CN106022262B - Remote Sensing Image Texture recognition methods and system based on interpolating wavelet - Google Patents
Remote Sensing Image Texture recognition methods and system based on interpolating wavelet Download PDFInfo
- Publication number
- CN106022262B CN106022262B CN201610338714.2A CN201610338714A CN106022262B CN 106022262 B CN106022262 B CN 106022262B CN 201610338714 A CN201610338714 A CN 201610338714A CN 106022262 B CN106022262 B CN 106022262B
- Authority
- CN
- China
- Prior art keywords
- image
- wavelet
- interpolation
- segmentation
- function
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 230000011218 segmentation Effects 0.000 claims abstract description 51
- 230000009466 transformation Effects 0.000 claims description 43
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000014509 gene expression Effects 0.000 claims description 4
- 230000015572 biosynthetic process Effects 0.000 claims description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims 4
- 230000009467 reduction Effects 0.000 abstract description 4
- 235000002017 Zea mays subsp mays Nutrition 0.000 description 11
- 240000008042 Zea mays Species 0.000 description 10
- 235000005824 Zea mays ssp. parviglumis Nutrition 0.000 description 10
- 235000005822 corn Nutrition 0.000 description 10
- 238000010586 diagram Methods 0.000 description 10
- 239000011159 matrix material Substances 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 230000007547 defect Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 3
- 241001057636 Dracaena deremensis Species 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 2
- 238000009395 breeding Methods 0.000 description 2
- 230000001488 breeding effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 241000482268 Zea mays subsp. mays Species 0.000 description 1
- 230000002146 bilateral effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 210000005069 ears Anatomy 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000010152 pollination Effects 0.000 description 1
- 238000012887 quadratic function Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000004800 variational method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/2163—Partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- 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
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20061—Hough transform
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of Remote Sensing Image Texture recognition methods and system based on interpolating wavelet, can carry out fast and accurately texture recognition to remote sensing images.The described method includes: being split using gray value threshold method to original farmland remote sensing images；It is rectangle by the image continuation that the segmentation obtains；Image in rectangular area is decomposed and reconstructed using multiple dimensioned inseparable interpolating wavelet, and noise reduction is carried out to the obtained image that reconstructs in the space Banach；Texture Segmentation is carried out to every block of image using gradient method, Hough transform is carried out to segmentation result, obtains class straight line texture recognition result.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a remote sensing image texture recognition method and system based on interpolation wavelets.
Background
In the corn field for seed production, every six rows of female corn plants are provided with one row of male plants at intervals, and in order to improve the breeding quality, the flower ears of the female corn plants are cut off after pollination. This is not required in high-yield field corn fields. Therefore, from the remote sensing image, the seed production corn field has a fuzzy straight line texture, as shown in fig. 1. Because the two corn fields have similar colors and gray values, the fuzzy straight-line-like texture in the seed corn field images becomes the main difference of the two corn field remote sensing images. Plots marked sm and gm in FIG. 1 are breeding corn and common corn plots, respectively.
Since a given image is a grayscale image and the texture of different plots is also relatively regular, it is difficult to distinguish from color and appearance. Two different plots, sm and gm, were carefully observed, and were slightly different in texture. To facilitate comparison and to find the difference between the two plots, the two plot images are stitched together as shown in FIG. 2 (sm on the left and gm on the right). sm has a more pronounced longitudinal linear texture, whereas gm has a faint, blurred transverse linear texture. Therefore, the selection of the linear texture as a basis for distinguishing different land areas is considered below.
Currently, popular image texture description algorithms include gray level co-occurrence matrix (GLCM), Gabor filter, direction wavelet, Curvelet (Curvelet) transform, shear wave (Shearlet) transform, and the like.
For directional wavelets and curvelets and shear waves, the description accuracy is high, but the wavelet transform speed is slow. The gray level co-occurrence matrix and the Gabor filter have similar properties, the texture in the specified direction is found through inner product operation, and the segmentation of the texture image is further realized through a clustering algorithm. The characteristics of such a division will be described below by taking a Gabor filter as an example. The image shown in fig. 2 is subjected to Gabor transform, and the transform result is shown in fig. 3. Obviously, the Gabor filter can capture the texture of sm land blocks very sensitively, but unlike what we observe, the captured texture is more focused on enclosing small areas, which has a direct relation with less specified directions in the algorithm, but the given directions are too many, which is necessary to reduce the algorithm efficiency as the directional wavelet type algorithm. In addition, even if the linear texture in the image can be accurately captured, the result of further adopting the clustering algorithm to segment the image is difficult to distinguish two different land parcels. The Gabor transform of the image shown in fig. 2 results in the image shown in fig. 3. The results do not show the straight line texture of the image, but do distinguish two different plots. If cluster segmentation is further applied, the result is shown in fig. 4. Therefore, the segmentation effect is not ideal, and the over-segmentation phenomenon appears.
Disclosure of Invention
In view of this, embodiments of the present invention provide a method and a system for identifying texture of a remote sensing image based on an interpolated wavelet, which can perform fast and accurate texture identification on the remote sensing image.
On one hand, the embodiment of the invention provides a remote sensing image texture recognition method based on interpolation wavelets, which comprises the following steps:
segmenting an original farmland remote sensing image by using a gray value threshold value method;
extending the image obtained by the segmentation into a rectangle;
decomposing and reconstructing an image in a rectangular region by using a multi-scale inseparable interpolation wavelet, and denoising the reconstructed image in a Banach space;
and (3) performing texture segmentation on each image by adopting a gradient method, and performing Hough transformation on a segmentation result to obtain a straight line-like texture recognition result.
On the other hand, the embodiment of the invention provides a remote sensing image texture recognition system based on an interpolation wavelet, which comprises the following steps:
the segmentation unit is used for segmenting the original farmland remote sensing image by utilizing a gray value threshold value method;
the continuation unit is used for extending the image obtained by the segmentation into a rectangle;
the reconstruction unit is used for decomposing and reconstructing the image in the rectangular region by utilizing the multi-scale insertable wavelet and denoising the reconstructed image in the Banach space;
and the transformation unit is used for performing texture segmentation on each image by adopting a gradient method and performing Hough transformation on the segmentation result to obtain a straight line-like texture recognition result.
The remote sensing image texture recognition method and system based on the interpolation wavelet provided by the embodiment of the invention have the advantages that on one hand, the image obtained by segmenting the original farmland remote sensing image is extended into a rectangle, so that the influence of the boundary on texture recognition during subsequent Hough transformation can be avoided, and on the other hand, by constructing the non-interpolation wavelet function, the defect that the common tensor product wavelet does not have directionality is overcome, the multi-scale interpolation operator designed on the basis can realize the directional decomposition and reconstruction of the image, and the reconstructed image is subjected to noise reduction in Banach space, so that false identification of a straight line direction caused by the existence of noise when Hough transformation is adopted to identify similar straight line textures subsequently is avoided, aliasing in Hough transformation can be avoided, and accurate texture identification of the remote sensing image can be realized by combining the two aspects. In addition, because the non-interpolation wavelet has interpolation characteristics, the transformation speed is higher than that of the wavelet with directions, and therefore the efficiency of image texture recognition is improved.
Drawings
FIG. 1 is a sample map of different parcel identifications;
FIG. 2 is a schematic diagram of two plots labeled SM and GM;
FIG. 3 shows Gabor transform results for different texture images;
FIG. 4 is a segmentation result of Gabor transform in combination with clustering algorithm;
FIG. 5 is a schematic flow chart of an embodiment of a remote sensing image texture recognition method based on an interpolated wavelet according to the present invention;
FIG. 6 is a schematic diagram before and after segmentation of an image by directly adopting a sobel operator;
FIG. 7 is a schematic diagram of a sobel operator before and after segmentation of a large image;
FIG. 8 is a diagram illustrating the recognition result of Hough transform on the quasi-linear texture in an image;
FIG. 9 is a triangular area;
FIG. 10 is a diagram illustrating the segmentation result obtained in S1 in FIG. 5;
FIG. 11 is a schematic view of the processed image of S2 in FIG. 5;
FIG. 12 is a schematic view of FIG. 5 after processing at S4;
fig. 13 is a schematic structural diagram of an embodiment of the remote sensing image texture recognition system based on the interpolated wavelet.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments, but not all embodiments, of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 5, the present embodiment discloses a remote sensing image texture recognition method based on interpolated wavelet, which includes:
s1, segmenting the original farmland remote sensing image by using a gray value threshold value method;
s2, extending the image obtained by segmentation into a rectangle;
s3, decomposing and reconstructing the image in the rectangular region by utilizing the multi-scale inseparable interpolation wavelet, and denoising the reconstructed image in a Banach space (Panach space);
in a specific application, a hard threshold method can be adopted for noise reduction.
And S4, performing texture segmentation on each image by adopting a gradient method (for example, the texture segmentation can be performed by using a sobel operator), and performing Hough transformation on a segmentation result (a binary image) to obtain a straight line-like texture recognition result.
Although the ideal segmentation effect cannot be obtained by using various existing continuous transformations, if the gradient method is directly used for segmentation, a sobel operator (sobel operator) can indeed obtain a more ideal linear texture segmentation effect (images before and after segmentation are respectively shown in (a) and (b) of fig. 6). Wherein the left half of the (a) and (b) images is an image marked by sm and the right half is an image marked by gm. Obviously, two kinds of land blocks can be simply divided from the number of the straight line textures.
The whole image is segmented by using a sobel operator, and only the land blocks with large gray level difference are segmented, so that the textures in the land blocks cannot be distinguished. In fact, the threshold of any segmentation algorithm is set for the whole image, and for the image given by the item, only messy segmentation results can be obtained even if the threshold of the segmentation algorithm is reduced. Therefore, before the linear texture segmentation is performed by using the sobel operator, the land blocks with different gray levels are firstly segmented according to the gray levels, and then the sobel segmentation is performed on the different land blocks, and the images before and after segmentation are respectively shown as (a) diagram and (b) diagram in fig. 7. Hough is an effective tool for identifying linear texture, Hough transformation is carried out on a sobel image segmentation result, geometric characteristics such as length and density of linear texture are calculated, and the geometric characteristics are used for identifying different land parcels, so that the method is simple, easy to implement and efficient.
Due to the fact that the straight-line-like texture is fuzzy, the direction is not unique and noise interference exists, the two different corn fields cannot be accurately identified through a common texture description method and a texture image segmentation method. Because the interpolation wavelet function is defined in Banach space, and the ordinary wavelet is defined in Hilbert space, the multi-scale interpolation wavelet transform has higher precision and application range compared with the ordinary wavelet transform. Therefore, a linear texture recognition method combining multi-scale interpolation wavelet and Hough transformation is constructed. The multi-scale interpolation wavelet transform can reduce noise points around the linear texture, thereby reducing the aliasing phenomenon in Houghting transform. Hough transformation is an effective means for detecting straight line textures in images, and identification rules of seed production corn field images are constructed through parameters such as density, length and parallelism of straight lines in the image textures, so that separation of the two corn field images is realized.
The Hough transformation is defined on the rectangular area, and when the Hough transformation is carried out, in order to avoid the influence of the boundary on texture recognition, the rectangular image area should not have a prominent boundary. While the tile image we extract is not a rectangular region, it may be a parallelogram or more irregular image region. The generalized variational principle can be adopted to reasonably extend the image, the reasonable extension should not affect the result of identifying the linear texture by the Hough transformation, and fig. 8 shows a schematic diagram of the identification result of the Hough transformation on the similar linear texture in the image.
The remote sensing image texture recognition method based on interpolation wavelet provided by the embodiment can avoid the influence of the boundary on texture recognition when Hough transformation is subsequently carried out by extending the image obtained by segmenting the original farmland remote sensing image into a rectangle, and can overcome the defect that common tensor product wavelet does not have directionality by constructing the non-interpolation wavelet function. In addition, because the non-interpolation wavelet has interpolation characteristics, the transformation speed is higher than that of the wavelet with directions, and therefore the efficiency of image texture recognition is improved.
Optionally, in another embodiment of the remote sensing image texture recognition method based on interpolated wavelet, the performing Hough transform on the segmentation result to obtain a straight line-like texture recognition result includes:
calculating the transformation attributes of each image block Hough, such as the density, the length, the parallelism and the like of a similar straight line;
determining the attribute of the corresponding image block according to the conversion attribute of the image block Hough;
and obtaining a similar straight line texture recognition result of the image block according to the attribute of the image block.
Hereinafter, a part of the process of the present invention will be described in detail.
(1) Constructing an indivisible interpolation wavelet
The function defined over the two-dimensional area Ω is denoted u (x, y). Let us say that the two-dimensional region omega has been triangulated, the vertices A of all the cellsiIs noted as (x)i,yi) The function value at these vertices is u (x)i,yi). Let given triangle E ═ A1A2A3And the midpoint B of the corresponding edge1,B2,B3It is required to interpolate the wavelet function U (x, y) so that U (A)i)＝u(Ai)，i＝1,2,3。
Because the interpolation conditions are 6, incomplete binary quadratic expressions can be selected as interpolation functions
Q(x,y)＝a1+a2x+a3y+a4xy+a5x2+a6y2 (1)
A in the above formula1、a2、a3、a4、a5And a6Is a constant term.
The form of the interpolated wavelet function U (x, y) can be expressed as:
whereinψiAll belong to function class (1) and satisfy
ψi(Aj)＝0,ψi(Bj)＝δi,j,i,j＝1,2,3，
Delta in the above formulai,jAnd deltai,jIs a constant term.
The following areIs given as an exampleψiThe method of construction of (1). See the triangular area shown in FIG. 9, due to A2,A3,B1On the y-axis; b2, B3On the straight line x is 0.5, soContains a factor x (x-0.5) to satisfyNeed to get
For the same reason haveLet t be 1-x-y, have
Similarly, available ψ1(x,y)＝4yt，ψ2(x,y)＝4xt，ψ3(x,y)＝4xy。
It is easy to see that the interpolation wavelet not only has interpolation characteristic and smoothness, but also has better directionality, overcomes the defect of isotropy based on tensor product wavelet, and has important function in the identification and reconstruction of texture images.
(2) Construction of multiscale interpolation operator based on inseparable interpolation wavelet function
From a sequence of wavelet functions psijkThe linear space of formation is defined as
For an arbitrary function f ∈ C0(0,1) one can always find a sufficiently large J (i.e., J) such that fj(x)∈VjSufficiently close to f (x), i.e.
j0Is a constant, coefficient βjkAnd αjkAre respectively defined as:
βj0k＝f(xj0k),αjk＝f(yjk)-Ijf(yjk) (4)
wherein xjkAnd yjkRepresenting discrete points, x, within a defined fieldjk∈[xmin,xmax],yjk∈[ymin,ymax]Then, then xmin、xmax、yminAnd ymaxIs a constant number of times, and is,
this equation shows wavelet coefficients αjkMeasures interpolated yjkThe error between the function value and the true value. In other words, the size of the interpolated wavelet coefficients reflects the local regularity of the analyzed function.
By utilizing interpolation wavelet transform theory, the interpolation operator I of multilayer interpolation wavelet can be obtainediThe calculation formula of (a) is as follows:
Zj＝0,1,2,...,2j (5)
wherein
j＝j0+1,j0+2,...,J-1,k＝0,1,2,...,2j-1 (6)
Obviously, when j is equal to j0Time of flight
For limiting the operator, define as
Since the limiting operator is known, the numerical solution of the interpolation operator can be easily obtained by using equations (5) to (8). The m-th derivative of the interpolation operator can be obtained by
The calculation formula of the adaptive interpolation operator of the multilayer interpolation wavelet is now given. Defining subsets of integersIndicating the configuration point index set corresponding to the wavelet with the wavelet coefficient larger than the threshold value epsilon at the j resolution level,and the index set represents configuration points corresponding to wavelets with wavelet coefficients larger than the threshold value epsilon on all resolution layers. Thus, it is only necessary to separatelyAnd ZcInstead of Z in the formulae (5) to (9)jAnd obtaining a calculation formula of the self-adaptive interpolation operator for solving the multilayer interpolation wavelet.
(3) Image decomposition and reconstruction based on multi-scale wavelet interpolation operator
And (3) image decomposition process:
f represents the gray value function of the image, and the interpolation wavelet coefficient α is not difficult to write by the interpolation wavelet transform theoryjkIs composed of
When j is equal to j0Time of flight
And (3) image reconstruction:
according to the definition of interpolation operator, approximating expression f of image fJCan be written as follows
Wherein Ii(x) Is an adaptive interpolation operator.
By means of multi-scale decomposition and reconstruction of the image, description and noise reduction processing of the image in the Banach space can be achieved.
(4) Image texture description based on generalized variational principle, Hough transformation and multi-scale interpolation wavelet decomposition and reconstruction
The Hough transform is a parameter estimation technique using the voting principle. The principle is that the detection problem in the image space is converted into the parameter space by using the point-line pair duality of the image space and the Hough parameter space. The straight line is detected by a method of performing simple accumulation statistics in a parameter space and then searching an accumulator peak value in a Hough parameter space. The essence of Hough transformation is to cluster pixels with a certain relationship in the image space and find the parameter space accumulation corresponding points which can link the pixels in a certain analytic form. This transformation has the desired effect in the case where the parameter space does not exceed two dimensions.
The application of the generalized variation principle aims to solve the problem of continuation of the land image.
Discussing the "variational principle" of a continuous medium problem, first a scalar functional pi is created, which is determined by the following form of integration
U is an unknown function, n is a positive integer, F and E are specific operators, omega is a solving domain, Γ is a boundary of omega, pi is called a functional of the unknown function u and changes along with the change of the function u, and the solution u of the continuous medium problem enables the functional pi to obtain a standing value for the tiny change deltau, namely the 'variation' of the functional is equal to zero
δΠ＝0 (14)
This method of solving the continuous medium problem is called the variational principle or variational method.
If u and its derivative have a maximum power of 2 in functional Π, the function is called a quadratic function. The functional in a large number of engineering and physical problems belongs to a quadratic functional.
The unknown function can be approximately expressed as
Where a is the parameter to be determined and N is the trial function.
Subject to additional constraints if the unknown function u is still
C (u) ═ 0 u is defined in Ω
Then another functional can be constructed
λ is a function vector of a set of independent coordinates in the Ω domain, called Lagrange multiplier, Π*Referred to as a correction functional. The modified functional constructed using Lagrange multipliers consists of two partial unknowns u and λ, both of which require a heuristic function to construct their approximate solution, e.g.
Where a, b are the parameters to be determined, N andis a trial function.
Correcting functional variations to zero to obtain a set of equations
c represents a vector formed by the parameters a and b to be determined, and two groups of parameters a and b can be obtained by solving an equation.
The Euler equation of the original functional pi is assumed to be
A(u)＝0
Wherein A represents the functional pi after variation treatment. (18)
With the proviso that the system of differential equations
C(u)＝L1(u)+C1＝0 (19)
Wherein L is1(u) denotes a linear differential operator, C1A constant representing the boundary condition is determined, and (17), (18) and (19) are substituted into the formula (15) to obtain
Since the above formula holds for all the variations δ a and δ b, there are
(21) The first term in the formula is an approximate equation of a natural variation of a linear equation set A (u) of 0
Ka＝P (23)
Equations (21) (22) can be approximated as the following set of equations
Wherein
K＝∫ΩNTA(N)dΩ (25)
Consider the one-dimensional problem u (x). Solving of domainsOmega is [ m, n ]]Without taking the number of discrete points as 2j+1, (where j ∈ Z), then the discrete point of variable x is defined as:
if a quasi-Shannon wavelet with interpolation characteristic is taken as a test function, that is to say
Where r is a constant, typically taken to be 3.2, then any element in the matrix K can be represented as
Kk,n＝A(wj(k-n)) (30)
Matrix GTAny one element in (1) can be represented as
WhereinIs a Lagrange basis function.
The wavelet transform is defined in a bilateral infinite interval, when the wavelet transform is performed on a limited-length image signal, the wavelet transform coefficient at the boundary is often very large, and the calculation amount and the calculation error are increased. To ensure the calculation accuracy, the scalar functional addition condition of the problem can be defined as
On the boundary Γ (32)
WhereinRepresenting a known boundary condition (which may be the gray value of the image at the boundary), then
WhereinIs a known value. Recalculating the formula (24) according to the additional condition to obtain
Matrix GTAny one element in (1) can be represented as
Due to the fact thatAnd N have interpolation characteristics, and according to the basic idea of the collocation method, in the formula (34), a is a vector formed by an approximate value of an unknown function u at a discrete point of x, and b is a vector formed by u towards extension points on two sides of a boundary. Assuming that in a given interval m, n]Taking R2j+1(j belongs to Z) discrete points x0,x1,…,x2jAnd L continuation points x are respectively taken at two sides of the boundary-L-1,x-L,…,x-1And xR,xR+1,…,xR+L。
The image can be extended to the rectangular area by utilizing the principle.
As shown in fig. 10, 11 and 12, which are schematic diagrams after processing in steps S1, S2 and S4, respectively, the present invention has the following advantages;
(1) an insertable wavelet function is constructed. Compared with tensor product wavelets, the method has good directivity, and lays a foundation for identification of farmland linear textures.
(2) The description of the image in the Banach space can be realized by adopting the inseparable multi-scale interpolation wavelet transform, so that the image processing range is greatly enlarged, and the method is particularly suitable for identifying and describing the texture of the farmland remote sensing image.
(3) The false recognition of the straight line direction caused by the existence of noise when the Hough transformation is adopted to recognize the straight line texture is avoided, and the aliasing phenomenon in the Hough transformation is solved.
(4) The generalized variational principle is adopted to extend the irregular image into rectangular image blocks, thus laying a foundation for the use of Hough transformation and not influencing the final texture recognition.
As shown in fig. 13, the present embodiment discloses a remote sensing image texture recognition system based on interpolated wavelet, which includes:
the dividing unit 1 is used for dividing the original farmland remote sensing image by using a gray value threshold value method;
the continuation unit 2 is used for extending the image obtained by the segmentation into a rectangle;
the reconstruction unit 3 is used for decomposing and reconstructing the image in the rectangular region by utilizing the multi-scale insertable wavelet and denoising the reconstructed image in the Banach space;
and the transformation unit 4 is used for performing texture segmentation on each image by adopting a gradient method and performing Hough transformation on the segmentation result to obtain a straight line-like texture recognition result.
The remote sensing image texture recognition system based on interpolation wavelet provided by the embodiment can avoid the influence of the boundary on texture recognition when Hough transformation is subsequently carried out by extending the image obtained by segmenting the original farmland remote sensing image into a rectangle, and can overcome the defect that common tensor product wavelet does not have directionality by constructing the non-interpolation wavelet function. In addition, because the non-interpolation wavelet has interpolation characteristics, the transformation speed is higher than that of the wavelet with directions, and therefore the efficiency of image texture recognition is improved.
Optionally, in another embodiment of the remote sensing image texture recognition system based on interpolated wavelet, the continuation unit is specifically configured to continue the image obtained by segmentation into a rectangle based on the principle of generalized variation.
Optionally, in another embodiment of the interpolated wavelet-based remote sensing image texture recognition system according to the present invention, the transformation unit is specifically configured to perform texture segmentation on each image by using a sobel operator.
Optionally, in another embodiment of the remote sensing image texture recognition system based on interpolated wavelet, the transformation unit is specifically configured to:
calculating the transformation attribute of each image block Hough;
determining the attribute of the corresponding image block according to the conversion attribute of the image block Hough;
and obtaining a similar straight line texture recognition result of the image block according to the attribute of the image block.
Optionally, in another embodiment of the interpolated wavelet based remote sensing image texture recognition system of the present invention, the transformation attributes include density, length and parallelism of the straight-like lines.
Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.
Claims (8)
1. A remote sensing image texture recognition method based on interpolation wavelet is characterized by comprising the following steps:
segmenting an original farmland remote sensing image by using a gray value threshold value method;
extending the image obtained by the segmentation into a rectangle;
decomposing and reconstructing an image in a rectangular region by using a multi-scale inseparable interpolation wavelet, and denoising the reconstructed image in a Banach space;
performing texture segmentation on each image by adopting a gradient method, and performing Hough transformation on a segmentation result to obtain a straight line-like texture identification result;
wherein, the extending the image obtained by the segmentation into a rectangle comprises:
extending the image obtained by segmentation into a rectangle based on the generalized variational principle;
the step of decomposing and reconstructing the image in the rectangular region by using the multi-scale inseparable interpolation wavelet specifically comprises the following steps:
step a, constructing an inseparable interpolation wavelet function:
the function defined over the two-dimensional region Ω is denoted u (x, y), and assuming that the two-dimensional region Ω has been triangulated with vertices A of all cellsiIs noted as (x)i,yi) The function value at these vertices is u (x)i,yi) Let a given triangle E ═ A1A2A3And the midpoint B of the corresponding edge1,B2,B3It is required to interpolate the wavelet function U (x, y) so that U (A)i)＝u(Ai)，i＝1,2,3，
Selecting incomplete binary quadratic as interpolation function
Q(x,y)＝a1+a2x+a3y+a4xy+a5x2+a6y2 (1)
A in formula (1)1、a2、a3、a4、a5And a6Is a constant term and is a constant number,
the form of the interpolated wavelet function U (x, y) is expressed as:
whereinψiAll belong to function class (1) and satisfy
ψi(Aj)＝0,ψi(Bj)＝δi,j,i,j＝1,2,3,
Wherein, deltai,jIs a constant term;
step b, constructing a multi-scale interpolation operator based on the non-interpolation wavelet function:
from a sequence of wavelet functions psijkThe linear space of formation is defined as
For an arbitrary function f ∈ C0(0,1) find a J (i.e., J) large enough for fj(x)∈VjSufficiently close to f (x), i.e.
j0Is a constant, coefficient βjkAnd αjkAre respectively defined as:
wherein x isjkAnd yjkRepresenting discrete points, x, within a defined fieldjk∈[xmin,xmax],yjk∈[ymin,ymax]Then, thenxmin、xmax、yminAnd ymaxIs a constant number of times, and is,
utilizing interpolation wavelet transform theory to obtain interpolation operator I of multilayer interpolation waveletiThe calculation formula of (a) is as follows:
wherein,
when j is equal to j0When the temperature of the water is higher than the set temperature,
for limiting the operator, define as
Wherein the m-order derivative of the interpolation operator is obtained by the following formula
Step c, image decomposition and reconstruction based on the multi-scale wavelet interpolation operator:
the image decomposition process specifically comprises the following steps:
solving for interpolated wavelet coefficients α according to the interpolated wavelet transform theoryjkThe formula is as follows:
wherein f represents the gray value function of the image
When j is equal to j0When the temperature of the water is higher than the set temperature,
the image reconstruction process specifically comprises the following steps:
solving an approximation expression f of the image f according to the definition of the interpolation operatorJThe formula is as follows
Wherein Ii(x) Is an adaptive interpolation operator.
2. The remote sensing image texture recognition method based on the interpolated wavelet as recited in claim 1, wherein the texture segmentation of each image by the gradient method comprises:
and (5) carrying out texture segmentation on each image by using a sobel operator.
3. The remote sensing image texture recognition method based on the interpolation wavelet as claimed in claim 1, wherein the Hough transformation is performed on the segmentation result to obtain a straight line-like texture recognition result, comprising:
calculating the transformation attribute of each image block Hough;
determining the attribute of the corresponding image block according to the conversion attribute of the image block Hough;
and obtaining a similar straight line texture recognition result of the image block according to the attribute of the image block.
4. The method for remotely sensing image texture recognition based on interpolated wavelet as recited in claim 3, wherein said transformation attributes comprise density, length and parallelism of straight-like lines.
5. A remote sensing image texture recognition system based on interpolation wavelet is characterized by comprising:
the segmentation unit is used for segmenting the original farmland remote sensing image by utilizing a gray value threshold value method;
the continuation unit is used for extending the image obtained by the segmentation into a rectangle;
the reconstruction unit is used for decomposing and reconstructing the image in the rectangular region by utilizing the multi-scale insertable wavelet and denoising the reconstructed image in the Banach space;
the transformation unit is used for carrying out texture segmentation on each image by adopting a gradient method and carrying out Hough transformation on a segmentation result to obtain a straight line-like texture identification result;
the continuation unit is specifically used for extending the image obtained by segmentation into a rectangle based on a generalized variational principle;
wherein the reconstruction unit is specifically configured to:
step a, constructing an inseparable interpolation wavelet function:
the function defined over the two-dimensional region Ω is denoted u (x, y), and assuming that the two-dimensional region Ω has been triangulated with vertices A of all cellsiIs noted as (x)i,yi) The function value at these vertices is u (x)i,yi) Let a given triangle E ═ A1A2A3And the midpoint B of the corresponding edge1,B2,B3It is required to interpolate the wavelet function U (x, y) so that U (A)i)＝u(Ai)，i＝1,2,3，
Selecting incomplete binary quadratic as interpolation function
Q(x,y)＝a1+a2x+a3y+a4xy+a5x2+a6y2 (1)
A in formula (1)1、a2、a3、a4、a5And a6Is a constant term and is a constant number,
the form of the interpolated wavelet function U (x, y) is expressed as:
whereinψiAll belong to function class (1) and satisfy
ψi(Aj)＝0,ψi(Bj)＝δi,j,i,j＝1,2,3,
Wherein, deltai,jIs a constant term;
step b, constructing a multi-scale interpolation operator based on the non-interpolation wavelet function:
from a sequence of wavelet functions psijkThe linear space of formation is defined as
For an arbitrary function f ∈ C0(0,1) find a J (i.e., J) large enough for fj(x)∈VjSufficiently close to f (x), i.e.
j0Is a constant, coefficient βjkAnd αjkAre respectively defined as:
wherein x isjkAnd yjkRepresenting discrete points, x, within a defined fieldjk∈[xmin,xmax],yjk∈[ymin,ymax]Then, thenxmin、xmax、yminAnd ymaxIs a constant number of times, and is,
using interpolation wavelet transform theory to obtain multiple layersInterpolation operator I of interpolated waveletsiThe calculation formula of (a) is as follows:
wherein,
when j is equal to j0When the temperature of the water is higher than the set temperature,
for limiting the operator, define as
Wherein the m-order derivative of the interpolation operator is obtained by the following formula
Step c, image decomposition and reconstruction based on the multi-scale wavelet interpolation operator:
the image decomposition process specifically comprises the following steps:
solving for interpolated wavelet coefficients α according to the interpolated wavelet transform theoryjkThe formula is as follows:
wherein f represents the gray value function of the image
When j is equal to j0When the temperature of the water is higher than the set temperature,
the image reconstruction process specifically comprises the following steps:
solving an approximation expression f of the image f according to the definition of the interpolation operatorJThe formula is as follows
Wherein Ii(x) Is an adaptive interpolation operator.
6. The interpolated wavelet based remote sensing image texture recognition system of claim 5, wherein said transformation unit is specifically configured to perform texture segmentation on each image using a sobel operator.
7. The remote sensing image texture recognition system based on interpolated wavelets according to claim 5, wherein the transformation unit is specifically configured to:
calculating the transformation attribute of each image block Hough;
determining the attribute of the corresponding image block according to the conversion attribute of the image block Hough;
and obtaining a similar straight line texture recognition result of the image block according to the attribute of the image block.
8. The interpolated wavelet based remote sensing image texture recognition system of claim 7, wherein the transformation attributes comprise the density, length and parallelism of straight-like lines.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610338714.2A CN106022262B (en) | 2016-05-20 | 2016-05-20 | Remote Sensing Image Texture recognition methods and system based on interpolating wavelet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610338714.2A CN106022262B (en) | 2016-05-20 | 2016-05-20 | Remote Sensing Image Texture recognition methods and system based on interpolating wavelet |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106022262A CN106022262A (en) | 2016-10-12 |
| CN106022262B true CN106022262B (en) | 2019-06-04 |
Family
ID=57096295
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610338714.2A Expired - Fee Related CN106022262B (en) | 2016-05-20 | 2016-05-20 | Remote Sensing Image Texture recognition methods and system based on interpolating wavelet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106022262B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107067426B (en) * | 2017-04-17 | 2018-02-06 | 东华理工大学 | A kind of image texture characteristic extracting method based on nested three-legged structure |
| CN107248187B (en) * | 2017-05-22 | 2020-12-08 | 武汉地大信息工程股份有限公司 | Method for quickly cutting and recombining three-dimensional model textures |
-
2016
- 2016-05-20 CN CN201610338714.2A patent/CN106022262B/en not_active Expired - Fee Related
Non-Patent Citations (2)
| Title |
|---|
| 基于采样三通道不可分小波的多光谱图像融合;刘斌等;《仪器仪表学报》;20150331;第645-648页 |
| 遥感图像中水上桥梁目标识别与毁伤分析研究;曹海梅;《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑》;20051115;正文第3-5,17-30,34-35页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106022262A (en) | 2016-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | A new wavelet-based image denoising using undecimated discrete wavelet transform and least squares support vector machine | |
| CN105574534A (en) | Significant object detection method based on sparse subspace clustering and low-order expression | |
| CN112308872B (en) | Image edge detection method based on multi-scale Gabor first derivative | |
| CN113506246B (en) | Concrete 3D printing component fine detection method based on machine vision | |
| Kamble et al. | Performance evaluation of wavelet, ridgelet, curvelet and contourlet transforms based techniques for digital image denoising | |
| Yang et al. | An improved weighted nuclear norm minimization method for image denoising | |
| Nandeesh et al. | Image fusion algorithms for medical images-a comparison | |
| CN106022262B (en) | Remote Sensing Image Texture recognition methods and system based on interpolating wavelet | |
| CN107451986B (en) | Single infrared image enhancement method based on fusion technology | |
| CN115994870A (en) | Image processing method for enhancing denoising | |
| Apdilah et al. | A study of Frei-Chen approach for edge detection | |
| Kim et al. | Image denoising feedback framework using split Bregman approach | |
| Gupta et al. | A noise robust edge detector for color images using hilbert transform | |
| Sayed et al. | Image object extraction based on curvelet transform | |
| Shire et al. | A review paper on: agricultural plant leaf disease detection using image processing | |
| Ayhan et al. | Spectral and spatial quality analysis in pan sharpening process | |
| Dhawan et al. | Single-image super-resolution using rational fractal interpolation and adaptive wiener filtering | |
| Kasturiwala et al. | Adaptive image superresolution for agrobased application | |
| Lindblad et al. | Exact linear time euclidean distance transforms of grid line sampled shapes | |
| CN106097272B (en) | Image processing method and device based on interpolation shearing wave | |
| CN109658421B (en) | Multi-target image segmentation method | |
| Kakadiya et al. | Shape Extraction methods for fruits: technical review | |
| Papari et al. | Reduced inverse distance weighting interpolation for painterly rendering | |
| CN107067374B (en) | Matrix completion two dimensional image processing method based on two-stage interpolation | |
| Dharani et al. | Morphological edge enhancement techniques on satillite image |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210610 Address after: 101500 room 208-826, Jugezhuang town government, Miyun District, Beijing Patentee after: Beijing Zhongjing Century Technology Co.,Ltd. Address before: 056038 No. 199 Guangming South Street, Hebei, Handan Patentee before: HEBEI University OF ENGINEERING |
|
| TR01 | Transfer of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190604 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |