CN103927538A - Threshold selection method for improving spectral angle mapping precision - Google Patents
Threshold selection method for improving spectral angle mapping precision Download PDFInfo
- Publication number
- CN103927538A CN103927538A CN201410099909.7A CN201410099909A CN103927538A CN 103927538 A CN103927538 A CN 103927538A CN 201410099909 A CN201410099909 A CN 201410099909A CN 103927538 A CN103927538 A CN 103927538A
- Authority
- CN
- China
- Prior art keywords
- rad
- target image
- value
- threshold
- atural object
- 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
Abstract
The invention belongs to the technical field of remote sensing information science, and particularly discloses a threshold selection method for improving spectral angle mapping precision. The threshold selection method for improving the spectral angle mapping precision comprises the following steps of obtaining pure wave spectrum end member curves of a target image; classifying the wave spectrum end member curves, and building a wave spectrum curve set of each kind of ground features; calculating spectral angle radian similar values between different kinds of ground features; obtaining the maximum value of spectral angle similar radian value sets of every two target ground features; according to the maximum value of the spectral angle similar radian value sets, obtaining an upper limit value of extraction thresholds of the ground features; calculating spectral angle similar radian values of the same kind of ground features; according to the spectral angle similar radian values of the same kind of ground features, obtaining the minimum value of the similar radian value set of each kind of ground features; according to the minimum value of the similar radian value set of each kind of ground features, obtaining a lower limit value of the extraction thresholds of the ground features; comparing the upper limit value with the lower limit value, selecting the extraction thresholds, and guaranteeing accuracy and integrity of information extraction of the ground features. The method can guarantee the accuracy and the integrity of information extraction of the ground features of the target image.
Description
Technical field
The invention belongs to remote sensing information science technical field.Be specifically related to a kind of Research on threshold selection of Spectral angle mapper method.
Background technology
Spectral angle mapper method is in imaging spectrum treatment technology, one of main method that atural object is identified.This algorithm is that image wave spectrum is direct with a kind of interactive class method with reference to wave spectrum coupling, is the automatic classification method of ground-object spectrum in a kind of movement images wave spectrum and ground-object spectrum or wave spectrum storehouse.The method takes full advantage of the information of spectrum dimension, has emphasized the shape facility of spectrum, has greatly reduced characteristic information.
Current, Spectral angle mapper method has successfully been applied in the atural object identification in a plurality of fields, is mainly because the method is mainly paid close attention to the form of spectrum, has reduced the impact of spectral signature gain and drift.In Spectral angle mapper technology, the selection of threshold value is most important.Threshold value is little, and information extraction result has disappearance, and threshold value is large, and the accuracy of extraction reduces.Therefore, need a kind of new method to carry out selected threshold, in order to improve accuracy and the integrality of Spectral angle mapper result.
Summary of the invention
The object of the invention is to the setting defect for the threshold value at spectrum angle, a kind of Research on threshold selection that improves Spectral angle mapper precision is provided, the method can guarantee accuracy and the integrality of target image ground object information extraction.
For achieving the above object, technical scheme of the present invention is as follows: a kind of Research on threshold selection that improves Spectral angle mapper precision, and the method comprises the following steps:
Step 1, obtains the pure wave spectrum end member of target image curve;
Step 2, according to the form of wave spectrum curve and feature locations, classifies to the end member wave spectrum curve of the target image atural object obtaining in above-mentioned steps (1), determines the atural object species number of target image, and builds the wave spectrum collection of curves of every class target image atural object;
Step 3, utilize spectral analysis tool, calculate respectively spectrum angle radian similar value rad between different types of target image atural object A, B, C, wherein, rad (A, B), rad (A, C), rad (B, C) represent respectively spectrum angle radian similar value set between A and B, A and C, B and C tertiary target image two class atural objects;
Step 4, obtains spectrum angle similar radian value set rad (A, B), rad (A, C) between any two of A and B, A and C, B and C tertiary target image atural object in above-mentioned steps (3), the maximal value δ of rad (B, C)
aB, δ
aC, δ
bC;
Step 5, according to the maximal value δ of the approximate radian value set in the spectrum angle obtaining in above-mentioned steps (4)
aB, δ
aC, δ
bC, obtain the extraction threshold value A t of target image atural object A, B, C
hreshold, B
threshold, C
thresholdhigher limit A
thresholdmax, B
thresholdmax, C
thresholdmax;
Step 6, utilizes spectral analysis tool, calculates generic target image object spectrum angle approximate radian value rad (A, A), rad (B, B), rad (C, C);
Step 7: according to approximate radian value rad (A, A), rad (B, B), the rad (C, C) at the generic target image object spectrum angle obtaining in above-mentioned steps (6), obtain the minimum value β of the approximate radian value set of every class target image atural object
aA, β
bB, β
cC;
Step 8, according to the minimum value β of the approximate radian value set of the every class target image atural object obtaining in above-mentioned steps (7)
aA, β
bB, β
cC, obtain the extraction threshold value A of target image atural object A, B, C
threshold, B
threshold, C
thresholdlower limit A
thresholdmin, B
thresholdmin, C
thresholdmin;
Step 9: the extraction upper threshold value A relatively obtaining in above-mentioned steps (8)
thresholdmax, B
thresholdmax, C
thresholdmaxand the lower limit A obtaining in above-mentioned steps (5)
thresholdmin, B
thresholdmin, C
thresholdmithe size of n, thereby selective extraction threshold value, accuracy and the integrality of assurance target image ground object information extraction.
In described step (1), adopt the method for hourglass method or continuous maximum angular convex cone to obtain the pure end member wave spectrum of Target scalar curve.
The concrete steps of described step (2): silhouette target ground species has three kinds to be respectively A, B, C, and the wave spectrum collection of curves of every class silhouette target atural object A, B, C is expressed as SA={a
1, a
2, a
3a
i, SB={b
1, b
2, b
3b
j, SC={c
1, c
2, c
3c
k.
The concrete formula of rad (A, B) in described step (3), rad (A, C), rad (B, C) is as follows respectively:
δ in described step (4)
aB, δ
aC, δ
bCconcrete formula as follows respectively::
δ
AB＝max(rad(A,B))＝max{rad(a
1,b
1),rad(a
1,b
2),rad(a
1,b
3)...rad(a
i,b
j)}
δ
AC＝max(rad(A,C))＝max{rad(a
1,c
1),rad(a
1,c
2),rad(a
1,c
3)...rad(a
i,c
k)}
δ
BC＝max(rad(B,C))＝max{rad(b
1,c
1),rad(b
1,c
2),rad(b
1,c
3)...rad(b
j,c
k)}。
A in described step (5)
thresholdmax, B
thresholdmax, C
thresholdmaxconcrete formula as follows respectively:
A
thresholdmax＝1-Max(δ
AB,δ
AC)
B
thresholdmax＝1-Max(δ
AB,δ
BC)
C
thresholdmax＝1-Max(δ
AC,δ
BC)。
The concrete formula of rad (A, A) in described step (6), rad (B, B), rad (C, C) is as follows respectively:
Rad (A, A)={ rad (a
n, a
m), wherein, n=1,2 ... i-1, m=n+1, n+2 ... i and n ≠ m}
Rad (B, B)={ rad (b
n, b
m), wherein, n=1,2 ... j-1, m=n+1, n+2 ... j and n ≠ m}
Rad (C, C)={ rad (c
n, c
m), wherein, n=1,2 ... k-1, m=n+1, n+2 ... k and n ≠ m}
β in described step (7)
aA, β
bB, β
cCconcrete formula as follows respectively:
β
aA=min (rad (A, A))=min{rad (a
n, a
m), wherein, n=1,2 ... i-1, m=n+1, n+2 ... i and n ≠ m}
β
bB=min (rad (B, B))=min{rad (b
n, b
m), wherein, n=1,2 ... j-1, m=n+1, n+2 ... j and n ≠ m}
β
cC=min (rad (C, C))=min{rad (c
n, c
m), wherein, n=1,2 ... k-1, m=n+1, n+2 ... k and n ≠ m}.
A in described step (8)
thresholdmin, B
thresholdmin, C
thresholdminconcrete formula as follows respectively:
A
thresholdmin＝1-β
AA
B
thresholdmin＝1-β
BB
C
thresholdmin＝1-β
CC。
In described step (9), specifically comprise following two kinds of situations:
(9.1), when the higher limit of the extraction threshold value of certain class target image atural object is less than lower limit, cannot guarantee accuracy and the integrality of such target image ground object information extraction simultaneously;
(9.2) when the higher limit of certain class target image atural object is more than or equal to lower limit, the threshold value of the Spectral angle mapper of such target image atural object equals the arbitrary value in interval in these two values or capping value and lower limit, can guarantee integrality and the accuracy of such target image ground object information extraction simultaneously.
Beneficial effect of the present invention is as follows: method of the present invention is by higher limit and the lower limit of the threshold value in the map plotting method of chosen spectrum angle, and judge the size of higher limit and lower limit, can guarantee that extracted Target scalar does not exist the situation that is mixed with other atural objects, improve accuracy and the integrality of the Spectral angle mapper result of target image atural object, make up because of the improper phenomenon of Lou carrying and carrying that exists of threshold value selection by mistake, thus the precision of raising target image ground object information extraction.
Embodiment
Below in conjunction with accompanying drawing and case study on implementation, the present invention is described further.
A kind of Research on threshold selection that is applicable to improve Spectral angle mapper precision provided by the present invention, comprises the steps:
Step 1, obtains the pure wave spectrum end member of target image curve
The object of obtaining the pure wave spectrum end member of target image curve is not reduce under the prerequisite of nicety of grading, farthest reducing the operand in later stage.A common width image picture element number is thousands of or larger, and pure wave spectrum end member number is generally less than 100.
Can adopt the method for hourglass method or continuous maximum angular convex cone to obtain the pure end member wave spectrum of the Target scalar curve in actual image.
Step 2, according to the form of wave spectrum curve and feature locations, classifies to the end member wave spectrum curve of the target image atural object obtaining in above-mentioned steps (1), determines the atural object species number of target image, and builds the wave spectrum collection of curves of every class target image atural object.
Suppose that the pure wave spectrum end member number obtaining in step (1) is 50, actual silhouette target ground species has three kinds to be respectively A, B, C, and the wave spectrum collection of curves of every class silhouette target atural object A, B, C is expressed as SA={a
1, a
2, a
3a
i, SB={b
1, b
2, b
3b
j, SC={c
1, c
2, c
3c
k.Wherein a, b, c are respectively the loose point curve of two-dimentional wave spectrum of target image atural object A, B, C, a
1represent article one curve of atural object A, the like, i, j, k are respectively the ripple curve number of target image atural object A, B, C.0<i<50,0<j<50,0<k<50, and three's sum equals 50.
Step 3, utilizes spectral analysis tool, calculates respectively spectrum angle radian similar value (Radians Similarity is abbreviated as rad) between different types of target image atural object A, B, C
The span of rad is (0,1), retains 2 significant digits.The similarity degree that is worth larger two curves of spectrum is higher, and the radian similar value of identical two curves of spectrum is 1.
Rad (A, B), rad (A, C), rad (B, C) represent respectively spectrum angle radian similar value set between A and B, A and C, B and C tertiary target image two class atural objects, rad (A, B), rad (A, C), the concrete formula of rad (B, C) is as follows respectively:
Rad (a
1, b
1) represent article one wave spectrum curve spectrum angle radian similar value between any two in A, B two class atural objects, and rad (a
1, b
1)=rad (b
1, a
1).I in above-mentioned formula, j, k implication are identical with step (2).
The spectrum angle radian similar value of two class target object spectrum curves is less, shows that this two classes target image atural object separability is larger, and this two classes object spectrum tracing pattern differs greatly; Spectrum angle radian similar value is larger, proves that this two classes target image atural object separability is less, and this two classes object spectrum tracing pattern difference is less.
Adopt ENVI spectral analysis tool directly to calculate the spectrum angle radian similar value of two wave spectrum curves.
Step 4, obtains spectrum angle similar radian value set rad (A, B), rad (A, C) between any two of A and B, A and C, B and C tertiary target image atural object in above-mentioned steps (3), the maximal value δ of rad (B, C)
aB, δ
aC, δ
bC, concrete formula is as follows respectively:
δ
AB＝max(rad(A,B))＝max{rad(a
1,b
1),rad(a
1,b
2),rad(a
1,b
3)...rad(a
i,b
j)}
δ
AC＝max(rad(A,C))＝max{rad(a
1,c
1),rad(a
1,c
2),rad(a
1,c
3)...rad(a
i,c
k)}
δ
BC＝max(rad(B,C))＝max{rad(b
1,c
1),rad(b
1,c
2),rad(b
1,c
3)...rad(b
j,c
k)}
Wherein, δ
aB, δ
aC, δ
bCrepresented respectively the spectrum radian similarity between every two class target image atural objects.For category-A atural object, extract threshold value, work as δ
aB> δ
aC, illustrating that the separability of category-A atural object and C class atural object is larger, the threshold value of extracting category-A atural object should not be greater than 1-δ
aB, otherwise extract result, can sneak into part category-B atural object, the accuracy of extraction reduces; Together should δ
aB< δ
aC, illustrating that the separability of category-A atural object and category-B atural object is larger, the threshold value of extracting category-A atural object should not be greater than 1-δ
aC, otherwise extract result, can sneak into part C class atural object, the accuracy of extraction reduces.Other atural object extracts accuracy, and the i in above-mentioned formula, j, k implication are identical with step (2).
Step 5, according to the maximal value δ of the approximate radian value set in the spectrum angle obtaining in above-mentioned steps (4)
aB, δ
aC, δ
bC, obtain the extraction threshold value A of target image atural object A, B, C
threshold, B
threshold, C
thresholdhigher limit A
thresholdmax, B
thresholdmax, C
thresholdmax, A
thresholdmax, B
thresholdmax, C
thresholdmaxconcrete formula as follows respectively:
A
thresholdmax＝1-Max(δ
AB,δ
AC)
B
thresholdmax＝1-Max(δ
AB,δ
BC)
C
thresholdmax＝1-Max(δ
AC,δ
BC)
Suppose the δ in above-mentioned formula
aB, δ
aC, δ
bCbe respectively 0.83,0.91,0.88, by calculating A
thresholdmax=0.09, B
thresholdmax=0.12, C
thresholdmax=0.09.If do not calculated, do not extract upper threshold value, and directly adopt default threshold value in ENVI software (Maximum radians) 0.1 to extract Target scalar, can all be used as Target scalar and extract being more than or equal to 0.9 pixel with Target scalar curve of spectrum radian similarity.
While extracting category-A atural object, be taken as category-A atural object with the A similarity C class pixel that is 0.91 and extract.While extracting category-B atural object, the atural object that does not have other atural object classification is taken as the situation that category-B atural object extracts; While extracting C class atural object, be taken as C class atural object with the C similarity category-A pixel that is 0.91 and extract.
Step 6, utilize spectral analysis tool, calculate generic target image object spectrum angle approximate radian value rad (A, A), rad (B, B), rad (C, C), rad (A, A), rad (B, B), rad (C, C) represent respectively the approximate radian value set in spectrum angle of the curve of spectrum between A, B, C tertiary target image atural object class, concrete formula is as follows respectively:
Rad (A, A)={ rad (a
n, a
m), wherein, n=1,2 ... i-1, m=n+1, n+2 ... i and n ≠ m}
Rad (B, B)={ rad (b
n, b
m), wherein, n=1,2 ... j-1, m=n+1, n+2 ... j and n ≠ m}
Rad (C, C)={ rad (c
n, c
m), wherein, n=1,2 ... k-1, m=n+1, n+2 ... k and n ≠ m}
I in above-mentioned formula, j, k implication are identical with step (2), and m, n are integer variables, and in different ground species, span is different.
The degree of variation of object spectrum curve between the approximate radian value rad (A, A) at spectrum angle, rad (B, B), rad (C, C) representation class, the approximate radian value at spectrum angle is larger, shows that the variation of every class object spectrum curve is less, otherwise larger.
Step 7: according to approximate radian value rad (A, A), rad (B, B), the rad (C, C) at the generic target image object spectrum angle obtaining in above-mentioned steps (6), obtain the minimum value β of the approximate radian value set of every class target image atural object
aA, β
bB, β
cC, concrete formula is as follows respectively:
β
aA=min (rad (A, A))=min{rad (a
n, a
m), wherein, n=1,2 ... i-1, m=n+1, n+2 ... i and n ≠ m}
β
bB=min (rad (B, B))=min{rad (b
n, b
m), wherein, n=1,2 ... j-1, m=n+1, n+2 ... j and n ≠ m}
β
cC=min (rad (C, C))=min{rad (c
n, c
m), wherein, n=1,2 ... k-1, m=n+1, n+2 ... k and n ≠ m}
I in formula, j, k, m, n implication are identical with step (6).
Above-mentioned three value β
aA, β
bB, β
cCrepresented respectively the degree of approximation of spectrum radian in every class target image atural object, value is less represents that the wave spectrum variation of such atural object is larger, and when the Threshold value of extracting every class atural object should be more than or equal to 1-β, guarantee atural object extracts result and do not have the phenomenon of Lou putting forward.
Step 8, according to the minimum value β of the approximate radian value set of the every class target image atural object obtaining in above-mentioned steps (7)
aA, β
bB, β
cC, obtain the extraction threshold value A of target image atural object A, B, C
threshold, B
threshold, C
thresholdlower limit A
thresholdmin, B
thresholdmin, C
thresholdmin, A
thresholdmin, B
thresholdmin, C
thresholdminconcrete formula as follows respectively:
A
thresholdmin＝1-β
AA
B
thresholdmin＝1-β
BB
C
thresholdmin＝1-β
CC
Suppose the β in above-mentioned formula
aA,
β bB ,
β cC be respectively 0.93,0.88,0.89, by calculating A
thresholdmin=0.07, B
thresholdmin=0.12, C
thresholdmin=0.11.If do not calculated, do not extract threshold value lower limit, directly according to the higher limit A in step (5)
thresholdmax=0.09, B
thresholdmax=0.12, C
thresholdmax=0.09 threshold value that arbitrarily setting satisfies condition is A
threshold=0.05, B
threshold=0.05, C
threshold=0.05, these three threshold values can guarantee the accuracy that every class is extracted, but cannot guarantee to extract the integrality of result.
While extracting category-A atural object, be more than or equal to 0.95 goal pels with category-A typical curve similarity and be extracted, and the goal pels that degree is 0.93～0.95 is not similarly extracted; While extracting category-B atural object, be more than or equal to 0.95 goal pels with category-B typical curve similarity and be extracted, and the goal pels that degree is 0.88～0.95 is not similarly extracted; While extracting C class atural object, be more than or equal to 0.95 goal pels with category-B typical curve similarity and be extracted, and the goal pels that degree is 0.89～0.95 is not similarly extracted; So can only improve merely the accuracy of Spectral angle mapper classification according to upper threshold value, cannot guarantee the integrality of Spectral angle mapper classification.
Step 9: the extraction upper threshold value A relatively obtaining in above-mentioned steps (8)
thresholdmax, B
thresholdmax, C
thresholdmaxand the lower limit A obtaining in above-mentioned steps (5)
thresholdmin, B
thresholdmin, C
thresholdminsize, thereby selective extraction threshold value guarantees accuracy and the integrality of target image ground object information extraction.
(9.1), when the higher limit of the extraction threshold value of certain class target image atural object is less than lower limit, cannot guarantee accuracy and the integrality of such target image ground object information extraction simultaneously.
For example, the C class atural object in step (5) and step (8) extracts threshold value to have above-mentioned situation is C
thresholdmin=0.11, C
thresholdmax=0.09, C
thresholdmin> C
thresholdmax, in leaching process, can only meet the integrality of C class ground object information extraction, i.e. C
threshold=0.11; Or meet the accuracy of C class ground object information extraction, i.e. C
threshold=0.09.
(9.2), when the higher limit of certain class target image atural object is more than or equal to lower limit, can guarantee integrality and the accuracy of such target image ground object information extraction simultaneously;
When the higher limit of certain class target image atural object equals lower limit, the threshold value of the Spectral angle mapper of such target image atural object equals this two values.For example, the category-B atural object in step (5) and step (8) extracts threshold value to have above-mentioned situation is B
thresholdmin=0.12, B
thresholdmax=0.12, B
thresholdmin=B
thresholdmax, common setting threshold B in leaching process
threshold=0.12, can guarantee that category-B atural object extracts result and do not have the phenomenon of omitting and putting forward by mistake, can guarantee the integrality of category-B ground object information extraction; Can guarantee the accuracy of category-B ground object information extraction again simultaneously.
When the higher limit of certain class target image atural object is greater than lower limit, the threshold value of Spectral angle mapper is the arbitrary value in interval in higher limit and lower limit.For example, the category-A atural object in step (5) and step (8) extracts threshold value and has above-mentioned situation, i.e. A
thresholdmin=0.07, A
thresholdmax=0.09, A
thresholdmin< A
thresholdmaxif, at leaching process setting threshold A
threshold=0.08, can guarantee to extract result and not have the phenomenon of omitting and putting forward by mistake, can guarantee the integrality of category-A ground object information extraction; Can guarantee again integrality and accuracy that category-A atural object extracts simultaneously.
Son is explained in detail the present invention in conjunction with the embodiments above, but the present invention is not limited to above-described embodiment, in the ken possessing, can also under the prerequisite that does not depart from aim of the present invention, make a variety of changes those of ordinary skills.The content not being described in detail in the present invention all can adopt prior art.
Claims (10)
1. a Research on threshold selection that improves Spectral angle mapper precision, is characterized in that: the method comprises the following steps:
Step 1, obtains the pure wave spectrum end member of target image curve;
Step 2, according to the form of wave spectrum curve and feature locations, classifies to the end member wave spectrum curve of the target image atural object obtaining in above-mentioned steps (1), determines the atural object species number of target image, and builds the wave spectrum collection of curves of every class target image atural object;
Step 3, utilize spectral analysis tool, calculate respectively spectrum angle radian similar value rad between different types of target image atural object A, B, C, wherein, rad (A, B), rad (A, C), rad (B, C) represent respectively spectrum angle radian similar value set between A and B, A and C, B and C tertiary target image two class atural objects;
Step 4, obtains spectrum angle similar radian value set rad (A, B), rad (A, C) between any two of A and B, A and C, B and C tertiary target image atural object in above-mentioned steps (3), the maximal value δ of rad (B, C)
aB, δ
aC, δ
bC;
Step 5, according to the maximal value δ of the approximate radian value set in the spectrum angle obtaining in above-mentioned steps (4)
aB, δ
aC, δ
bC, obtain the extraction threshold value A of target image atural object A, B, C
threshold, B
threshold, C
thresholdhigher limit A
thresholdmax, B
thresholdmax, C
thresholdmax;
Step 6, utilizes spectral analysis tool, calculates generic target image object spectrum angle approximate radian value rad (A, A), rad (B, B), rad (C, C);
Step 7: according to approximate radian value rad (A, A), rad (B, B), the rad (C, C) at the generic target image object spectrum angle obtaining in above-mentioned steps (6), obtain the minimum value β of the approximate radian value set of every class target image atural object
aA, β
bB, β
cC;
Step 8, according to the minimum value β of the approximate radian value set of the every class target image atural object obtaining in above-mentioned steps (7)
aA, β
bB, β
cC, obtain the extraction threshold value A of target image atural object A, B, C
threshold, B
threshold, C
thresholdlower limit A
thresholdmin, B
thresholdmin, C
thresholdmin;
Step 9: the extraction upper threshold value At relatively obtaining in above-mentioned steps (8)
hresholdmax, B
thresholdmax, C
thresholdmaxand the lower limit A obtaining in above-mentioned steps (5)
thresholdmin, B
thresholdmin, C
thresholdminsize, thereby selective extraction threshold value guarantees accuracy and the integrality of target image ground object information extraction.
2. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 1, is characterized in that: in described step (1), adopt the method for hourglass method or continuous maximum angular convex cone to obtain the pure end member wave spectrum of Target scalar curve.
3. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 2, it is characterized in that: the concrete steps of described step (2): silhouette target ground species has three kinds to be respectively A, B, C, and the wave spectrum collection of curves of every class silhouette target atural object A, B, C is expressed as SA={a
1, a
2, a
3a
i, SB={b
1, b
2, b
3b
j, SC={c
1, c
2, c
3c
k.
4. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 3, is characterized in that: the concrete formula of the rad (A, B) in described step (3), rad (A, C), rad (B, C) is as follows respectively:
5. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 4, is characterized in that: the δ in described step (4)
aB, δ
aC, δ
bCconcrete formula as follows respectively::
δ
AB＝max(rad(A,B))＝max{rad(a
1,b
1),rad(a
1,b
2),rad(a
1,b
3)...rad(a
i,b
j)}
δ
AC＝max(rad(A,C))＝max{rad(a
1,c
1),rad(a
1,c
2),rad(a
1,c
3)...rad(a
i,c
k)}
δ
BC＝max(rad(B,C))＝max{rad(b
1,c
1),rad(b
1,c
2),rad(b
1,c
3)...rad(b
j,c
k)}。
6. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 5, is characterized in that: the A in described step (5)
thresholdmax, B
thresholdmax, C
thresholdmaxconcrete formula as follows respectively:
A
thresholdmax＝1-Max(δ
AB,δ
AC)
B
thresholdmax＝1-Max(δ
AB,δ
BC)
C
thresholdmax＝1-Max(δ
AC,δ
BC)。
7. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 6, is characterized in that: the concrete formula of the rad (A, A) in described step (6), rad (B, B), rad (C, C) is as follows respectively:
Rad (A, A)={ rad (a
n, a
m), wherein, n=1,2 ... i-1, m=n+1, n+2 ... i and n ≠ m}
Rad (B, B)={ rad (b
n, b
m), wherein, n=1,2 ... j-1, m=n+1, n+2 ... j and n ≠ m}
Rad (C, C)={ rad (c
n, c
m), wherein, n=1,2 ... k-1, m=n+1, n+2 ... k and n ≠ m}.
8. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 7, is characterized in that: the β in described step (7)
aA, β
bB, β
cCconcrete formula as follows respectively:
β
aA=min (rad (A, A))=min{rad (a
n, a
m), wherein, n=1,2 ... i-1, m=n+1, n+2 ... i and n ≠ m}
β
bB=min (rad (B, B))=min{rad (b
n, b
m), wherein, n=1,2 ... j-1, m=n+1, n+2 ... j and n ≠ m}
β
cC=min (rad (C, C))=min{rad (c
n, c
m), wherein, n=1,2 ... k-1, m=n+1, n+2 ... k and n ≠ m}.
9. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 8, is characterized in that: the A in described step (8)
thresholdmin, B
thresholdmin, C
thresholdminconcrete formula as follows respectively:
A
thresholdmin＝1-β
AA
B
thresholdmin＝1-β
BB
C
thresholdmin＝1-β
CC。
10. a kind of Research on threshold selection that improves Spectral angle mapper precision according to claim 9, is characterized in that: in described step (9), specifically comprise following two kinds of situations:
(9.1), when the higher limit of the extraction threshold value of certain class target image atural object is less than lower limit, cannot guarantee accuracy and the integrality of such target image ground object information extraction simultaneously;
(9.2) when the higher limit of certain class target image atural object is more than or equal to lower limit, the threshold value of the Spectral angle mapper of such target image atural object equals the arbitrary value in interval in these two values or capping value and lower limit, can guarantee integrality and the accuracy of such target image ground object information extraction simultaneously.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410099909.7A CN103927538B (en) | 2014-03-18 | 2014-03-18 | Threshold selection method for improving spectral angle mapping precision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410099909.7A CN103927538B (en) | 2014-03-18 | 2014-03-18 | Threshold selection method for improving spectral angle mapping precision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103927538A true CN103927538A (en) | 2014-07-16 |
| CN103927538B CN103927538B (en) | 2017-05-10 |
Family
ID=51145754
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410099909.7A Active CN103927538B (en) | 2014-03-18 | 2014-03-18 | Threshold selection method for improving spectral angle mapping precision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103927538B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105372672A (en) * | 2015-12-07 | 2016-03-02 | 广州地理研究所 | Time sequence data-based southern winter crop planting area extraction method |
| CN106846291A (en) * | 2015-12-04 | 2017-06-13 | 核工业北京地质研究院 | A kind of method that threshold value is chosen in spectral modeling matching |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6940999B2 (en) * | 2000-06-13 | 2005-09-06 | American Gnc Corp. | Method for target detection and identification by using proximity pixel information |
| CN101950361A (en) * | 2010-09-06 | 2011-01-19 | 中国科学院遥感应用研究所 | Adaptive extraction method of remote sensing image thematic information based on spectrum matching degree |
| CN102708544A (en) * | 2012-04-21 | 2012-10-03 | 西北工业大学 | Spectral angle mapping method used for correcting negative correlation of hyperspectral remote sensing image by wavebands |
| CN103559493A (en) * | 2013-10-21 | 2014-02-05 | 中国农业大学 | Method for extracting linear ground objects |
-
2014
- 2014-03-18 CN CN201410099909.7A patent/CN103927538B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6940999B2 (en) * | 2000-06-13 | 2005-09-06 | American Gnc Corp. | Method for target detection and identification by using proximity pixel information |
| CN101950361A (en) * | 2010-09-06 | 2011-01-19 | 中国科学院遥感应用研究所 | Adaptive extraction method of remote sensing image thematic information based on spectrum matching degree |
| CN102708544A (en) * | 2012-04-21 | 2012-10-03 | 西北工业大学 | Spectral angle mapping method used for correcting negative correlation of hyperspectral remote sensing image by wavebands |
| CN103559493A (en) * | 2013-10-21 | 2014-02-05 | 中国农业大学 | Method for extracting linear ground objects |
Non-Patent Citations (2)
| Title |
|---|
| 冯维一等: "基于三维光谱角统计的光谱图像信息提取方法", 《光谱学与光谱分析》 * |
| 樊彦国等: "基于多波段分析的无阈值自动光谱角制图分类法", 《地理与地理信息科学》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846291A (en) * | 2015-12-04 | 2017-06-13 | 核工业北京地质研究院 | A kind of method that threshold value is chosen in spectral modeling matching |
| CN105372672A (en) * | 2015-12-07 | 2016-03-02 | 广州地理研究所 | Time sequence data-based southern winter crop planting area extraction method |
| CN105372672B (en) * | 2015-12-07 | 2017-09-15 | 广州地理研究所 | Southern winter kind crops planting area extracting method based on time series data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103927538B (en) | 2017-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9754192B2 (en) | Object detection utilizing geometric information fused with image data | |
| US8934669B2 (en) | Self-adaptive image-based obstacle detection method | |
| JP2016110232A (en) | Object recognition device, object recognition method, and program | |
| US9477899B2 (en) | System and method for differentiating type of vegetation from remotely sensed data | |
| CN104966054B (en) | Detection method of small target in unmanned plane visible images | |
| CN105574521A (en) | House contour extraction method and apparatus thereof | |
| Sadiq et al. | Edge detection: A collection of pixel based approach for colored images | |
| CN105787470A (en) | Method for detecting power transmission line tower in image based on polymerization multichannel characteristic | |
| CN103810716A (en) | Image segmentation method based on grey scale flitting and Renyi entropy | |
| JP2016194858A5 (en) | ||
| CN105141913A (en) | Method and system for visually and remotely controlling touch control equipment and relevant equipment | |
| CN108121972A (en) | A kind of target identification method under the conditions of partial occlusion | |
| CN105184317A (en) | License plate character segmentation method based on SVM classification | |
| CN103927538A (en) | Threshold selection method for improving spectral angle mapping precision | |
| JP5712801B2 (en) | Image feature amount extraction device and marker detection device using image processing using the same | |
| Choudhary et al. | Image processing algorithm for fruit identification | |
| Yang et al. | Field-based plot extraction using UAV RGB images | |
| CN106485639A (en) | The method and apparatus differentiating forged certificate picture | |
| CN103533332A (en) | Image processing method for converting 2D video into 3D video | |
| CN103679724A (en) | Slope approximant straight line detection method | |
| CN104851102A (en) | Infrared small target detection method based on human visual system | |
| JP6326622B2 (en) | Human detection device | |
| Bo et al. | Research on weed recognition method based on invariant moments | |
| TW201627912A (en) | Image processing system and method for license plate recognition | |
| Li et al. | A soft shadow detection method based on MRF for remote sensing images |
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 |