CN106340005A - High-resolution remote sensing image unsupervised segmentation method based on scale parameter automatic optimization - Google Patents
High-resolution remote sensing image unsupervised segmentation method based on scale parameter automatic optimization Download PDFInfo
- Publication number
- CN106340005A CN106340005A CN201610664398.8A CN201610664398A CN106340005A CN 106340005 A CN106340005 A CN 106340005A CN 201610664398 A CN201610664398 A CN 201610664398A CN 106340005 A CN106340005 A CN 106340005A
- Authority
- CN
- China
- Prior art keywords
- image
- value
- segmentation
- pixel
- yardstick
- 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 123
- 230000011218 segmentation Effects 0.000 title claims abstract description 80
- 238000005457 optimization Methods 0.000 title abstract 2
- 238000000605 extraction Methods 0.000 claims abstract description 11
- 230000008569 process Effects 0.000 claims abstract description 11
- 238000003709 image segmentation Methods 0.000 claims abstract description 8
- 230000006978 adaptation Effects 0.000 claims description 9
- 238000011002 quantification Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 4
- 230000010339 dilation Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 3
- AIMMVWOEOZMVMS-UHFFFAOYSA-N cyclopropanecarboxamide Chemical compound NC(=O)C1CC1 AIMMVWOEOZMVMS-UHFFFAOYSA-N 0.000 claims 1
- 238000002474 experimental method Methods 0.000 abstract description 15
- 230000004807 localization Effects 0.000 abstract 1
- 238000004458 analytical method Methods 0.000 description 14
- 238000001228 spectrum Methods 0.000 description 13
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical group [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 9
- 230000003595 spectral effect Effects 0.000 description 8
- 238000011156 evaluation Methods 0.000 description 6
- 241000894007 species Species 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 239000000284 extract Substances 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- IXSZQYVWNJNRAL-UHFFFAOYSA-N etoxazole Chemical compound CCOC1=CC(C(C)(C)C)=CC=C1C1N=C(C=2C(=CC=CC=2F)F)OC1 IXSZQYVWNJNRAL-UHFFFAOYSA-N 0.000 description 2
- 238000010191 image analysis Methods 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 241000908464 Geobia Species 0.000 description 1
- 208000031481 Pathologic Constriction Diseases 0.000 description 1
- 244000173948 Sanguisorba canadensis Species 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000007596 consolidation process Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 210000001215 vagina Anatomy 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a high-resolution remote sensing image unsupervised segmentation method based on scale parameter automatic optimization. The high-resolution remote sensing image unsupervised segmentation method mainly comprises three steps of 1) J-value self-adaptive SP selection based on a local homogeneity index; 2) image segmentation based on an inter-scale object boundary constraint strategy; and 3) region merging based on multiple features. The method is more accurate in object edge localization and more completed in object contour extraction through the experiment of multiple sets of high-resolution remote sensing images of different sensor types and comparison with the well-known commercial software eCognition and the conventional supervised segmentation method, and the segmentation process has no artificial intervention so that the high-resolution remote sensing image unsupervised segmentation method is a high-generality and effective unsupervised solution.
Description
Technical field
The present invention relates to a kind of non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal, belong to distant
Sense image division technology field.
Background technology
In recent years, object-based graphical analyses obia (object-based image analysis) is in giscience
(geographic information science) and remote sensing fields (especially high-resolution remote sensing image application)
In be just increasingly taken seriously.And image segmentation is one of core procedure in obia, it achieves geographic object in scene
Profile information extracts, and is basis and the premise subsequently carrying out feature extraction and target recognition.Compared with normal image, remote sensing image
Possess cover a wide range, many characteristics such as multiband, many spatial resolution, the ground species comprising are also abundanter, therefore
Traditional dividing method is difficult to directly apply on remote sensing image.Meanwhile, with remote sensing satellite spatial resolution not
Disconnected raising, such as spot 5, ikonos, quickbird etc. are for representing meter level, the high-resolution data of sub-meter grade is widely used to
The every field of the social lifes such as crop yield investigation, city land plan, disaster monitoring and early warning, is therefore directed to high-resolution
The image Segmentation Technology of rate remote sensing image has become as one of study hotspot of remote sensing fields.
With in, compared with low resolution remote sensing image, high-resolution remote sensing image brings more abundant spectrum and texture
Feature, the spatial detail information such as the structure of object, shape can clearly be expressed, the mixed pixel problem in the past projecting
Obtain effectively solving, thus improve the inter-class separability of adjacent atural object.On the other hand, figure is also given in the raising of spatial resolution
As segmentation causes new difficulty and challenge: " the different spectrum of jljl " phenomenon is more prominent, i.e. spectral signature between the atural object of identical type
It is likely to be of significant difference；While detailed information increases, the interference factor such as atural object shade, noise, cloud layer covering impacts
Also more significantly；In City scenarios, changeable ecological environment, abundant ground species and baroque artificial atural object etc. is all
Cause difficulty to the accurate geographical objects that extracts.
For this reason, scholars have been proposed that some solve countermeasure, one of most important of which means are introduced into multiple dimensioned dividing
Cut strategy, thus preferably disclosing spatial structure characteristic under different scale for the object.For example, c.burnet etc. proposes one kind
Based on the multi-scale division algorithm of point shape, with heterogeneity and it is iterated excellent by the homogeneity estimating regional area spectral signature
Change, achieve good effect[1]；Well-known remote sensing business software ecognition employs fractal net work evolution algorithmic
(fractal net evolution algorithm, fnea) carries out multi-scale division, takes full advantage of spectrum, the stricture of vagina of object
Information between reason, shape, level and class.It is pointed out that multi-scale segmentation method is required for by artificial in prior art
Determining scale parameter (scale parameter, sp), these methods all can not be referred to as automatization for interpretation or trial-and-error method
Image segmentation.And the multi-scale division solution of sp parameter adaptive is also rare at present, this also becomes restriction obia skill
One of wide variety of Main Bottleneck of art.
List of references
[1]burnett c,blaschke t.a multi-scale segmentation/object
relationship modelling methodology for landscape analysis[j].ecological
modelling,2003,168(3):233-249.
[2] shao p, yang g, niu x, et al.information extraction of high-
resolution remotely sensed image based on multiresolution segmentation[j]
.sustainability, 2014,6 (8): 5300-5310.
[3]deng y,manjunath b s.unsupervised segmentation of color-texture
regions in images and video[j].pattern analysis and machine intelligence,ieee
transactions on,2001,23(8):800-810.
[4]baraldi a,boschetti l.operational automatic remote sensing image
understanding systems:beyond geographic object-based and object-oriented
image analysis(geobia/geooia).part 2:novel system architecture,information/
knowledge representation,algorithm design and implementation[j].remote
sensing,2012,4(9):2768-2817.
Content of the invention
Goal of the invention: for problems of the prior art and deficiency, the present invention passes through conventional color Texture Segmentation
Multiple dimensioned j-image sequence in method jseg is introduced high-resolution remote sensing image and splits it is proposed that being referred to based on local homogeneity
Object bounds constraints policy and the region merging technique based on multiple features between the self adaptation sp selection strategy of target j-value, yardstick
Strategy is it is achieved that the multi-scale division of automatization.Carried out by the remote sensing image that different sensors type, different spaces are differentiated
Experiment, and be compared with dividing method (the ecognition and document 2) experimental result of two kinds of supervision it was demonstrated that being proposed calculation
It is more accurate that method not only positions target edges, and extracting object profile is more complete and without manual intervention, improves cutting procedure
Automaticity and robustness.
Technical scheme: a kind of non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal, main bag
Include three steps: 1) self adaptation sp of the j-value based on local homogeneity index selects, so that it is determined that optimal multiple dimensioned j-
Image sequence；2) between based on yardstick, the image segmentation of object bounds constraints policy is it is achieved that arrive smart multi-scale division by thick；
3) region merging technique based on multiple features, to tackle over-segmentation phenomenon that may be present in segmentation result.
Sp is adaptively selected
Select j-image sequence as multiscale analysis platform, and propose a kind of adaptively selected strategy of sp.
The calculating process of multiple dimensioned j-image is as follows: carries out color quantizing to raw video in luv space first.In amount
Change the window z setting a size of m × m centered on pixel z (m is sp) pixel in image, and by each pixel in window
Coordinate z (x, y) as its pixel value, and z (x, y) ∈ z.Angle point in window is removed simultaneously.
If gray level sum is p in quantification image, make zpFor belonging to the set of all pixels of gray level p, m in window zp
For all pixel corresponding pixel averages belonging to gray level p, then belong to the variance of same gray-level pixels and can in window z
It is expressed as:
In window z, the population variance of all pixels is represented by:
Wherein, m is the average of all pixels in window z.Then homogeneity index j-value in local may be defined as:
J-value=(st-sw)/sw(3)
Now, using corresponding for pixel z j-value as the pixel value of this pixel, travel through view picture quantification image, sp can be obtained
For j-image during m, multiple dimensioned j-image sequence can be obtained by changing sp.
Self adaptation sp selection strategy based on j-value:
Step1: calculating sp is j-image sequence during m (m=5,6....n), and wherein m=5 is that j-image allows
Small window size, n represents the j-image of coarse scale.
Step2: calculate the pixel j-value average under all yardstick j-imageAnd buildCurve.
Step3: numerousIn point of inflexion on a curve, only select some flex points the most prominent, these flex points should meet
The multi-scale division of constraint
In the segmentation stage, the segmentation strategy of object bounds constraint between proposing based on yardstick.If optimal j-image sequence comprises
L yardstick, is represented by sk(k=1,2...l), implements process as follows:
Step1: first to coarse scale s1Split.According to formula (4) threshold value t1Carry out seed region to carry
Take, wherein μkAnd σkRepresent yardstick s respectivelykThe j-value average of middle all pixels and variance.
tk=μk-0.2σk, (k=1,2...l) (4)
In s1In, all j-value values are less than t1Pixel adopt four methods of attachment to constitute UNICOM regions, as one by one
Seed region.With seed region as starting point, according to four direction up and down, with j-value value, order from small to large carries out area
Domain increases, and border when adjacent area crosses just constitutes s1Under segmentation result.
Step2: the object bounds of a upper yardstick are mapped to current scale and are modified.By current scale j-image
It is converted into a width bianry image, only retain the object bounds by mapping extraction between yardstick, and carry out morphological dilation.Swollen
Swollen construction unit is dimensioned so as to m × m pixel, and m is the sp of current scale.Using the border after expanding by current scale j-
Image is divided into seed region independent one by one, and carries out region to these seed regions from small to large according to j-value value
Increase, border when adjacent area crosses is the result that border is revised.
Now, the segmentation under current scale is only carried out inside by the object revising back boundary extraction.And in order to avoid mistake
For the higher object of internal homogenizing degree, segmentation phenomenon, thinks that it is mated with actual type of ground objects, under current scale
No longer split.Judgment rule is that the j-value average within this object should be less than corresponding threshold value t of current scalek(referring to
Formula 4).On this basis, according to threshold value tkSplit to remaining object, cutting procedure is identical with yardstick 1, and will split
Result is mapped to next yardstick.
Step3: repeat the cutting procedure of step2, until yardstick l segmentation finishes, thus obtaining preliminary segmentation result.
The region merging technique of multiple features
Raw video each wave band correspondence multiple dimensioned j-image sequence is calculated according to optimal sp.If raw video comprises f
Individual wave band, for any one object q, defined feature vector jqf=(jq1,jq2..., jqf), wherein each component represent right
As j-value average under l yardstick j-image of each wave band for the q, then for any wave band f, (f=1,2 ... f), hasAccording to the definition of j-value, j-value concentrated expression regional area (object)
Spectrum, texture and dimensional information, therefore pass through to judge adjacent object qaAnd qbCharacteristic vector between Euclidean distance judging its phase
Like degree, such as shown in formula (5).
Region merging technique is carried out using rag (region adjacency graphics), detailed process is as follows:
Step1: according to the segmentation result in yardstick l, generate the rag of all adjacent object.
Step2: select and any object qaAdjacent all objects, calculate Euclidean distance according to formula (5).
Step3: if there is object qbMeet d (qa,qb)≤0.1 is then it is assumed that qaAnd qbBelong to same target, merge qaAnd qb
And generate new rag.Otherwise, return step2.
Step4: repeat step2 to step3, travel through all objects, obtain final segmentation result.
Brief description
Fig. 1 the inventive method flow chart；
Fig. 2 is specific dimensions window z during m=9；
Fig. 3 is qucikbird image in 2005；
Fig. 4 isCurve and optimal sp select；
Fig. 5 is the inventive method segmentation result；
Fig. 6 is method two segmentation result；
Fig. 7 is method three segmentation result；
Fig. 8 is qucikbird image in 2005；
Fig. 9 isCurve and optimal sp select；
Figure 10 is the inventive method segmentation result；
Figure 11 is method two segmentation result；
Figure 12 is method three method segmentation result；
Figure 13 is experiment onePrecision evaluation；
Figure 14 is experiment twoPrecision evaluation.
Specific embodiment
With reference to specific embodiment, it is further elucidated with the present invention it should be understood that these embodiments are merely to illustrate the present invention
Rather than restriction the scope of the present invention, after having read the present invention, the various equivalences to the present invention for the those skilled in the art
The modification of form all falls within the application claims limited range.
As shown in figure 1, the non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal, mainly include three
Individual step: 1) self adaptation sp of the j-value based on local homogeneity index selects, so that it is determined that optimal multiple dimensioned j-
Image sequence；2) between based on yardstick, the image segmentation of object bounds constraints policy is it is achieved that arrive smart multi-scale division by thick；
3) region merging technique based on multiple features, to tackle over-segmentation phenomenon that may be present in segmentation result.
Sp is adaptively selected
It is that piece image is divided into independent ground by sp that multi-scale division generally relies primarily on a core control parameter
Reason object.Sp is controlled by the average-size of the spectrum homogeneity within object in segmentation result and object, also actual shadow
Ring the quantity of object in segmentation result.Therefore, select how suitable multiscale analysis instrument and sp determine it is that the present invention is first
First need one of key issue of solution.
Multiple dimensioned j-image sequence
The multiple dimensioned j-image sequence that jseg adopts can fully reflect the homogeneity of regional area spectral distribution, also may be used
Avoid using traditional multiscale transform analytical tool (as wavelet transformation, profile wave convert) calculate image fusion when exist only right
The limitation of the high-frequency information sensitivity in indivedual directions, but similarly it is faced with the rational selection of sp.Therefore, this method selects j-
Image sequence is as multiscale analysis platform, and proposes a kind of adaptively selected strategy of sp.
The calculating process of multiple dimensioned j-image is as follows: carries out color quantizing to raw video in luv space first.In amount
Change the window z setting a size of m × m centered on pixel z (m is sp) pixel in image, and by each pixel in window
Coordinate z (x, y) as its pixel value, and z (x, y) ∈ z.Meanwhile, the angle in order to ensure the concordance of all directions, in window
Point is removed., the window z centered on pixel z is as shown in Figure 2 taking m=9 as a example:
If gray level sum is p in quantification image, make zpFor belonging to the set of all pixels of gray level p, m in window zp
For all pixel corresponding pixel averages belonging to gray level p, then belong to the variance of same gray-level pixels and can in window z
It is expressed as:
In window z, the population variance of all pixels is represented by:
Wherein, m is the average of all pixels in window z.Then homogeneity index j-value in local may be defined as:
J-value=(st-sw)/sw(3)
Now, using corresponding for pixel z j-value as the pixel value of this pixel, travel through view picture quantification image, sp can be obtained
For j-image during m, multiple dimensioned j-image sequence can be obtained by changing sp.
Best scale based on j-value selects
Using j-image sequence as multiscale analysis instrument it is proposed that a kind of self adaptation sp based on j-value selects
Strategy.
Step1: calculating sp is j-image sequence during m (m=5,6....n), and wherein m=5 is that j-image allows
Small window size, n (suitably can be adjusted according to real image size, set n=30 herein) represents coarse scale
J-image.
Step2: calculate the pixel j-value average under all yardstick j-imageAnd buildCurve.
According to the definition of j-value,Point of inflexion on a curve reflects compared with front and back yardstick, the homogenizing journey of spectral distribution under current scale
Spend to have and suddenly increase.So we assume that some representative ground species are suitable in these flex points explanation scene
Split under current scale.Namely at these flex points, the object in segmentation result is mated with actual type of ground objects just,
There is same or like spectrum homogenizing degree, and these representative atural objects can be rightCurve makes a significant impact.
Step3: in numerous flex points, only select some flex points the most prominent, these flex points should meetTo make most representational atural object species in scene be effectively extracted as far as possible.In addition,
In order to retain the detailed information of image, yardstick (i.e. m=5) the finest is chosen all the time.At this point it is possible to comprehensive selected this
A little corresponding sp of flex point and fine dimension to determine most preferably multiple dimensioned j-image sequence jointly.
The multi-scale division of constraint
In the segmentation stage, the segmentation strategy of object bounds constraint between proposing based on yardstick.If optimal j-image sequence comprises
L yardstick, is represented by sk(k=1,2...l), implements process as follows:
Step1: first to coarse scale s1Split.According to formula (4) threshold value t1Carry out seed region to carry
Take, wherein μkAnd σkRepresent yardstick s respectivelykThe j-value average of middle all pixels and variance.
tk=μk-0.2σk, (k=1,2...l) (4)
In s1In, all j-value values are less than t1Pixel adopt four methods of attachment to constitute UNICOM regions, as one by one
Seed region.With seed region as starting point, according to four direction up and down, with j-value value, order from small to large carries out area
Domain increases, and border when adjacent area crosses just constitutes s1Under segmentation result.
Step2: the object bounds of a upper yardstick are mapped to current scale and are modified.Definition according to j-image
Understand, the multiple dimensioned sequence of j-image is all of the same size with raw video.Therefore go up a coarse scale segmentation result to be carried
The same position that the object bounds taking can be mapped to current fine dimension according to coordinate up, and under current scale point
Undercutting row constraint.Although and the border extracted under coarse scale can determine position and its general profile of object, being difficult to standard
Determine the edge of position object it is therefore desirable to be modified under current scale, process is as follows.
Current scale j-image is converted into a width bianry image, only retains the object edges by mapping extraction between yardstick
Boundary, and carry out morphological dilation.Expansion structure unit size is set as m × m pixel, and m is the sp of current scale.Using swollen
Current scale j-image is divided into seed region independent one by one by the border after swollen, and to these seed regions according to j-
Value value carries out region growth from small to large, and border when adjacent area crosses is the result that border is revised.
Now, the segmentation under current scale is only carried out inside by the object revising back boundary extraction.And in order to avoid mistake
For the higher object of internal homogenizing degree, segmentation phenomenon, thinks that it is mated with actual type of ground objects, under current scale
No longer split.Judgment rule is that the j-value average within this object should be less than corresponding threshold value t of current scalek(referring to
Formula 4).On this basis, according to threshold value tkSplit to remaining object, cutting procedure is identical with yardstick 1, and will split
Result is mapped to next yardstick.
Step3: repeat the cutting procedure of step2, until yardstick l segmentation finishes, thus obtaining preliminary segmentation result.
The region merging technique of multiple features
Although higher to internal homogenizing degree first object is differentiated, too before to each multi-scale segmentation
Cut phenomenon to be still difficult to avoid that in addition it is also necessary to further region merging technique is processed.Due to prominent " jljl in high-resolution remote sensing image
Different spectrum " and " same object different images " phenomenon, may produce consolidation problem by mistake merely with the spectral signature within object, it is therefore proposed that one
Plant the region merging technique strategy of multiple features.
Raw video each wave band correspondence multiple dimensioned j-image sequence is calculated according to optimal sp.If raw video comprises f
Individual wave band, for any one object q, defined feature vector jqf=(jq1,jq2..., jqf), wherein each component represent right
As j-value average under l yardstick j-image of each wave band for the q, then for any wave band f, (f=1,2 ... f), hasAccording to the definition of j-value, the j-value concentrated expression light of regional area (object)
Spectrum, texture and dimensional information, therefore pass through to judge adjacent object qaAnd qbCharacteristic vector between Euclidean distance similar to judge it
Degree, such as shown in formula (5).
We adopt rag (region adjacency graphics) to carry out region merging technique, and detailed process is as follows:
Step1: according to the segmentation result in yardstick l, generate the rag of all adjacent object.
Step2: select and any object qaAdjacent all objects, calculate Euclidean distance according to formula (5).
Step3: if there is object qbMeet d (qa,qb)≤0.1 is then it is assumed that qaAnd qbBelong to same target, merge qaAnd qb
And generate new rag.Otherwise, return step2.
Step4: repeat step2 to step3, travel through all objects, obtain final segmentation result.
Experiment and analysis
For verify proposed method validity and reliability, to two groups of different resolutions, different sensors type many
Spectrum high resolution remote sensing image is tested, and with business software econgnition (referred to as " method two ") and shao
The high-resolution remote sensing image dividing method (referred to as " method three ") of the supervision that p et al. et al. (document 2) proposes is compared
Relatively.
Wherein, econgnition is a internationally recognizable OO of German definiens imaging exploitation
Classification of remote-sensing images software, the fnea segmentation strategy that it adopts fully utilizes the features such as spectrum, texture and the shape of object,
In the segmentation of high-resolution remote sensing image, there is excellent performance.Econgnition is mainly subject to three parameters in segmentation
Control it may be assumed that scale parameter (scale parameter), the object average-size in major control segmentation result；Form parameter
(shape parameter), contributes to the integrity of keeping object profile in cutting procedure；Degree of compacting parameter (compactness
Parameter), it is favorably improved the inter-class separability of object.The method that shao p et al. et al. proposes is by traditional edge
Detection is introduced in high-resolution remote sensing image segmentation, achieves the Multi resolution feature extraction of object by structure object hierarchy and divides
Cut, in the segmentation of Chinese zy-3 satellite image, achieve good effect.In both the above method, the setting of relevant parameter is both needed to
Human interpretation to be passed through realizes, and the present invention determines that optimal parameter combines in an experiment by trial-and-error method.
Test a result and analysis
Experiment one adopts the qucikbird tetra- wave band color integration data of 2005, and location is Wuhan, China, space
Resolution is 2.4m, and picture size is 512 × 512 pixels.Image is mainly the typical urban scene areas under complex background, bag
The ground species enriched containing road, playground, water body and baroque man-made target etc., as shown in Figure 3.
In the inventive methodCurve is as shown in figure 4, describe constantly to increase with scale parameter m, indexChange
Situation.Wherein vertical dotted line has corresponded to optimal sp, and these yardsticks and fine dimension together form most preferably multiple dimensioned j-image
Sequence, corresponding scale parameter is m ∈ [5,13,18,28], and final segmentation result is as shown in Figure 5.
In method two, setting scale parameter as 77, shape parameter is 50, compactness
Parameter is 40.Each wave band proportion is set identical, parameter " scale parameter " is 30, parameter in document 2
" shape heterogeneous degree " is 0.4, parameter " compactness parameter " and " smoothness
Parameter " is 0.5.Two methods experimental result is respectively as shown in Figure 6, Figure 7.
For the ease of being compared to distinct methods experimental result, the present invention is carried out to the typical subject in scene or position
Mark.Wherein, position a, b is sports ground, and position c, d, f are building, position e is road, and position g is man-made lake.Pass through
Visual analysis can be seen that the playground areas that three kinds of methods are all effectively extracted position a, wherein this method and method two and positions
The obvious ratio method in edge three on lawn is more accurate, and method three has certain over-segmentation phenomenon；Playground area for position b
Domain, method two does not extract the profile on lawn, and part runway zone and lawn are then obscured by method three；The building of position c
Complex structure, only this method not only maintain the integrity of object outline, have been accurately positioned target edges simultaneously, method two,
Three are respectively present segmentation and over-segmentation phenomenon by mistake；The building shape rule of position d, f, this method and method two positioning roof
Edge more accurate, but there is less divided phenomenon in position f method two；Only method three is effectively extracted the artificial of position g
The profile information in lake.For synthesis, the edge detail information ability side of being substantially better than of the inventive method and method two positioning object
Method three, and the inventive method is more complete for the profile holding of bulk homogenous area.Method three can effectively identify and belong to not
With species but the similar adjacent atural object of spectral signature, but there is also that prominent positioning precision is low and over-segmentation problem.
Test two results and analysis
Experiment two selection three wave band high-resolution air remote sensings dom (digital orthophoto map) image, data
Acquisition time is in March, 2009, and spatial resolution is 0.6m, a size of 512 × 512 pixels, and location is Nanjing of China, such as schemes
Shown in 8.By decreasing as can be seen that testing the two data background complexities adopting with Fig. 3 contrast, but variety classes ground
The minutias such as the spectrum of thing, texture, edge are more notable, and object includes the large area homogenous area of regular shape and structure is answered
The man-made structures that miscellaneous, textural characteristics enrich, are therefore accurately positioned the complete of target edges and keeping object profile to partitioning algorithm
Property is proposed higher requirement.
In this methodCurve is as shown in Figure 9.According toCurve understands, the corresponding scale parameter of most preferably multiple dimensioned sequence is
M ∈ [5,10,12,21,25,27], segmentation result is as shown in Figure 10.
In method two, setting scale parameter as 100, shape parameter is 50, compactness
Parameter is 50.Each wave band proportion is set identical, scale parameter is 50, shape in document 14
Heterogeneous degree is 0.5 for 0.3, compactness parameter and smoothness parameter.
Two methods experimental result is respectively as shown in Figure 11, Figure 12.
One identical with experiment, the present invention in scene typical subject or position marked.By to not Tongfang
The visual analysis of method experimental result can be seen that three kinds of methods and has all accurately been partitioned into the playground lawn of position a and the behaviour of position b
Runway, but only the inventive method effectively maintains the integrity of playground profile, method two two larger cut zone it
Between there are some long and narrow false units, such as adjacent with lawn outside runway region；For the fritter lawn of position c, method
Two have less divided phenomenon；For the building roof of position d, three kinds of method segmentation effects are close, and method two, three carries further
Take the texture information within roof, but method three there is over-segmentation phenomenon and edge positioning is inaccurate；The playground of position e, f is seen
Platform complex structure, texture information enriches, and only the inventive method maintains the complete of the ceiling area of grandstand and is effectively extracted
The minutia of both sides auxiliary building；The inventive method and method two are accurately extracted the road area positioned at position g, h, method
Three have segmentation problem by mistake；Large area vegetation region for the court of position i, the tennis court of position j and position k
Domain, three kinds of methods are all more accurate to the edge positioning of object, but occur in that long and narrow in method two in the segmentation in tennis court again
False unit problem.Comprehensive above analysis can draw the conclusion similar with testing, demonstrates proposed algorithm further
Reliability.
Precision evaluation
Mainly analyze the segmentation effect to distinct methods by visual observation above to be evaluated, herein will be using precision index
Experimental result is done with further quantitative analyses.In the method that deng et al. proposes (document 3), index j-value is not only
For calculating multiple dimensioned j-image sequence, also as the evaluation index of segmentation precision, it is defined as follows:
Wherein, u is the sum of all pixels in image, and r is the region sum in segmentation result, wrAnd jrIt is respectively r-th region
Internal sum of all pixels and its corresponding j-value.When the corresponding precision index of segmentation resultMore hour, illustrates segmentation knot
In fruit, the averaged spectrum homogeneity within object is higher, then segmentation effect is better.Due to atural object huge number in remote sensing image,
In different application scenarios, the evaluation angle of segmentation precision is also not quite similar with standard, and what deng et al. et al. proposedOn the whole spectral distribution situation in segmentation result is evaluated, there is good versatility[4].Therefore, the present invention adopts
WithPrecision evaluation is carried out to the experimental result of three kinds of methods.
To j in segmentation resultrDistribution situation is analyzed, by corresponding for all objects j-value value in [0,1] is interval
Uniform quantization is 20 units, then jrDistribution curve is as shown in Figure 13, Figure 14:
In figure is represented respectively with the solid line of different colours respectively and belongs to the interval object of different j-value in three kinds of methods
Shared proportion in segmentation result, and dotted line then to represent three kinds of methods correspondingIndex.By compare Figure 13,14 can
To find out: in two groups of experiments, the present invention proposes method precision indexIt is significantly better than that other two methods, with visual point
Solution result is consistent.J in two groups of experimentsrCurvilinear trend is roughly the same, and difference major embodiment typical feature in the scene is concentrated
[0.2,0.55] in interval, such as experiment one is interval and [0.1,0.4] that test in two is interval.In addition in experiment two, three
The segmentation precision of kind algorithm is relatively tested one and is all significantly increased, and main reason is that the remote sensing image spatial discriminations of experiment two employing
Rate is higher, and background is relatively easy and atural object minutia is more prominent, therefore tests the object of extraction and its border is more nearly
Type of ground objects in actual scene.
For the automatic segmentation of high-resolution remote sensing image, the present invention proposes non-supervisory multiple dimensioned point of a kind of novelty
Segmentation method.The method fully utilizes the spectrum of object and textural characteristics it is proposed that based on passing based on local homogeneity index
Self adaptation scale parameter (sp) selection strategy of j-value, enable typical feature type in scene with its
Split in best scale j-image joined.On this basis, the multi-scale division strategy being proposed makes region segmentation be subject to
A upper yardstick extracts the constraint of institute object bounds, and these borders are modified under current scale, it is to avoid between yardstick
The accumulation of error.And the region merging technique strategy based on multiple features then can effectively distinguish the variety classes ground with similar spectral feature
Thing, it is to avoid by mistake merge phenomenon.Experiment shows, compared with ecognition and traditional supervised segmentation method, proposed method is fixed
Accurately and extracting object profile is more complete for position target edges, has higher segmentation precision, is capable of automatization simultaneously
High-resolution remote sensing image is split, whole without manual intervention, be a kind of generic and effective non-supervisory solution.
Claims (4)
1. a kind of non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal is it is characterised in that main wrap
Include three steps: 1) self adaptation sp of the j-value based on local homogeneity index selects, so that it is determined that optimal multiple dimensioned j-
Image sequence；2) between based on yardstick, the image segmentation of object bounds constraints policy is it is achieved that arrive smart multi-scale division by thick；
3) region merging technique based on multiple features, to tackle over-segmentation phenomenon that may be present in segmentation result.
2. the non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal as claimed in claim 1, it is special
Levy and be, during self adaptation sp of the j-value based on local homogeneity index selects:
The calculating process of multiple dimensioned j-image is as follows: carries out color quantizing to raw video in luv space first；Quantifying shadow
In picture, set the window z of a size of m × m centered on pixel z (m is sp) pixel, and the seat by each pixel in window
Mark z (x, y) is as its pixel value, and z (x, y) ∈ z；Angle point in window is removed simultaneously；
If gray level sum is p in quantification image, make zpFor belonging to the set of all pixels of gray level p, m in window zpFor institute
There is the pixel belonging to gray level p corresponding pixel average, then belong to the variance of same gray-level pixels in window z and can represent
For:
In window z, the population variance of all pixels is represented by:
Then homogeneity index j-value in local may be defined as:
J-value=(st-sw)/sw(3)
Now, using corresponding for pixel z j-value as the pixel value of this pixel, travel through view picture quantification image, can obtain sp is m
When j-image, by change sp can obtain multiple dimensioned j-image sequence；
The step of the self adaptation sp selection strategy based on j-value is:
Step1: calculating sp is j-image sequence during m (m=5,6....n), and wherein m=5 is the min window that j-image allows
Mouth size, n represents the j-image of coarse scale；
Step2: calculate the pixel j-value average under all yardstick j-imageAnd buildCurve.
Step3: numerousIn point of inflexion on a curve, only select some flex points the most prominent, these flex points should meet
3. the non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal as claimed in claim 2, it is special
Levy and be, in the segmentation stage, the segmentation strategy of object bounds constraint between proposing based on yardstick；If optimal j-image sequence comprises l
Individual yardstick, is represented by sk(k=1,2...l), implements process as follows:
Step1: first to coarse scale s1Split.According to formula (4) threshold value t1Carry out seed region extraction, its
Middle μkAnd σkRepresent yardstick s respectivelykThe j-value average of middle all pixels and variance；
tk=μk-0.2σk, (k=1,2...l) (4)
In s1In, all j-value values are less than t1Pixel adopt four methods of attachment to constitute UNICOM regions, as seed one by one
Region；With seed region as starting point, according to four direction up and down, with j-value value, order from small to large carries out region increasing
Long, border when adjacent area crosses just constitutes s1Under segmentation result；
Step2: the object bounds of a upper yardstick are mapped to current scale and are modified；Current scale j-image is converted
For a width bianry image, only retain the object bounds by mapping extraction between yardstick, and carry out morphological dilation.Expand knot
Structure unit size is set as m × m pixel, and m is the sp of current scale；Using the border after expanding, current scale j-image is drawn
It is divided into seed region independent one by one, and according to j-value value, region growth, phase are carried out from small to large to these seed regions
Border when neighbouring region crosses is the result that border is revised；
The higher object of internal homogenizing degree is thought that it is mated with actual type of ground objects, no longer enters under current scale
Row segmentation；Judgment rule is that the j-value average within this object should be less than corresponding threshold value t of current scalek；Here basis
On, according to threshold value tkRemaining object is split, cutting procedure is identical with yardstick 1, and segmentation result is mapped to next chi
Degree；
Step3: repeat the cutting procedure of step2, until yardstick l segmentation finishes, thus obtaining preliminary segmentation result.
4. the non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal as claimed in claim 3, it is special
Levy and be, raw video each wave band correspondence multiple dimensioned j-image sequence is calculated according to optimal sp；If raw video comprises f
Individual wave band, for any one object q, defined feature vector jqf=(jq1,jq2..., jqf), wherein each component represent right
As j-value average under l yardstick j-image of each wave band for the q, then for any wave band f, (f=1,2 ... f), hasBy judging adjacent object qaAnd qbCharacteristic vector between Euclidean distance judging its similar journey
Degree, such as shown in formula (5).
Region merging technique is carried out using rag, detailed process is as follows:
Step1: according to the segmentation result in yardstick l, generate the rag of all adjacent object；
Step2: select and any object qaAdjacent all objects, calculate Euclidean distance according to formula (5)；
Step3: if there is object qbMeet d (qa,qb)≤0.1 is then it is assumed that qaAnd qbBelong to same target, merge qaAnd qbAnd it is raw
The rag of Cheng Xin；Otherwise, return step2；
Step4: repeat step2 to step3, travel through all objects, obtain final segmentation result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610664398.8A CN106340005B (en) | 2016-08-12 | 2016-08-12 | The non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610664398.8A CN106340005B (en) | 2016-08-12 | 2016-08-12 | The non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106340005A true CN106340005A (en) | 2017-01-18 |
| CN106340005B CN106340005B (en) | 2019-09-17 |
Family
ID=57824778
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610664398.8A Active CN106340005B (en) | 2016-08-12 | 2016-08-12 | The non-supervisory dividing method of high score remote sensing image based on scale parameter Automatic Optimal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106340005B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107610118A (en) * | 2017-09-25 | 2018-01-19 | 中国科学院遥感与数字地球研究所 | One kind is based on dMImage segmentation quality evaluating method |
| CN107657616A (en) * | 2017-08-28 | 2018-02-02 | 南京信息工程大学 | A kind of high score Remote Sensing Image Segmentation towards geographic object |
| CN109859219A (en) * | 2019-02-26 | 2019-06-07 | 江西理工大学 | In conjunction with the high score Remote Sensing Image Segmentation of phase and spectrum |
| CN109918449A (en) * | 2019-03-16 | 2019-06-21 | 中国农业科学院农业资源与农业区划研究所 | A kind of agricultural disaster information remote sensing extracting method and system based on Internet of Things |
| CN109948415A (en) * | 2018-12-30 | 2019-06-28 | 中国科学院软件研究所 | Remote sensing image object detection method based on filtering background and scale prediction |
| CN109993753A (en) * | 2019-03-15 | 2019-07-09 | 北京大学 | The dividing method and device of urban function region in remote sensing image |
| CN116681711A (en) * | 2023-04-25 | 2023-09-01 | 中国科学院地理科学与资源研究所 | Multi-scale segmentation method for high-resolution remote sensing image under partition guidance |
| CN118366059A (en) * | 2024-06-20 | 2024-07-19 | 山东锋士信息技术有限公司 | Crop water demand calculating method based on optical and SAR data fusion |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632363A (en) * | 2013-08-27 | 2014-03-12 | 河海大学 | Object-level high-resolution remote sensing image change detection method based on multi-scale fusion |
| CN104361589A (en) * | 2014-11-12 | 2015-02-18 | 河海大学 | High-resolution remote sensing image segmentation method based on inter-scale mapping |
| CN105335966A (en) * | 2015-10-14 | 2016-02-17 | 南京信息工程大学 | Multi-scale remote-sensing image segmentation method based on local homogeneity index |
-
2016
- 2016-08-12 CN CN201610664398.8A patent/CN106340005B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632363A (en) * | 2013-08-27 | 2014-03-12 | 河海大学 | Object-level high-resolution remote sensing image change detection method based on multi-scale fusion |
| CN104361589A (en) * | 2014-11-12 | 2015-02-18 | 河海大学 | High-resolution remote sensing image segmentation method based on inter-scale mapping |
| CN105335966A (en) * | 2015-10-14 | 2016-02-17 | 南京信息工程大学 | Multi-scale remote-sensing image segmentation method based on local homogeneity index |
Non-Patent Citations (3)
| Title |
|---|
| YINING DENG等: "Color Image Segmentation", 《1999 IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| YINING DENG等: "Unsupervised Segmentation of Color-Texture Regions in Images and Video", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| 黄飞等: "基于 HIS空间的彩色图象分割", 《小型微型计算机系统》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107657616A (en) * | 2017-08-28 | 2018-02-02 | 南京信息工程大学 | A kind of high score Remote Sensing Image Segmentation towards geographic object |
| CN107657616B (en) * | 2017-08-28 | 2019-10-01 | 南京信息工程大学 | A kind of high score Remote Sensing Image Segmentation towards geographic object |
| CN107610118A (en) * | 2017-09-25 | 2018-01-19 | 中国科学院遥感与数字地球研究所 | One kind is based on dMImage segmentation quality evaluating method |
| CN107610118B (en) * | 2017-09-25 | 2020-12-08 | 中国科学院遥感与数字地球研究所 | Based on dMImage segmentation quality evaluation method |
| CN109948415A (en) * | 2018-12-30 | 2019-06-28 | 中国科学院软件研究所 | Remote sensing image object detection method based on filtering background and scale prediction |
| CN109859219A (en) * | 2019-02-26 | 2019-06-07 | 江西理工大学 | In conjunction with the high score Remote Sensing Image Segmentation of phase and spectrum |
| CN109993753A (en) * | 2019-03-15 | 2019-07-09 | 北京大学 | The dividing method and device of urban function region in remote sensing image |
| CN109993753B (en) * | 2019-03-15 | 2021-03-23 | 北京大学 | Method and device for segmenting urban functional area in remote sensing image |
| CN109918449A (en) * | 2019-03-16 | 2019-06-21 | 中国农业科学院农业资源与农业区划研究所 | A kind of agricultural disaster information remote sensing extracting method and system based on Internet of Things |
| CN116681711A (en) * | 2023-04-25 | 2023-09-01 | 中国科学院地理科学与资源研究所 | Multi-scale segmentation method for high-resolution remote sensing image under partition guidance |
| CN116681711B (en) * | 2023-04-25 | 2024-01-30 | 中国科学院地理科学与资源研究所 | Multi-scale segmentation method for high-resolution remote sensing image under partition guidance |
| CN118366059A (en) * | 2024-06-20 | 2024-07-19 | 山东锋士信息技术有限公司 | Crop water demand calculating method based on optical and SAR data fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106340005B (en) | 2019-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106340005A (en) | High-resolution remote sensing image unsupervised segmentation method based on scale parameter automatic optimization | |
| CN105335966B (en) | Multiscale morphology image division method based on local homogeney index | |
| Yang et al. | Urban surface water body detection with suppressed built-up noise based on water indices from Sentinel-2 MSI imagery | |
| CN103632363B (en) | Object level high-resolution remote sensing image change detecting method based on Multiscale Fusion | |
| Lu et al. | Land cover classification in a complex urban-rural landscape with QuickBird imagery | |
| CN103971115B (en) | Automatic extraction method for newly-increased construction land image spots based on NDVI and PanTex index | |
| Mallinis et al. | Object-based classification using Quickbird imagery for delineating forest vegetation polygons in a Mediterranean test site | |
| CN103578110B (en) | Multiband high-resolution remote sensing image dividing method based on gray level co-occurrence matrixes | |
| CN104899562B (en) | Radar remote sensing image culture's recognizer based on Texture Segmentation fusion | |
| CN101840581B (en) | Method for extracting profile of building from satellite remote sensing image | |
| CN104361590B (en) | High-resolution remote sensing image registration method with control points distributed in adaptive manner | |
| CN104361589A (en) | High-resolution remote sensing image segmentation method based on inter-scale mapping | |
| Zhou et al. | Individual tree parameters estimation for plantation forests based on UAV oblique photography | |
| CN108197583A (en) | The building change detecting method of optimization and image structure feature is cut based on figure | |
| CN107016403B (en) | A method of completed region of the city threshold value is extracted based on nighttime light data | |
| CN103839267B (en) | Building extracting method based on morphological building indexes | |
| CN104504718B (en) | The automatic method for extracting roads of high-resolution Airborne Data Classification | |
| CN107564017A (en) | A kind of city high score remote sensing image shadow Detection and dividing method | |
| CN110889840A (en) | Effectiveness detection method of high-resolution 6 # remote sensing satellite data for ground object target | |
| Chen et al. | Land cover mapping in urban environments using hyperspectral APEX data: A study case in Baden, Switzerland | |
| CN109360184A (en) | In conjunction with the remote sensing image variation detection method of shadow compensation and Decision fusion | |
| CN107992856A (en) | High score remote sensing building effects detection method under City scenarios | |
| CN109859219A (en) | In conjunction with the high score Remote Sensing Image Segmentation of phase and spectrum | |
| CN103456029B (en) | The Mean Shift tracking of a kind of anti-Similar color and illumination variation interference | |
| Ming et al. | Cropland extraction based on OBIA and adaptive scale pre-estimation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Gu Aihua Inventor before: Gu Aihua Inventor before: Wang Chao Inventor before: Li Shujun |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240326 Address after: 224000 No. 617, building 1, Longyuan Xincun business district, Yannan high tech Zone, Yancheng City, Jiangsu Province (CNW) Patentee after: Jiangsu Youji Technology Co.,Ltd. Country or region after: China Address before: 224002 No. 50, open avenue, Jiangsu, Yancheng City Patentee before: YANCHENG TEACHERS University Country or region before: China |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240703 Address after: Room 502, Unit 1, Building 2, Tongtai Meiling Garden, No. 181 Furong South Road, Tianxin District, Changsha City, Hunan Province, 410000 Patentee after: Hunan Liren Land Consulting Co.,Ltd. Country or region after: China Address before: 224000 No. 617, building 1, Longyuan Xincun business district, Yannan high tech Zone, Yancheng City, Jiangsu Province (CNW) Patentee before: Jiangsu Youji Technology Co.,Ltd. Country or region before: China |
|
| TR01 | Transfer of patent right |