CN104680151B - A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account - Google Patents
A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account Download PDFInfo
- Publication number
- CN104680151B CN104680151B CN201510108620.1A CN201510108620A CN104680151B CN 104680151 B CN104680151 B CN 104680151B CN 201510108620 A CN201510108620 A CN 201510108620A CN 104680151 B CN104680151 B CN 104680151B
- Authority
- CN
- China
- Prior art keywords
- mrow
- change
- pixel
- remote 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.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account, the present invention is made full use of according on panchromatic image the characteristics of snow cover, detection is changed using the change detection detection method based on textural characteristics, region of variation is extracted.The accumulated snow region tested on new and old image is extracted using the Chan Vese dividing methods based on Level Set Method, according to snow cover situation in the result and new and old image of change detection, remove pseudo- change caused by the change of part snow cover, extract region of variation, the accuracy of change detection and the automaticity of map revision are improved, the data update cycle is shortened.
Description
Technical field
The invention belongs to image processing field, it is related to a kind of remote sensing image processing and change detecting method, more particularly to one
Plant the panchromatic remote sensing image region of variation extracting method of high-resolution that taking snow covering influence into account based on high-resolution remote sensing image.
Background technology
The Northeast latitude is high, close to winter wind regime, it is one of the main accumulated snow area in China's winter, the accumulated snow of large area
There is important influence to water resource utilization, diastrous weather and the atmospheric circulation of Northeast Regional.Become to Northeast Regional
When changing detection, the covering of accumulated snow also generates strong influence to the result of change detection.The change of Snow-Cover causes
Change in testing result and produce the pseudo- change in part.The influence of snow cover how is reduced, the change of snow cover situation is removed and produces
Raw pseudo- change, more accurately extracts region of variation, is the difficult point of Northeast Regional change detection, for very cold region
Research have important meaning.
High-resolution remote sensing image extensive use, traditional change detection based on pixel has great limitation on yardstick
Property, a kind of object-based change detecting method arises at the historic moment, and what the change detecting method of object-oriented first had to solve is point
The problem of cutting, currently without the universality algorithm for high-resolution remote sensing image.When being changed detection to the Northeast, row
Except accumulated snow influence is even more important.Accumulated snow region how is determined, is to exclude the problem of accumulated snow influence first has to solve.Many scholars couple
Snow remote sensing knows method for distinguishing and has carried out beneficial exploration, has higher reflection in visible ray and near infrared band according to accumulated snow
Characteristic, the relatively low characteristic of reflectivity at middle-infrared band passes through certain digital image processing techniques and obtains snow cover
Information.Dozier proposes the snow lid index concept using accumulated snow reflectivity Characteristics, to distinguish accumulated snow and other atural objects.Hall
Complete snow-cover mapping SNOMAP algorithms have been developed using snow lid index.Shaded side accumulated snow can not due to the influence of the factors such as landform
Identification, can interpret the accumulated snow in shade according to the distribution of snow lid and the substantial connection of landform, research and propose correction illumination change
Method interpretation shade in accumulated snow.Any of the above algorithm respectively has its advantage and disadvantage, and institute can be applied to without any one algorithm
Some changes detect case.The characteristics of image that different sensors is obtained has respective, the scope of application is different, atural object itself
The influence of complexity and environmental factor so that the process of change detection becomes particularly complicated.Remote sensing image is used to the Northeast
Detection is changed, the covering of accumulated snow causes the difficulty increase of change detection, and single algorithm can not accurately extract change
Region.
The content of the invention
In order to overcome the deficiencies in the prior art, the present invention proposes a kind of panchromatic remote sensing of high-resolution for taking snow covering influence into account
Remote sensing imagery change detection method.The present invention use the Chan-Vese dividing methods based on level set, taken into full account pixel with
It faces the relation in domain, is partitioned into the accumulated snow region on double phase remote sensing images, by itself and the result that is detected based on group's strategy change
Compare, eliminate target to be detected due to the pseudo- change that Snow-Cover changes and produces, improve the precision of change detection.
The technical solution adopted in the present invention is：A kind of panchromatic remote sensing image change inspection of the high-resolution for taking snow covering influence into account
Survey method, it is characterised in that comprise the following steps：
Step 1：Input two width remote sensing images of same place different times；
Step 2：Detection is changed using group's strategy to two width remote sensing images of input；To two width remote sensing images of input
Accumulated snow region segmentation is carried out using the Chan-Vese dividing methods based on level set, it is considered to which pixel faces the relation in domain with it, segmentation
The accumulated snow region gone out on double phase remote sensing images, by itself and the results contrast detected based on group's strategy change, determines two width remote sensing
The change type of image, described change type is divided into：Change class, the class that do not change, pseudo- change class and suspicious do not change class；It is described
Change type determining standard be：
Change class：If two width remote sensing images are without snow, and testing result is that target is changed, then it can be assumed that being mesh
Mark has changed；
Do not change class：If two width remote sensing images are without snow, and testing result is that target does not change, then it can be assumed that target does not have
Change；
Puppet change class：If in two width remote sensing images one it is snowy, and testing result changes for target, this may is that by
In pseudo- change caused by snow covering, result of variations needs further examine；
It is suspicious not change class：If in two width remote sensing images one it is snowy, and testing result be target do not change, this
The situation of kind may is that because snow deposit masks the change of actual atural object, it is necessary to further detection；
Step 3：Take the panchromatic image change detection of snow into account, be classified as when on two width remote sensing images all by snow cover
Suspicious region, is not processed, and determines that detection target becomes due to the puppet that snow cover situation changes and produces during knock-out experiment
Change；
Step 4：Using overall accuracy, Kappa coefficients, change class accuracy and do not change class accuracy and commented to carry out precision
Valency.
Preferably, the two width remote sensing images to input described in step 2 are changed detection using group's strategy, it has
Body, which is realized, includes following sub-step：
Step 2.1.1：The textural characteristics figure of two described width remote sensing images of generation, calculates image greyscale co-occurrence matrix, by
Pixel takes texture eigenvalue needed for window calculation, and the texture eigenvalue of calculating then is assigned into central point pixel, after the completion of, will
Window moves a pixel, forms another new window, recalculates, the like, thus constitute one and artwork
The eigenvalue matrix of an equal-sized gray level co-occurrence matrixes；
Step 2.1.2：Using omnidirectional's texture value (i.e. the directive average value of institute), 0 °, 45 °, 90 °, 135 ° is calculated respectively
The average of the texture of four direction, compared with the texture template image of single direction, omnidirectional's texture has rotational invariance；
Step 2.1.3：The appropriate pixel distance of selection and window size obtain the textural characteristics needed；
Step 2.1.4：Texture feature extraction, is respectively adopted following three kinds of method of discrimination and obtains difference image, using group's plan
Slightly integrate three kinds of indexs and judge object variations；(i, j) is the gray value of different two pixels, and p (i, j) is gray level co-occurrence matrixes：
(4-a) second order is away from (Eneryg), the quadratic sum of gray level co-occurrence matrixes element value, and it is formulated as：
(4-b) entropy (Entropy), image has a measurement of the randomness of information content, represents the non-homogeneous of texture in image
Degree or complexity；If texture is complicated, entropy is big；If conversely, element size difference in uniform gray level in image, co-occurrence matrix
Greatly, entropy is small；The calculation formula of entropy is：
(4-c) homogeneity degree (Homogeneity) is also known as contrast square, reflects the uniformity of gradation of image；When pixel is to equal
When even, the value of homogeneity degree is of a relatively high；Its calculation formula is：
Step 2.1.5：Given threshold iterative, its algorithmic formula is f (i, j)=gt1(i,j)-gt2(i, j), wherein
gt1(i, j) and gt2(i, j) is the gray value of the pixel arranged on textural characteristics image in the i-th row jth, and f (i, j) is that texture is poor
It is worth the gray value of the pixel of the i-th row jth row on image, iterative method is chosen herein and calculates threshold value progress binaryzation, the mistake of iterative method
Cheng Wei：
Step 2.1.5.1：Choose the maximum of gray scale and the average T of minimum value on image and be used as initial threshold value；
Step 2.1.5.2：With T as Threshold segmentation image, grey scale pixel value is used as foreground part, pixel ash more than T
Angle value is used as background parts less than T；
Step 2.1.5.3：Calculate the gray average g of foreground part and background parts1And g2, new threshold value T ' is calculated, its
In：
When T ' and T difference is less than setting threshold values M, stop iteration, no person goes back to step 2.1.5.2 and continues to calculate, until
Condition is met, stops iteration, threshold value now is required；
Step 2.1.6：Group decision detects object variations, and single feature is respectively adopted and is changed detection, then to three
The result of individual change detection detection is merged, and obtains more accurately changing testing result, three kinds of testing result importance one
Cause, use majority rule, when there is two kinds or more to judge that targets to be detected are changed in three kinds of testing results,
Then think that target to be detected is changed.
Preferably, the appropriate pixel distance of selection and window size described in step 2.1.3 obtain the texture needed
Feature, wherein pixel distance 1, window size 5x5.
Preferably, setting threshold values M as 0.01 described in step 2.1.5.3.
Preferably, the two width remote sensing images to input described in step 2 are using the Chan-Vese based on level set points
Segmentation method carry out accumulated snow region segmentation, its implement including for：By causing the energy functional E in following formulaCVMinimum is realized
Segmentation：
Fixed λ1=λ2=0, μ are normal number；c1And c2It is the average value of the inside and outside pixel gray scales of curve C, to enable
Measure functional ECV(c1,c2, C) and it is minimum, useTo represent that curve C, x, y are the coordinate of theorem in Euclid space, then above formula can be represented
For：
Wherein δε(z) be regularization Di Lake functions, Hε(z) be Hai Shi functions regularization form,
Then the EVOLUTION EQUATION of level set is：
WithIt is the gray average of inside or outside of curve partial pixel.
Preferably, use overall accuracy, Kappa coefficients described in step 4, change class accuracy and not changing class just
True rate carries out precision evaluation, wherein：
Overall accuracy (OA) is the Baidu point ratio that the pixel number correctly classified accounts for all total pixel numbers, and it reflects total physical examination
The degree of accuracy of survey；The pixel number wherein correctly classified diagonally is distributed in confusion matrix, is M11 and M22 in table 1 below
With；The change pixel number correctly detected in table is M11；The number for the change pixel that error detection goes out is M12；Error detection
What is gone out does not change pixel number for M12；What is correctly detected does not change pixel number for M22；
Table 1
Overall accuracy OA calculation formula is expressed as：
Kappa coefficients are the methods of another evaluating precision, and the data calculation formula in confusion matrix is：
Accuracy is defined as the change class pixel detected and non-changing class pixel accounts for the percentage of very respective real image member number
Than；Its calculation formula is：
The present invention is made full use of according on panchromatic image the characteristics of snow cover, is detected using the change based on textural characteristics
Detection method is changed detection, extracts region of variation.Extracted using the Chan-Vese dividing methods based on Level Set Method
The accumulated snow region on new and old image is tested, according to snow cover situation in the result and new and old image of change detection, part is removed
Pseudo- change, extracts region of variation caused by snow cover change, improves the accuracy and map revision of change detection
Automaticity, shortens the data update cycle.
Brief description of the drawings
Fig. 1：For the schematic flow sheet of the embodiment of the present invention.
Fig. 2：For the aviation image of the somewhere different times 1m spatial resolutions of the embodiment of the present invention.
Fig. 3：Change testing result for the second moment of the embodiment of the present invention.
Fig. 4：For the Entropy Changes testing result figure of the embodiment of the present invention.
Fig. 5：Change testing result figure for the homogeneity degree of the embodiment of the present invention.
Fig. 6：Change testing result figure for three Fusion Features of the embodiment of the present invention.
Fig. 7：For the early stage image accumulated snow region of the same place different times of the embodiment of the present invention.
Fig. 8：For the later stage image accumulated snow region of the same place different times of the embodiment of the present invention.
Fig. 9：Change testing result for the textural characteristics of the embodiment of the present invention.
Figure 10：For the change testing result for taking snow covering influence into account of the embodiment of the present invention.
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 present embodiment is mainly changed detection using entropy, three textural characteristics of homogeneity degree and second moment, passes through group decision
Plurality rule the change testing result of each texture is merged to extract region of variation.With the Chan-Vese based on level set
Dividing method extracts the accumulated snow region on two width remote sensing images of same place different times, is carried out after the influence of snow removing is gone
Change detection.Its flow is shown in Fig. 1.
The resolution ratio that the present embodiment chooses image is 1m, and the size of image is 1000 × 1000 pixels.In this group image
Building, house, bare area, road are contained, change detection specific implementation step is as follows：
Step 1：Input two width remote sensing images of same place different times；
Step 2：Detection is changed using group's strategy to two width remote sensing images of input；To two width remote sensing images of input
Accumulated snow region segmentation is carried out using the Chan-Vese dividing methods based on level set, it is considered to which pixel faces the relation in domain with it, segmentation
The accumulated snow region gone out on double phase remote sensing images, by itself and the results contrast detected based on group's strategy change, determines two width remote sensing
The change type of image, described change type is divided into：Change class, the class that do not change, pseudo- change class and suspicious do not change class；
Described change type determining standard is：
Change class：If two width remote sensing images are without snow, and testing result is that target is changed, then it can be assumed that being mesh
Mark has changed；
Do not change class：If two width remote sensing images are without snow, and testing result is that target does not change, then it can be assumed that target does not have
Change；
Puppet change class：If in two width remote sensing images one it is snowy, and testing result changes for target, this may is that by
In pseudo- change caused by snow covering, result of variations needs further examine；
It is suspicious not change class：If in two width remote sensing images one it is snowy, and testing result be target do not change, this
The situation of kind may is that because snow deposit masks the change of actual atural object, it is necessary to further detection；
Detection is changed using group's strategy to two width remote sensing images of input in step 2, it is implemented including following
Sub-step：
Step 2.1.1：The textural characteristics figure of two described width remote sensing images of generation, calculates image greyscale co-occurrence matrix, by
Pixel takes texture eigenvalue needed for window calculation, and the texture eigenvalue of calculating then is assigned into central point pixel, after the completion of, will
Window moves a pixel, forms another new window, recalculates, the like, thus constitute one and artwork
The eigenvalue matrix of an equal-sized gray level co-occurrence matrixes；
Step 2.1.2：Using omnidirectional's texture value (i.e. the directive average value of institute), 0 °, 45 °, 90 °, 135 ° is calculated respectively
The average of the texture of four direction, compared with the texture template image of single direction, omnidirectional's texture has rotational invariance；
Step 2.1.3：The appropriate pixel distance of selection and window size obtain the textural characteristics needed, wherein pixel distance
For 1, window size is 5x5.
Step 2.1.4：Texture feature extraction, is respectively adopted following three kinds of method of discrimination and obtains difference image, using group's plan
Slightly integrate three kinds of indexs and judge object variations；(i, j) is the gray value of different two pixels, and p (i, j) is gray level co-occurrence matrixes；
(4-a) second order is away from (Eneryg), the quadratic sum of gray level co-occurrence matrixes element value, and it is formulated as：
(4-b) entropy (Entropy), image has a measurement of the randomness of information content, represents the non-homogeneous of texture in image
Degree or complexity；If texture is complicated, entropy is big；If conversely, element size difference in uniform gray level in image, co-occurrence matrix
Greatly, entropy is small；The calculation formula of entropy is：
(4-c) homogeneity degree (Homogeneity) is also known as contrast square, reflects the uniformity of gradation of image；When pixel is to equal
When even, the value of homogeneity degree is of a relatively high；Its calculation formula is：
Gray level co-occurrence matrixes are to describe the common method of texture by studying the spatial correlation characteristic of gray scale, and it is pixel
The matrix function of distance and angle, it is by calculating the correlation in image between certain distance and 2 gray scales of certain orientation
Property, to reflect integrated information of the image on direction, interval, amplitude of variation and speed.On the image of N × N sizes, certain point
(x, y), and deviate its another point (x+a, y+b), if this to gray value be (g1,g2).Point (x, y) is made in whole picture
Upper movement, then can obtain various (g1,g2) value, if the series of gray value is k, then (g1,g2) combination have k2Kind.For whole
Picture, counts each (g1,g2) number of times that value occurs, a square formation is then arranged in, with (g1,g2) occur it is total time
They are normalized to the probability P of appearance by number, and such square formation is referred to as gray level co-occurrence matrixes.Difference is taken apart from difference value (b, b)
Combinations of values, the joint probability matrix under different situations can be obtained.Gray level co-occurrence matrixes are carried out in experiment as follows
Normalization：
P (i, j, d, θ) is that gray scale is i and j, number of times of the picture point that distance is d and direction is θ to appearance.
Step 2.1.5：Given threshold iterative, its algorithmic formula is f (i, j)=gt1(i,j)-gt2(i, j), wherein
gt1(i, j) and gt2(i, j) is the gray value of the pixel arranged on textural characteristics image in the i-th row jth, and f (i, j) is that texture is poor
It is worth the gray value of the pixel of the i-th row jth row on image, iterative method is chosen herein and calculates threshold value progress binaryzation, the mistake of iterative method
Cheng Wei：
Step 2.1.5.1：Choose the maximum of gray scale and the average T of minimum value on image and be used as initial threshold value；
Step 2.1.5.2：With T as Threshold segmentation image, grey scale pixel value is used as foreground part, pixel ash more than T
Angle value is used as background parts less than T；
Step 2.1.5.3：Calculate the gray average g of foreground part and background parts1And g2, new threshold value T ' is calculated, its
In：
When T ' and T difference is less than setting threshold values M=0.01, stop iteration, no person goes back to step 2.1.5.2 and continues to count
Calculate, until meeting condition, stop iteration, threshold value now is required；
Step 2.1.6：Group decision detects object variations, and single feature is respectively adopted and is changed detection, then to three
The result of individual change detection detection is merged, and obtains more accurately changing testing result, three kinds of testing result importance one
Cause, use majority rule, when there is two kinds or more to judge that targets to be detected are changed in three kinds of testing results,
Then think that target to be detected is changed.
The Chan-Vese dividing methods based on level set are used to carry out accumulated snow area two width remote sensing images of input in step
Regional partition, its implement including for：By causing the energy functional E in following formulaCVMinimum realizes segmentation：
Fixed λ1=λ2=0, μ are normal number；c1And c2It is the average value of the inside and outside pixel gray scales of curve C, to enable
Measure functional ECV(c1,c2, C) and it is minimum, useTo represent that curve C, x, y are the coordinate of theorem in Euclid space, then above formula can be represented
For：
Wherein δε(z) be regularization Di Lake functions, Hε(z) be Hai Shi functions regularization form,
Then the EVOLUTION EQUATION of level set is：
WithIt is the gray average of inside or outside of curve partial pixel.
To level set imbedding function derivation：
Wherein △ t are time step, are had using central difference schemes：
Wherein k1And k2For the step-length on x and y directions.ForThere is equation below：
Wherein：
Full scale equation is substituted into obtain：
Compared with the result of gray level threshold segmentation, the accumulated snow split using the Chan-Vese methods based on level set
Region is more complete, has taken into full account that pixel faces the relation in domain with it, more conforms to the distribution of Snow Cover Over situation on ground.
Step 3：Take the panchromatic image change detection of snow into account, be classified as when on two width remote sensing images all by snow cover
Suspicious region, is not processed, and determines that detection target becomes due to the puppet that snow cover situation changes and produces during knock-out experiment
Change；
Step 4：Using overall accuracy, Kappa coefficients, change class accuracy and do not change class accuracy and commented to carry out precision
Valency；
Wherein overall accuracy (OA) is the Baidu point ratio that the pixel number correctly classified accounts for all total pixel numbers, and it is reflected always
The degree of accuracy that physical examination is surveyed；The pixel number wherein correctly classified diagonally is distributed in confusion matrix, be in table 1 below M11 and
M22 sum；The change pixel number correctly detected in table is M11；The number for the change pixel that error detection goes out is M12；It is wrong
What flase drop was measured does not change pixel number for M12；What is correctly detected does not change pixel number for M22；
Table 1
Overall accuracy OA calculation formula is expressed as
Kappa coefficients are the methods of another evaluating precision, and the data calculation formula in confusion matrix is：
Accuracy is defined as the change class pixel detected and non-changing class pixel accounts for the percentage of very respective real image member number
Than；Its calculation formula is：
Table 2 is the precision index of the evaluation of the embodiment of the present invention：
Table 2
Result in table 2 is each coefficient result of calculation when being changed detection to this group of data, wherein taking snow influence into account
As a result the accuracy of overall accuracy, Kappa coefficients and detection has small size raising, illustrates to take snow covering change detecting method pair into account
Precision is improved to some extent.
The reason for raising of precision is limited is during puppet is gone, to eliminate fraction and change snow cover simultaneously in itself
The ground target that situation has also changed, the precision to change detection exerts a certain influence.
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 (5)
1. a kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account, it is characterised in that including following
Step：
Step 1：Input two width remote sensing images of same place different times；
Step 2：Detection is changed using group's strategy to two width remote sensing images of input；Two width remote sensing images of input are used
Chan-Vese dividing methods based on level set carry out accumulated snow region segmentation, it is considered to which pixel faces the relation in domain with it, are partitioned into double
Accumulated snow region on phase remote sensing image, by itself and the results contrast detected based on group's strategy change, determines two width remote sensing images
Change type, described change type is divided into：Change class, the class that do not change, pseudo- change class and suspicious do not change class；Described change
Changing type criterion is：
Change class：If two width remote sensing images are without snow, and testing result is that target is changed, then it can be assumed that for target
Change；
Do not change class：If two width remote sensing images are without snow, and testing result is that target does not change, then it can be assumed that target is not sent out
Changing；
Puppet change class：If in two width remote sensing images one it is snowy, and testing result changes for target, and this may is that due to snow
Pseudo- change caused by covering, result of variations needs further examine；
It is suspicious not change class：If in two width remote sensing images one it is snowy, and testing result be target do not change, this feelings
Condition may is that because snow deposit masks the change of actual atural object, it is necessary to further detection；It is wherein described two distant to input
Sense image is changed detection using group's strategy, and it is implemented including following sub-step：
Step 2.1.1：The textural characteristics figure of two described width remote sensing images of generation, calculates image greyscale co-occurrence matrix, pixel-by-pixel
Texture eigenvalue needed for window calculation is taken, the texture eigenvalue of calculating is then assigned to central point pixel, after the completion of, by window
A mobile pixel, forms another new window, recalculates, the like, thus constitute one and artwork size
The eigenvalue matrix of an equal gray level co-occurrence matrixes；
Step 2.1.2：Using omnidirectional's texture value (i.e. the directive average value of institute), 0 °, 45 °, 90 °, 135 ° four is calculated respectively
The average of the texture in direction, compared with the texture template image of single direction, omnidirectional's texture has rotational invariance；
Step 2.1.3：The appropriate pixel distance of selection and window size obtain the textural characteristics needed；
Step 2.1.4：Texture feature extraction, is respectively adopted following three kinds of method of discrimination and obtains difference image, comprehensive using group's strategy
Close three kinds of indexs and judge object variations；(i, j) is the gray value of different two pixels, and p (i, j) is gray level co-occurrence matrixes；
(4-a) second order is away from (Eneryg), the quadratic sum of gray level co-occurrence matrixes element value, and it is formulated as：
<mrow>
<mi>E</mi>
<mi>n</mi>
<mi>e</mi>
<mi>r</mi>
<mi>y</mi>
<mi>g</mi>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<munder>
<mo>&Sigma;</mo>
<mi>j</mi>
</munder>
<mi>p</mi>
<msup>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>;</mo>
</mrow>
(4-b) entropy (Entropy), image has the measurement of the randomness of information content, represents the non-uniform degree of texture in image
Or complexity；If texture is complicated, entropy is big；Conversely, if element size difference is big in uniform gray level in image, co-occurrence matrix,
Entropy is small；The calculation formula of entropy is：
<mrow>
<mi>E</mi>
<mi>n</mi>
<mi>t</mi>
<mi>r</mi>
<mi>o</mi>
<mi>p</mi>
<mi>y</mi>
<mo>=</mo>
<mo>-</mo>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<munder>
<mo>&Sigma;</mo>
<mi>j</mi>
</munder>
<mi>p</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mi>l</mi>
<mi>o</mi>
<mi>g</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
(4-c) homogeneity degree (Homogeneity) is also known as contrast square, reflects the uniformity of gradation of image；When pixel to it is uniform when,
The value of homogeneity degree is of a relatively high；Its calculation formula is：
<mrow>
<mi>H</mi>
<mi>o</mi>
<mi>m</mi>
<mi>o</mi>
<mi>g</mi>
<mi>e</mi>
<mi>n</mi>
<mi>e</mi>
<mi>i</mi>
<mi>t</mi>
<mi>y</mi>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<munder>
<mo>&Sigma;</mo>
<mi>j</mi>
</munder>
<mfrac>
<mn>1</mn>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>-</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</mfrac>
<mi>p</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
Step 2.1.5：Given threshold iterative, its algorithmic formula is f (i, j)=gt1(i,j)-gt2(i, j), wherein gt1(i,
And g j)t2(i, j) is the gray value of the pixel arranged on textural characteristics image in the i-th row jth, and f (i, j) is texture Difference image
The gray value of the pixel of upper i-th row jth row, chooses iterative method and calculates threshold value progress binaryzation, the process of iterative method is herein：
Step 2.1.5.1：Choose the maximum of gray scale and the average T of minimum value on image and be used as initial threshold value；
Step 2.1.5.2：With T as Threshold segmentation image, grey scale pixel value is used as foreground part, grey scale pixel value more than T
Background parts are used as less than T；
Step 2.1.5.3：Calculate the gray average g of foreground part and background parts1And g2, new threshold value T ' is calculated,
Wherein：
When T ' and T difference is less than setting threshold values M, stop iteration, no person goes back to step 2.1.5.2 and continues to calculate, until meeting
Condition, stops iteration, and threshold value now is required；
Step 2.1.6：Group decision detects object variations, and single feature is respectively adopted and is changed detection, and then three are become
The result for changing detection detection is merged, and obtains more accurately changing testing result, and three kinds of testing result importance are consistent, adopt
It is majority rule, when there is two or more to judge that target to be detected is changed in three kinds of testing results,
Then think that target to be detected is changed；
Step 3：Take the panchromatic image change detection of snow into account, be classified as suspicious when on two width remote sensing images all by snow cover
Region, is not processed, and the pseudo- change that detection target is produced due to the change of snow cover situation is determined during knock-out experiment；
Step 4：Do not change using overall accuracy, Kappa coefficients, change class accuracy and class accuracy and carry out precision evaluation.
2. the high-resolution panchromatic remote sensing image variation detection method according to claim 1 for taking snow covering influence into account, it is special
Levy and be：The appropriate pixel distance of selection and window size described in step 2.1.3 obtain the textural characteristics needed, wherein as
Element distance is 1, and window size is 5x5.
3. the high-resolution panchromatic remote sensing image variation detection method according to claim 1 for taking snow covering influence into account, it is special
Levy and be：Threshold values M is set as 0.01 described in step 2.1.5.3.
4. the high-resolution panchromatic remote sensing image variation detection method according to claim 1 for taking snow covering influence into account, it is special
Levy and be：The two width remote sensing images to input described in step 2 are carried out using the Chan-Vese dividing methods based on level set
Accumulated snow region segmentation, its implement including for：By causing the energy functional E in following formulaCVMinimum realizes segmentation：
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<msup>
<mi>E</mi>
<mrow>
<mi>C</mi>
<mi>V</mi>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<msub>
<mi>c</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>c</mi>
<mn>2</mn>
</msub>
<mo>,</mo>
<mi>C</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>&mu;</mi>
<mo>&CenterDot;</mo>
<mi>L</mi>
<mi>e</mi>
<mi>n</mi>
<mi>g</mi>
<mi>t</mi>
<mi>h</mi>
<mrow>
<mo>(</mo>
<mi>C</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>&lambda;</mi>
<mn>1</mn>
</msub>
<mo>&CenterDot;</mo>
<munder>
<mo>&Integral;</mo>
<mrow>
<mi>i</mi>
<mi>n</mi>
<mi>s</mi>
<mi>i</mi>
<mi>d</mi>
<mi>e</mi>
<mrow>
<mo>(</mo>
<mi>C</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<mo>|</mo>
<msub>
<mi>u</mi>
<mn>0</mn>
</msub>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>c</mi>
<mn>1</mn>
</msub>
<mo>|</mo>
<mi>d</mi>
<mi>x</mi>
<mi>d</mi>
<mi>y</mi>
<mo>+</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>&lambda;</mi>
<mn>2</mn>
</msub>
<mo>&CenterDot;</mo>
<munder>
<mo>&Integral;</mo>
<mrow>
<mi>o</mi>
<mi>u</mi>
<mi>t</mi>
<mi>s</mi>
<mi>i</mi>
<mi>d</mi>
<mi>e</mi>
<mrow>
<mo>(</mo>
<mi>C</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<mo>|</mo>
<msub>
<mi>u</mi>
<mn>0</mn>
</msub>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>c</mi>
<mn>2</mn>
</msub>
<mo>|</mo>
<mi>d</mi>
<mi>x</mi>
<mi>d</mi>
<mi>y</mi>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>;</mo>
</mrow>
Fixed λ1=λ2=0, μ are normal number；c1And c2It is the average value of the inside and outside pixel gray scales of curve C, to make energy general
Letter ECV(c1,c2, C) and it is minimum, useTo represent that curve C, x, y are the coordinate of theorem in Euclid space, then above formula can be expressed as：
Wherein δε(z) be regularization Di Lake functions, Hε(z) be Hai Shi functions regularization form,
<mrow>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mn>1</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
<mi> </mi>
<mi>z</mi>
<mo>&GreaterEqual;</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
<mi> </mi>
<mi>z</mi>
<mo><</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mi>d</mi>
<mrow>
<mi>d</mi>
<mi>z</mi>
</mrow>
</mfrac>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
2
Then the EVOLUTION EQUATION of level set is：
WithIt is the gray average of inside or outside of curve partial pixel.
5. the high-resolution panchromatic remote sensing image variation detection method according to claim 1 for taking snow covering influence into account, it is special
Levy and be：Described in step 4 using overall accuracy, Kappa coefficients, change class accuracy and do not change class accuracy and carry out
Precision evaluation, wherein：
Overall accuracy (OA) is the Baidu point ratio that the pixel number correctly classified accounts for all total pixel numbers, and it reflects overall detection
The degree of accuracy；The pixel number wherein correctly classified eye diagonal in confusion matrix is distributed, the sum for being M11 and M22 in table 1 below；
The change pixel number correctly detected in table is M11；The number for the change pixel that error detection goes out is M12；Error detection goes out
Do not change pixel number for M12；What is correctly detected does not change pixel number for M22；
Table 1
Overall accuracy OA calculation formula is expressed as：
<mrow>
<mi>O</mi>
<mi>A</mi>
<mo>=</mo>
<mfrac>
<mrow>
<mi>M</mi>
<mn>11</mn>
<mo>+</mo>
<mi>M</mi>
<mn>22</mn>
</mrow>
<mi>T</mi>
</mfrac>
<mo>&times;</mo>
<mn>100</mn>
<mi>%</mi>
<mo>;</mo>
</mrow>
Kappa coefficients are the methods of another evaluating precision, and the data calculation formula in confusion matrix is：
<mrow>
<mi>K</mi>
<mi>a</mi>
<mi>p</mi>
<mi>p</mi>
<mi>a</mi>
<mo>=</mo>
<mfrac>
<mrow>
<mi>T</mi>
<mo>&times;</mo>
<mrow>
<mo>(</mo>
<mi>M</mi>
<mn>11</mn>
<mo>+</mo>
<mi>M</mi>
<mn>22</mn>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>&times;</mo>
<mi>C</mi>
<mo>+</mo>
<mi>B</mi>
<mo>&times;</mo>
<mi>D</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msup>
<mi>T</mi>
<mn>2</mn>
</msup>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>&times;</mo>
<mi>C</mi>
<mo>+</mo>
<mi>B</mi>
<mo>&times;</mo>
<mi>D</mi>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
Accuracy is defined as the change class pixel detected and non-changing class pixel accounts for the percentage of very respective real image member number；Its
Calculation formula is：
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mi>t</mi>
<mi>c</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<mi>M</mi>
<mn>11</mn>
</mrow>
<mi>A</mi>
</mfrac>
<mo>&times;</mo>
<mn>100</mn>
<mi>%</mi>
<mo>;</mo>
<msub>
<mi>P</mi>
<mrow>
<mi>t</mi>
<mi>u</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<mi>M</mi>
<mn>22</mn>
</mrow>
<mi>B</mi>
</mfrac>
<mo>&times;</mo>
<mn>100</mn>
<mi>%</mi>
<mo>.</mo>
</mrow>
4
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510108620.1A CN104680151B (en) | 2015-03-12 | 2015-03-12 | A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510108620.1A CN104680151B (en) | 2015-03-12 | 2015-03-12 | A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104680151A CN104680151A (en) | 2015-06-03 |
| CN104680151B true CN104680151B (en) | 2017-08-25 |
Family
ID=53315169
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510108620.1A Expired - Fee Related CN104680151B (en) | 2015-03-12 | 2015-03-12 | A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104680151B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107103610B (en) * | 2017-04-25 | 2019-12-17 | 自然资源部国土卫星遥感应用中心 | automatic detection method for suspicious region matched with satellite images in stereo mapping |
| CN107610114B (en) * | 2017-09-15 | 2019-12-10 | 武汉大学 | optical satellite remote sensing image cloud and snow fog detection method based on support vector machine |
| CN108564002B (en) * | 2018-03-22 | 2020-08-21 | 中国科学院遥感与数字地球研究所 | Method and system for detecting time sequence change of remote sensing image |
| CN110909730A (en) * | 2018-09-14 | 2020-03-24 | 恒景科技股份有限公司 | Imaging system for movement detection |
| CN110070525B (en) * | 2019-04-16 | 2021-01-01 | 湖北省水利水电科学研究院 | Remote sensing image change detection method based on object-level semi-supervised CV model |
| CN111640159B (en) * | 2020-05-11 | 2022-04-01 | 武汉大学 | Remote sensing image change detection method based on twin convolutional neural network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101126812A (en) * | 2007-09-27 | 2008-02-20 | 武汉大学 | High resolution ratio remote-sensing image division and classification and variety detection integration method |
| CN101950364A (en) * | 2010-08-30 | 2011-01-19 | 西安电子科技大学 | Remote sensing image change detection method based on neighbourhood similarity and threshold segmentation |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120328161A1 (en) * | 2011-06-22 | 2012-12-27 | Palenychka Roman | Method and multi-scale attention system for spatiotemporal change determination and object detection |
-
2015
- 2015-03-12 CN CN201510108620.1A patent/CN104680151B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101126812A (en) * | 2007-09-27 | 2008-02-20 | 武汉大学 | High resolution ratio remote-sensing image division and classification and variety detection integration method |
| CN101950364A (en) * | 2010-08-30 | 2011-01-19 | 西安电子科技大学 | Remote sensing image change detection method based on neighbourhood similarity and threshold segmentation |
Non-Patent Citations (2)
| Title |
|---|
| Automatic change detection in remote sensing images using level set method with neighborhood constraints;Guo Cao et al.;《Journal of Applied Remote Sensing》;20140214;全文 * |
| 基于图像融合和改进水平集模型的遥感图像变化检测;翟路;《中国优秀硕士学位论文全文数据库 信息科技辑》;20131215;第2013年卷(第S2期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104680151A (en) | 2015-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104680151B (en) | A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account | |
| CN105930868B (en) | A kind of low resolution airport target detection method based on stratification enhancing study | |
| CN103578110B (en) | Multiband high-resolution remote sensing image dividing method based on gray level co-occurrence matrixes | |
| CN106228125B (en) | Method for detecting lane lines based on integrated study cascade classifier | |
| CN103727930B (en) | A kind of laser range finder based on edge matching and camera relative pose scaling method | |
| CN104361590B (en) | High-resolution remote sensing image registration method with control points distributed in adaptive manner | |
| CN104392228A (en) | Unmanned aerial vehicle image target class detection method based on conditional random field model | |
| CN103353988B (en) | Allos SAR scene Feature Correspondence Algorithm performance estimating method | |
| CN109255781B (en) | Object-oriented multispectral high-resolution remote sensing image change detection method | |
| CN109766936A (en) | Image change detection method based on information transmitting and attention mechanism | |
| CN105354841B (en) | A kind of rapid remote sensing image matching method and system | |
| CN104613944A (en) | Distributed water depth prediction method based on GWR (geographically weighted regression) and BP (back propagation) neural network | |
| CN108492298A (en) | Based on the multispectral image change detecting method for generating confrontation network | |
| CN111339827A (en) | SAR image change detection method based on multi-region convolutional neural network | |
| CN104680169B (en) | A kind of semi-supervised diagnostic feature selection approach towards Hi-spatial resolution remote sensing image Extracting Thematic Information | |
| CN104182985A (en) | Remote sensing image change detection method | |
| CN109766824B (en) | Active and passive remote sensing data fusion classification method based on fuzzy evidence theory | |
| Shaoqing et al. | The comparative study of three methods of remote sensing image change detection | |
| CN103426158A (en) | Method for detecting two-time-phase remote sensing image change | |
| CN111598098A (en) | Water gauge water line detection and effectiveness identification method based on full convolution neural network | |
| CN104899892A (en) | Method for quickly extracting star points from star images | |
| CN108734122B (en) | Hyperspectral urban water body detection method based on self-adaptive sample selection | |
| CN104751185A (en) | SAR image change detection method based on mean shift genetic clustering | |
| CN105139375A (en) | Satellite image cloud detection method combined with global DEM and stereo vision | |
| CN105095913A (en) | Remote sensing image classification method and system based on neighbor regular joint sparse representation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170825 Termination date: 20200312 |