CN107194405A - A kind of method that interactive semi-automatic high-resolution remote sensing image building is extracted - Google Patents
A kind of method that interactive semi-automatic high-resolution remote sensing image building is extracted Download PDFInfo
- Publication number
- CN107194405A CN107194405A CN201710305968.9A CN201710305968A CN107194405A CN 107194405 A CN107194405 A CN 107194405A CN 201710305968 A CN201710305968 A CN 201710305968A CN 107194405 A CN107194405 A CN 107194405A
- Authority
- CN
- China
- Prior art keywords
- building
- line
- remote sensing
- extracted
- sensing image
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 17
- 238000001514 detection method Methods 0.000 claims abstract description 13
- 230000011218 segmentation Effects 0.000 claims description 25
- 238000004422 calculation algorithm Methods 0.000 claims description 23
- 238000000605 extraction Methods 0.000 description 18
- 239000000284 extract Substances 0.000 description 16
- 238000010586 diagram Methods 0.000 description 7
- 241000196324 Embryophyta Species 0.000 description 6
- 238000003709 image segmentation Methods 0.000 description 6
- 230000003993 interaction Effects 0.000 description 5
- 238000012946 outsourcing Methods 0.000 description 4
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000000638 solvent extraction Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 241000270295 Serpentes Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of method that interactive semi-automatic high-resolution remote sensing image building is extracted, it is first determined the approximate range of building；Then the straight line in the range of detection building, obtains building principal direction and rotates building to horizontal direction；Then building figure spot is obtained；Finally according to building color characteristic to building regularization, to obtain accurate contour of building.The present invention realize by user on independent building setting-out rapidly and accurately to extract contour of building.
Description
Technical field
It is more particularly to a kind of new based on super-pixel segmentation the invention belongs to image procossing and target identification technology field
The method extracted with the interactive semi-automatic high-resolution remote sensing image building of regularization.
Background technology
Building is estimated as one of geographical space staple in city construction planning and building, change detection and the density of population
The fields such as meter, in occupation of consequence, are very important class targets in remote sensing image interpretation.With high spatial resolution
Remote sensing (remote sensing of abbreviation high score) satellite is developed rapidly with aviation image acquiring technology, and high-resolution remote sensing image information content is huge
Greatly, comprising atural object minutia, geometry, space characteristics, the information such as shape and textural characteristics can be retouched relatively precisely
State typical feature so that accurate identification is carried out to building using high-resolution remote sensing image and is possibly realized.At present, from high-resolution
The method of building is automatically extracted in remote sensing image to be had：Based on cutting techniques ([document 1-2]) and based on angle point, straight line and shade
The method ([document 3-5]) of feature.Wherein the extracting method based on cutting techniques makes full use of the spectrum, texture, shape of object
Etc. feature, building is extracted using object-oriented cutting techniques, but due to the influence of the factors such as noise, illumination and reflectivity, lead
Cause the building edge and corner that obtain inaccurate.Method combination Corner Detection and image segmentation extraction based on Corner Feature are built
Build thing；Method based on linear feature extracts the line segment on image by using line detection algorithm, then judges linear relation
Automatically extract building ([document 6-9])；Method based on shadow character is uniform zonule first by picture breakdown, then
According to the position of shade, fusion has the zonule of similar spectral feature so that its shape is approximately rectangle, so as to extract building
([document 10,11]).However, this method based on low-level feature is only applicable to, house is independent, and body is simply and between atural object
The few situation of masking.Due to the complexity of remote sensing image, building by noise, block, shade, low contrast are influenceed, it is complete from
The dynamic method for extracting building extraction can not obtain reliable result, and especially border precision is had higher requirements at some
Application field.
For this problem, a kind of mode feasible at present is to be combined Computer Automatic Extraction and man-machine interactively, i.e., real
Existing interactive mode Objects extraction.In remote sensing image building extraction process, the initial letter on the one hand making full use of operator to provide
Breath, while playing the advantage of computer disposal image, both, which combine, improves the efficiency that building is extracted.Some scholars are to interactive mode
Extract building to be studied, such as：The right angle flat-top semi-automatic building extraction side that geometrical constraint and Image Segmentation are combined
Method ([document 12])；Semi-automatic Building extracting method ([document based on object space geometrical constraint Least squares matching
13]), this method specifies house initial position by user, and building edge and object space geometrical model are obtained through algorithm process
Optimum Matching；Using two initial points being manually specified, the linear ground object of point-to-point transmission is extracted by least squares template matching
([document 14])；Building is extracted in joint Snake and Dynamic Programming, and user only needs to specify several seed points to build in Fang Jiaochu
Build thing Position Approximate to indicate, you can extract building precise boundary ([document 15]).Above method is by artificial
The initial position in given house, relies primarily on edge extraction building, but interaction is complicated, needs user to give the standard of building
True position, therefore application is limited.
Based on above-mentioned analysis, while considering that building is mostly this feature, this hair of rectangle on high-resolution remote sensing image
Computer Automatic Extraction and man-machine interactively are combined by bright proposition, and by user, one line drawing of simple graph goes out to build on building
Thing is built, compared to existing interactive semiautomatic extraction method, exact position of this method interaction simply without specifying building, you can
Contour of building is fast and accurately extracted, so as to improve the efficiency of interactive semi-automatic extraction building.
Bibliography：
[1]Blaschke,T.and J.Strobl,What's wrong with pixelsSome recent
developments interfacing remote sensing and GIS.GIS-Zeitschrift für
Geoinformationssysteme,2001.14(6):p.12-17.
[2]Kropatsch,W.G.and S.Ben Yacoub.A revision of pyramid
segmentation.in Pattern Recognition,1996.,Proceedings of the 13th
International Conference on.1996.
[3]Huertas A,Nevatia R.Detecting buildings in aerial images☆[J]
.Computer Vision Graphics&Image Processing,1988,41(2):131-152.
[4]Song,Y.,&Shan,J.(2010).Building extraction from high resolution
color imagery based on edge flow driven active contour and jseg.1.Blaschke,
T.and J.Strobl,What's wrong with pixelsSome recent developments interfacing
remote sensing and GIS.GIS-Zeitschrift für Geoinformationssysteme,2001.14(6):
p.12-17.
[5]Li J,Wang K,Zhang Z,et al.A method of building information
extraction based on mathematical morphology and multiscale[C]//International
Conference on Intelligent Earth Observing and Applications.2015:98082S.
[6]Wang J,Yang X,Qin X,et al.An Efficient Approach for Automatic
Rectangular Building Extraction From Very High Resolution Optical Satellite
Imagery[J].IEEE Geoscience&Remote Sensing Letters,2015,12(3):487-491.
[7]A.CroitoruDoytsher Y.Right-Angle Rooftop Polygon Extraction in
Regularised Urban Areas:Cutting the Corners[J].Photogrammetric Record,2004,19
(108):311–341.
[8]Ngo T T,Collet C,Mazet V.Automatic rectangular building detection
from VHR aerial imagery using shadow and image segmentation[C]//IEEE
International Conference on Image Processing.IEEE,2015:1483-1487.
[9] Pang Chihai, Li Guangyao, Zhao Jie waits contour of building in satellite photoes of the based on line detection algorithm to extract
[J] computer applications, 2008,28 (b06):190-192.
[10] Zhou Shaoguang, Sun Jinyan, all jasmines wait the contour of building information extracting method of high-resolution remote sensing images
[J] land resources remote sensing, 2015,27 (3):52-58.
[11]Singh G,Jouppi M,Zhang Z,et al.Shadow based building extraction
from single satellite image[J].Proceedings of SPIE-The International Society
for Optical Engineering,2015,9401:94010F-94010F-15.
[12] Zhang Yu, Zhang Zuxun geometrical constraints extract [J] Wuhan with the quick House Semi-automatic that Image Segmentation is combined
College journal information science version, 2000,25 (3):238-242.
[13] buildings of Zhang Zuxun, Zhang Jianqing, the Hu Xiangyun based on object space geometrical constraint Least squares matching half is certainly
Dynamic extracting method [J] Wuhan University Journals (information science version), 2001,26 (4):290-295.
[14] the Chinese image graphics of semi-automatic extraction [J] of linear ground object on Hu Xiangyun, Zhang Zuxun aviation images
Report:,2002,7(2):137-140.
[15]Fazan A J,Poz A P D.Rectilinear building roof contour extraction
based on snakes and dynamic programming[J].International Journal of Applied
Earth Observation&Geoinformation,2013,25(3):1–10.
The content of the invention
Requirement and existing semi-automatic friendship can not be met for automatically extracting building from high-resolution remote sensing image at present
The problem of mutual formula extracting method, the present invention proposes building with rectangular on a kind of interactive semi-automatic extraction high-resolution remote sensing image
The method of thing, realize by user on independent building setting-out rapidly and accurately to extract contour of building.
The technical solution adopted in the present invention is：What a kind of interactive semi-automatic high-resolution remote sensing image building was extracted
Method, it is characterised in that comprise the following steps：
Step 1：Determine the approximate range of building；
Step 2：The straight line in the range of building is detected, building principal direction is obtained and rotates building to level side
To；
Step 3：Obtain building figure spot；
Step 4：According to the color characteristic of building to building regularization, to obtain accurate contour of building.
Relative to prior art, the beneficial effects of the invention are as follows：
Present invention interaction is simple, it is only necessary to one line of simple graph on building, you can fast and accurately extract building
Profile, so as to improve the efficiency of interactive semi-automatic extraction building.
Brief description of the drawings
Fig. 1 is the approximate location schematic diagram of the building of the embodiment of the present invention；
Fig. 2 rotates schematic diagram for the principal direction detection of the embodiment of the present invention with image, wherein (a) user mutual, (b)
The line segment that EDLines algorithms are extracted, (c) seed linear distance figure, (d) building principal direction, (e) building is rotated to level side
To；
Fig. 3 is the SLIC cluster seeking schematic diagrams of the embodiment of the present invention, wherein (a) searches for view picture image, (b) search 2S X
2S scopes；
Fig. 4 is the SLIC over-segmentation striographs of the embodiment of the present invention；
Fig. 5 extracts flow chart for the building of the embodiment of the present invention；
Fig. 6 is the acquisition striograph to be split of the embodiment of the present invention；
Fig. 7 rotates schematic diagram for the principal direction detection of the embodiment of the present invention with building, wherein (a) EDLines algorithms are carried
The line segment taken, (b) seed linear distance figure and building principal direction, (c) building are rotated to horizontal direction；
Fig. 8 is the super-pixel pre-segmentation result schematic diagram of the embodiment of the present invention；
Fig. 9 is the GrabCut cutting procedure schematic diagrames of the embodiment of the present invention, wherein (a) is image to be split, (b) is
GrabCut segmentation results；
Figure 10 is the building regularisation procedure schematic diagram of the embodiment of the present invention, wherein (a) building figure spot, (b) figure spot
Regularization, (c) regularization result；
Figure 11 extracts result exemplary plot for the aviation image building of the embodiment of the present invention, wherein (a) plants sub-line, (b) is carried
Result is taken, (c) plants sub-line, and (d) extracts result；
Figure 12 extracts result exemplary plot for the satellite image building of the embodiment of the present invention, wherein (a) plants sub-line, (b) is carried
Result is taken, (c) plants sub-line, and (d) extracts result.
Embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, below in conjunction with the accompanying drawings and embodiment is to this hair
It is bright to be described in further detail, it will be appreciated that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
The method that a kind of interactive semi-automatic high-resolution remote sensing image building that the present invention is provided is extracted, passes through user
Interaction, principal direction detection, super-pixel segmentation, building patch obtain with the several steps of regularization to complete high-definition remote sensing shadow
As building is extracted.
Step 1：The approximate range of building is determined by user mutual.
Application claims user specifies kind of a sub-line, has some to require to kind of a sub-line：1) trend for planting sub-line is building pair
Linea angulata；2) seed line length is building cornerwise 2/3rds or so, 3) sub-line two-end-point is planted on building center pair
Claim.The Position Approximate of target structures thing is can determine that based on above-mentioned 3 points, and using the image in the region as image to be split, is such as schemed
Shown in 1.
Wherein building upper curve is the kind sub-line of user mutual, and Wseed and Hseed are respectively kind of a sub-line outsourcing rectangle S1
Wide and height, the approximate location S2 of building is a square centered on S1, and Wbuilding is the length of side of building,
Calculation is as follows：
Wbuilding=2*max (Wseed,Wseed) (1)
Step 2：The straight line in the range of building is detected using line detection algorithm, is obtained by counting line segment direction histogram
Obtain building principal direction and rotate building to horizontal direction；
The adjacent edge of Rectangle building is mutually perpendicular to, therefore building has two mutually perpendicular principal directions, therefore available line
Section direction histogram determines the principal direction of building, and building then is rotated into horizontal direction so that building is in horizontal stroke
Flat vertical state, is easy to segmentation.EDLines is a kind of line detection algorithm automatic, quick and without any parameter adjustment,
Line Algorithm is extracted relative to Hough transform and Lsd, the straight line on image can be more accurately extracted.Therefore, this hair
Bright use EDLines extracts the line segment L=(l1, l2 ... ..., ln) on view picture image, generally plants sub-line and is in building
The center of thing, the line segment separation of extraction is bigger in the probability on building from building center more peri position, therefore is carrying out
The nearer line segment of kind of sub-line of being adjusted the distance during the ballot of direction assigns bigger weights.All line segment midpoints are calculated in the range of building to building
The distance at Zhu Wu centers, assigns weights, line segment direction histogram value mode is as follows using distance map to line segment：
Wherein θ ∈ [0,180) represent line segment direction, while also representing histogram transverse axis, nθIt is that histogram direction value is θ
The intensity at place, li represents a line segment, and Len (li) represents line segment li length, D representative species sub-line normalized cumulant figures, D (x,
Y) distance value at point (x, y) place is then represented.(xli,yli) represent line segment li middle point coordinates.(3) formula of utilization calculates building
Principal direction, is then rotated raw video to horizontal direction using bilinear interpolation so that building is in smooth vertical and horizontal shape
State, as shown in Figure 2：
θM=arg max (nθ+nθ+90)θ∈[0,90] (3)
To improve the effect and performance of high-resolution remote sensing image, over-segmentation is frequently used to the step as a pretreatment
Suddenly, the present invention utilizes simple linear Iterative Clustering (SLIC) to raw video pre-segmentation.SLIC algorithms are a kind of realization sides
Just super-pixel algorithm, it is time-consuming less, need user's regulation parameter few, pre-segmentation can obtain a series of having similar features and not breaking
The uniform super-pixel of bad image boundary information.It closes on journey by consider similarity and locus of the pixel on color
Degree obtains super-pixel to carry out cluster.In a quintuple space complete super-pixel Local Clustering, this quintuple space be by
[l, a, b] value and pixel coordinate value [x, y] composition of CIELAB color spaces.It is similar using Fusion of Color in quintuple space
The normalized cumulant of degree and picture element position information is measured to carry out the cluster to pixel.Algorithm flow is as follows：
1) initial cluster center is given.In the image that a width has N number of pixel, it is assumed that the super-pixel number of pre-segmentation is
K, the size of each super-pixel is exactly N/K pixel, and the distance between each super-pixel center isSelect pixel
Center is：CK=[lk,ak,bk,xk,yk] k∈[1,K]
2) cluster centre is moved at the gradient of neighborhood minimum.In order to avoid edge of the cluster centre just at image
Position, and the possibility for choosing noise spot is reduced simultaneously, cluster centre is moved at the gradient of the neighborhood of 3 X 3 minimum by algorithm.
3) for each cluster centre, each pixel and nearest cluster centre pixel are calculated in 2S X 2S neighborhoods
Between similarity.Similarity between its pixel is calculated as follows according to Euclidean distance：
SLIC algorithm cluster seekings are shown in Fig. 3；
4) new cluster centre is calculated, and scans for mark again；
5) until the difference of new cluster centre and original cluster centre is less than threshold value；
It is the SLIC over-segmentation images of the embodiment of the present invention see Fig. 4.
Step 3：Building figure spot is obtained using the GrabCut dividing methods based on super-pixel；
The present invention obtains building figure spot using GrabCut partitioning algorithms.GrabCut algorithms divide the image into problem conversion
For energy function minimization problem, then energy minimization problem is converted into minimal cut problem, minimal cut problem can use maximum
Theory is flowed to solve.It uses colored gauss hybrid models (GMM) statistics substitution histogram that characteristics of image is described, it is considered to
Feature Covariance Plots between sampled point cause segmentation result more accurate.Because its interaction is simple, segmentation precision is high, can
Meet user and obtain satisfied segmentation result by relatively small number of interworking on image.Therefore, the present invention is with super-pixel
Initial segmentation result interacts formula selection target foreground pixel point and background pixel point, is realized using max-flow min-cut algorithm
The segmentation of target and background, i.e. Objective extraction.Energy function is expressed as：
E (α, k, θ, z)=U (α, k, θ, z)+V (α, z) (4)
K=(k1,k1,...,kN), ki∈ { 1,2 ..., K } (i ∈ [1, N]) represents that ith pixel belongs to GMM kthiIt is individual
Gauss model, K represents the number of Gaussian function in GMM.α represents that prospect background is marked, and z is image data, and θ represents GMM ginseng
Number, has：θ=π (α, k), μ (α, k), Σ (α, k) }
Wherein π represents the weight of each Gaussian function in GMM, μ, and Σ represents the average and covariance matrix of Gaussian function respectively.
Definition of data item is：One pixel is classified as the punishment of background or target, and form of calculation is as follows：
Border item is defined as：Border item embodies discontinuous punishment between neighborhood territory pixel m and n, if two neighborhood territory pixels are poor
Other very little, then they belong to same target or the possibility of same background is just very big, if difference is very big, illustrate this
Two pixels are likely in the marginal portion of target and background, therefore the possibility being partitioned from is than larger, therefore when two adjacent
Domain pixel differences are bigger, and energy is smaller.In rgb space, the similitude of two pixels is weighed, it is fixed using Euclidean distance (two norms)
Justice is as follows：
V (α, z)=γ ∑s [αn≠αm]exp-β||zm≠zn||2 (7)
GrabCut algorithm energy is to reach minimum by iteration, and the process of each iteration all causes foreground and background to build
The GMM parameters of mould are more excellent, and image segmentation result is more excellent, and three steps of algorithm point complete image segmentation：Initialization, iteration are minimized and used
Family is interacted.
1) initialize, user specifies a rectangle frame for including target on image, and the pixel of outer frame is all as background
Pixel, the pixel of inframe estimates the GMM of prospect background using these pixels respectively all as the pixel of " possible target ".
2) iteration is minimized, the Gaussian component distributed in GMM each pixel, recalculates prospect background GMM parameters, is led to
The Gibbs energy terms for crossing (1) formula build figure, and energy function is minimized using min-cut/max-flow methods, perform and extremely restrain
Untill, obtain initial segmentation estimation.
3) user is edited, if user is dissatisfied to segmentation result, and prospect background execution can be reassigned on segmentation result
Step 2), whole iterative process is repeated, untill user is satisfied with.
The present invention expands kind of sub-line outsourcing rectangle in the rectangle provided as user in algorithm initialization, and the rectangle will be built
Thing is included, and foreground pixel is the corresponding super-pixel of kind of sub-line, and background pixel is the super-pixel outside rectangle, utilizes preceding background picture
Element builds GMM model, realizes that building is split by min-cut/max-flow methods.
Step 4：According to the color characteristic of building to building regularization, to obtain accurate contour of building.
Followed by specific embodiment, the present invention is further elaborated；
1st, building extracts flow description；
Man-machine interactively formula building extraction based on super-pixel segmentation is detected by user mutual, principal direction and image rotates,
Completed based on super-pixel Image Segmentation, four steps of figure spot regularization, particular flow sheet is as shown in Figure 5；
2nd, implementation process；
(1) user mutual；
Because raw video is larger, the present invention extracts single building, therefore individually building is extracted in segmentation on small range image
Thing, determines the probable ranges of building by user's setting-out, and image to be split size.As shown in fig. 6, wherein building
Upper curve represents kind of a sub-line, and architecture enclosing square is building probable ranges, and the present invention is used building probable ranges
Expand twice image as image to be split.
(2) principal direction detection rotates with image；
Straight-line detection is carried out to image to be split using EDLines line detection algorithms, calculates in the range of building and owns
Weights are assigned to line segment using distance map, line segment direction histogram are counted, so that really to the distance at building center in line segment midpoint
Building principal direction is determined, as a result see Fig. 7；
(3) super-pixel segmentation
Super-pixel object is obtained to image over-segmentation to be split using SLIC super-pixel segmentations algorithm, as a result see Fig. 8；
(4) the GrabCut images segmentation based on super-pixel
The present invention interacts formula selection building target foreground pixel point with super-pixel initial segmentation result, will plant sub-line
Boundary rectangle expand as partitioning algorithm initialization procedure need user provide rectangle, foreground pixel point be and kind of a sub-line pixel
The super-pixel of the identical label of point, background pixel point is the super-pixel outside rectangle, and target and the back of the body are realized using GrabCut partitioning algorithms
The segmentation of scape, i.e. building are extracted.S1 is the boundary rectangle of kind of sub-line in Fig. 9 (a), and S2 is the rectangle comprising building, is schemed (b)
It is the building figure spot that segmentation is obtained.
(5) figure spot regularization；
Due to the complexity of remote sensing image, building periphery atural object is similar to building, therefore splits calculation merely with image
The building figure spot that method is obtained, is not building complete on image, it is therefore desirable to carry out regularization.In view of building sheet
The color of body and the difference of periphery atural object color, the present invention calculate pixel l, a, b in the outsourcing rectangle each edge of figure spot respectively
Similarity (being calculated using Euclidean distance) between l, a, b average value for all pixels that average value and building are included, setting
The four edges of outsourcing rectangle are translated until reaching threshold value by certain similarity threshold (empirical value is 0.75), so that
To complete building, regularisation procedure is as shown in Figure 10, wherein the green rectangle in (a) is building figure spot boundary rectangle,
(b) red line is the line segment after green line is translated in, and (c) is the building after regularization.
Extract and test using aerial image and satellite image as the different buildings of data source progress respectively, original shadow
Picture and result is extracted as is illustrated by figs. 11 and 12, test result indicates that the present invention proposes the side of interactive semi-automatic extraction building
Method can rapidly and accurately extract high-resolution remote sensing image contour of building.
It should be appreciated that the part that this specification is not elaborated belongs to prior art.
It should be appreciated that the above-mentioned description for preferred embodiment is more detailed, therefore it can not be considered to this
The limitation of invention patent protection scope, one of ordinary skill in the art is not departing from power of the present invention under the enlightenment of the present invention
Profit is required under protected ambit, can also be made replacement or be deformed, each fall within protection scope of the present invention, this hair
It is bright scope is claimed to be determined by the appended claims.
Claims (4)
1. a kind of method that interactive semi-automatic high-resolution remote sensing image building is extracted, it is characterised in that including following step
Suddenly：
Step 1：Determine the approximate range of building；
Step 2：The straight line in the range of building is detected, building principal direction is obtained and rotates building to horizontal direction；
Step 3：Obtain building figure spot；
Step 4：According to the color characteristic of building to building regularization, to obtain accurate contour of building.
2. the method that interactive semi-automatic high-resolution remote sensing image building according to claim 1 is extracted, its feature
It is：In step 1, the approximate range of building is determined by user mutual；User specifies kind of a sub-line first, wherein planting sub-line
Move towards as building diagonal, seed line length is building cornerwise 2/3rds or so, plant sub-line two-end-point on building
Build thing Central Symmetry.
3. the method that interactive semi-automatic high-resolution remote sensing image building according to claim 1 is extracted, its feature
It is：In step 2, the straight line in the range of building is detected using line detection algorithm, is obtained by counting line segment direction histogram
Take building principal direction and rotate building to horizontal direction.
4. the method that interactive semi-automatic high-resolution remote sensing image building according to claim 1 is extracted, its feature
It is：In step 3, first with simple linear Iterative Clustering (SLIC) to raw video pre-segmentation, then using being based on
The GrabCut dividing methods of super-pixel obtain building figure spot.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710305968.9A CN107194405B (en) | 2017-05-03 | 2017-05-03 | Interactive semi-automatic high-resolution remote sensing image building extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710305968.9A CN107194405B (en) | 2017-05-03 | 2017-05-03 | Interactive semi-automatic high-resolution remote sensing image building extraction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107194405A true CN107194405A (en) | 2017-09-22 |
| CN107194405B CN107194405B (en) | 2020-02-14 |
Family
ID=59873343
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710305968.9A Active CN107194405B (en) | 2017-05-03 | 2017-05-03 | Interactive semi-automatic high-resolution remote sensing image building extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107194405B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108022268A (en) * | 2017-11-21 | 2018-05-11 | 长沙理工大学 | Color of image edit methods and device |
| CN111178175A (en) * | 2019-12-12 | 2020-05-19 | 中国资源卫星应用中心 | Automatic building information extraction method and system based on high-view satellite image |
| CN111325684A (en) * | 2020-02-01 | 2020-06-23 | 武汉大学 | Semi-automatic high-spatial-resolution remote sensing image extraction method for buildings with different shapes |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101604450A (en) * | 2009-07-24 | 2009-12-16 | 武汉大学 | The method of integrated images and LiDAR data extract contour of building |
| CN104156988A (en) * | 2014-08-05 | 2014-11-19 | 陕西汇图测绘地理信息有限公司 | Urban building contour regularization method based on iteration minimum bounding rectangle |
| CN106127791A (en) * | 2016-07-06 | 2016-11-16 | 武汉大学 | A kind of contour of building line drawing method of aviation remote sensing image |
-
2017
- 2017-05-03 CN CN201710305968.9A patent/CN107194405B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101604450A (en) * | 2009-07-24 | 2009-12-16 | 武汉大学 | The method of integrated images and LiDAR data extract contour of building |
| CN104156988A (en) * | 2014-08-05 | 2014-11-19 | 陕西汇图测绘地理信息有限公司 | Urban building contour regularization method based on iteration minimum bounding rectangle |
| CN106127791A (en) * | 2016-07-06 | 2016-11-16 | 武汉大学 | A kind of contour of building line drawing method of aviation remote sensing image |
Non-Patent Citations (1)
| Title |
|---|
| YIHUA TAN ET AL.: "Semi-Automatic Building Extraction From Very High Resolution Remote Sensing Imagery Via Energy Minimization Model", 《IEEE》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108022268A (en) * | 2017-11-21 | 2018-05-11 | 长沙理工大学 | Color of image edit methods and device |
| CN108022268B (en) * | 2017-11-21 | 2020-03-17 | 长沙理工大学 | Image color editing method and device |
| CN111178175A (en) * | 2019-12-12 | 2020-05-19 | 中国资源卫星应用中心 | Automatic building information extraction method and system based on high-view satellite image |
| CN111325684A (en) * | 2020-02-01 | 2020-06-23 | 武汉大学 | Semi-automatic high-spatial-resolution remote sensing image extraction method for buildings with different shapes |
| CN111325684B (en) * | 2020-02-01 | 2022-04-26 | 武汉大学 | Semi-automatic high-spatial-resolution remote sensing image extraction method for buildings with different shapes |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107194405B (en) | 2020-02-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105761266B (en) | The method that Rectangle building is extracted from remote sensing images | |
| CN107705309B (en) | Forest parameter evaluation method in laser point cloud | |
| Li et al. | An edge embedded marker-based watershed algorithm for high spatial resolution remote sensing image segmentation | |
| CN104915636B (en) | Remote sensing image road recognition methods based on multistage frame significant characteristics | |
| Yang et al. | Semiautomated building facade footprint extraction from mobile LiDAR point clouds | |
| CN105335966B (en) | Multiscale morphology image division method based on local homogeney index | |
| CN107229917A (en) | A kind of several remote sensing image general character well-marked target detection methods clustered based on iteration | |
| CN106127791A (en) | A kind of contour of building line drawing method of aviation remote sensing image | |
| Jamil et al. | Tree species extraction and land use/cover classification from high-resolution digital orthophoto maps | |
| Harikumar et al. | A local projection-based approach to individual tree detection and 3-D crown delineation in multistoried coniferous forests using high-density airborne LiDAR data | |
| CN110309781A (en) | Damage remote sensing recognition method in house based on the fusion of multi-scale spectrum texture self-adaption | |
| CN103578110A (en) | Multi-band high-resolution remote sensing image segmentation method based on gray scale co-occurrence matrix | |
| Yang et al. | A multi-band watershed segmentation method for individual tree crown delineation from high resolution multispectral aerial image | |
| CN103077515A (en) | Multi-spectral image building change detection method | |
| CN106446925A (en) | Dolphin identity recognition method based on image processing | |
| CN108629297A (en) | A kind of remote sensing images cloud detection method of optic based on spatial domain natural scene statistics | |
| CN107194405A (en) | A kind of method that interactive semi-automatic high-resolution remote sensing image building is extracted | |
| CN107992856A (en) | High score remote sensing building effects detection method under City scenarios | |
| Zheng et al. | Building recognition of UAV remote sensing images by deep learning | |
| CN108710862A (en) | A kind of high-resolution remote sensing image Clean water withdraw method | |
| CN106875407A (en) | A kind of unmanned plane image crown canopy dividing method of combining form and marking of control | |
| CN108022245B (en) | Facial line primitive association model-based photovoltaic panel template automatic generation method | |
| CN110070545A (en) | A kind of method that textural characteristics density in cities and towns automatically extracts cities and towns built-up areas | |
| Dong et al. | Cloud detection method for high resolution remote sensing imagery based on the spectrum and texture of superpixels | |
| CN104036295A (en) | Road center line auto-detection method employing multispectral remote sensing images |
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 | ||
| CB02 | Change of applicant information |
Address after: 430040 1 new bridge four, Jin Yin Hu Street, Dongxihu District, Wuhan, Hubei Applicant after: POWERCHINA HUBEI ELECTRIC ENGINEERING Corp.,Ltd. Applicant after: WUHAN University Applicant after: XI'AN University OF SCIENCE AND TECHNOLOGY Address before: 430040 1 new bridge four, Jin Yin Hu Street, Dongxihu District, Wuhan, Hubei Applicant before: HUBEI ELECTRIC POWER SURVEY & DESIGN INSTITUTE Applicant before: Wuhan University Applicant before: Xi'an University of Science and Technology |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 430040 1 new bridge four, Jin Yin Hu Street, Dongxihu District, Wuhan, Hubei Patentee after: Hubei Electric Power Planning, Design and Research Institute Co.,Ltd. Patentee after: WUHAN University Patentee after: XI'AN University OF SCIENCE AND TECHNOLOGY Address before: 430040 1 new bridge four, Jin Yin Hu Street, Dongxihu District, Wuhan, Hubei Patentee before: POWERCHINA HUBEI ELECTRIC ENGINEERING Corp.,Ltd. Patentee before: WUHAN University Patentee before: XI'AN University OF SCIENCE AND TECHNOLOGY |
|
| CP01 | Change in the name or title of a patent holder |