11-17 湖南神帆科技有限公司 Optical remote sensing image woodland classification method based on cascade deep convolutional neural network
CN112001293A (en) * 2020-08-19 2020-11-27 四创科技有限公司 Remote sensing image ground object classification method combining multi-scale information and coding and decoding network
CN111931689A (en) * 2020-08-26 2020-11-13 北京建筑大学 Method for extracting video satellite data identification features on line
CN112233062A (en) * 2020-09-10 2021-01-15 浙江大华技术股份有限公司 Surface feature change detection method, electronic device, and storage medium
CN112464732A (en) * 2020-11-04 2021-03-09 北京理工大学重庆创新中心 Optical remote sensing image ground object classification method based on dual-path sparse hierarchical network
CN112329852A (en) * 2020-11-05 2021-02-05 西安泽塔云科技股份有限公司 Classification method and device for earth surface coverage images and electronic equipment
CN112464745A (en) * 2020-11-09 2021-03-09 中国科学院计算机网络信息中心 Ground feature identification and classification method and device based on semantic segmentation
CN112464745B (en) * 2020-11-09 2023-07-07 中国科学院计算机网络信息中心 Feature identification and classification method and device based on semantic segmentation
CN112598692A (en) * 2020-12-21 2021-04-02 陕西土豆数据科技有限公司 Remote sensing image segmentation post-processing algorithm based on marked pixel matrix
CN112712087A (en) * 2020-12-29 2021-04-27 哈尔滨工业大学 Remote sensing image ground object semantic segmentation method based on deep convolutional neural network
US11669990B2 (en) 2021-01-21 2023-06-06 Beijing Baidu Netcom Science And Technology Co., Ltd. Object area measurement method, electronic device and storage medium
CN112950655A (en) * 2021-03-08 2021-06-11 甘肃农业大学 Land use information automatic extraction method based on deep learning
CN113239815B (en) * 2021-05-17 2022-09-06 广东工业大学 Remote sensing image classification method, device and equipment based on real semantic full-network learning
CN113239815A (en) * 2021-05-17 2021-08-10 广东工业大学 Remote sensing image classification method, device and equipment based on real semantic full-network learning
CN113361373A (en) * 2021-06-02 2021-09-07 武汉理工大学 Real-time semantic segmentation method for aerial image in agricultural scene
CN113723411B (en) * 2021-06-18 2023-06-27 湖北工业大学 Feature extraction method and segmentation system for semantic segmentation of remote sensing image
CN113723411A (en) * 2021-06-18 2021-11-30 湖北工业大学 Feature extraction method and segmentation system for semantic segmentation of remote sensing image
CN113298042A (en) * 2021-06-22 2021-08-24 中国平安财产保险股份有限公司 Method and device for processing remote sensing image data, storage medium and computer equipment
CN113298042B (en) * 2021-06-22 2024-02-02 中国平安财产保险股份有限公司 Remote sensing image data processing method and device, storage medium and computer equipment
CN113920421A (en) * 2021-07-03 2022-01-11 桂林理工大学 Fast-classification full convolution neural network model
CN114092423A (en) * 2021-11-11 2022-02-25 中国电子科技集团公司第五十四研究所 Intelligent extraction method for remote sensing image information label
CN118378780A (en) * 2024-04-08 2024-07-23 广西壮族自治区自然资源遥感院 Environment comprehensive evaluation method and system based on remote sensing image

Also Published As

Publication number Publication date
CN109255334B (en) 2021-12-07

Similar Documents

Publication Publication Date Title
CN109255334A (en) Remote sensing image terrain classification method based on deep learning semantic segmentation network
CN107016405B (en) A kind of pest image classification method based on classification prediction convolutional neural networks
CN105512661B (en) A kind of Remote Image Classification based on multi-modal Fusion Features
KR102102161B1 (en) Method, apparatus and computer program for extracting representative feature of object in image
CN105488517B (en) A kind of vehicle brand type identifier method based on deep learning
CN111428781A (en) Remote sensing image ground object classification method and system
CN113065558A (en) Lightweight small target detection method combined with attention mechanism
CN107133955B (en) A kind of collaboration conspicuousness detection method combined at many levels
CN106372648A (en) Multi-feature-fusion-convolutional-neural-network-based plankton image classification method
CN107292343A (en) A kind of Classification of hyperspectral remote sensing image method based on six layers of convolutional neural networks and spectral space information consolidation
CN110287849A (en) A kind of lightweight depth network image object detection method suitable for raspberry pie
CN103106265B (en) Similar image sorting technique and system
CN104462494B (en) A kind of remote sensing image retrieval method and system based on unsupervised feature learning
CN108960404B (en) Image-based crowd counting method and device
CN107657257A (en) A kind of semantic image dividing method based on multichannel convolutive neutral net
CN106919920A (en) Scene recognition method based on convolution feature and spatial vision bag of words
CN108520114A (en) A kind of textile cloth defect detection model and its training method and application
CN110399819A (en) A kind of remote sensing image residential block extraction method based on deep learning
CN105955708A (en) Sports video lens classification method based on deep convolutional neural networks
CN106023224A (en) PCNN automatic segmentation method for microscopic image of traditional Chinese medicine
CN109670405A (en) A kind of complex background pedestrian detection method based on deep learning
CN112232328A (en) Remote sensing image building area extraction method and device based on convolutional neural network
CN106960176A (en) A kind of pedestrian's gender identification method based on transfinite learning machine and color characteristic fusion
CN106875481A (en) A kind of preparation method of three-dimensional visualization remote sensing image Surface classification model
CN107480679A (en) The road network extracting method of classification and connected component analysis based on convolutional neural networks

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