CN114359232B - Image change detection method and device based on context covariance matrix - Google Patents
Image change detection method and device based on context covariance matrix Download PDFInfo
- Publication number
- CN114359232B CN114359232B CN202210014121.6A CN202210014121A CN114359232B CN 114359232 B CN114359232 B CN 114359232B CN 202210014121 A CN202210014121 A CN 202210014121A CN 114359232 B CN114359232 B CN 114359232B
- Authority
- CN
- China
- Prior art keywords
- change detection
- phase information
- time phase
- covariance matrix
- context
- 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.)
- Active
Links
- 230000008859 change Effects 0.000 title claims abstract description 217
- 238000001514 detection method Methods 0.000 title claims abstract description 156
- 239000011159 matrix material Substances 0.000 title claims abstract description 154
- 238000001914 filtration Methods 0.000 claims abstract description 31
- 238000000034 method Methods 0.000 claims abstract description 29
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 19
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 16
- 239000013598 vector Substances 0.000 claims description 45
- 238000010586 diagram Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 description 10
- 238000004458 analytical method Methods 0.000 description 7
- 238000010276 construction Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000010287 polarization Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The application relates to an image change detection method, an image change detection device, computer equipment and a storage medium based on a context covariance matrix. The method comprises the following steps: extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix, and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information; performing eigenvalue decomposition on the image change matrix, respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by using the obtained eigenvalues to obtain a plurality of initial change detection amounts, and selecting the largest initial change detection amount as a final change detection amount; and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map. By adopting the method, the accuracy of image change detection can be improved.
Description
Technical neighborhood
The present disclosure relates to the field of remote sensing image change detection technologies, and in particular, to an image change detection method, device, computer device, and storage medium based on a context covariance matrix.
Background
The remote sensing image change detection is a technology for determining whether a ground object target is changed or not by analyzing differences among remote sensing images of different moments in the same region. The change detection is helpful for quickly knowing the change condition of a certain area, and has wide application in the fields of environmental monitoring, disaster assessment, agricultural investigation, military reconnaissance and the like.
However, in image change detection, the generation of a difference map and the analysis of the difference map are two key steps. Traditional change detection studies have focused on methods of analysis of the difference map. In recent years, a more basic difference map generation stage has been paid attention to. Classical disparity map generation algorithms include the difference method and the ratio method. The difference method directly subtracts the pixel values of the corresponding positions of the different phase images, and the absolute value of each pixel in the obtained difference image represents the change degree and the sign represents the change direction. The difference method is widely applied to remote sensing image change detection because of simplicity and easiness, but is easily affected by noise. In contrast, the ratio operator can better overcome the defect of sensitivity to noise, but can cause the conditions of edge blurring and the like, and the method is not accurate enough in image change detection.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a method, apparatus, computer device, and storage medium for detecting image changes based on a context covariance matrix, which can improve the accuracy of detecting image changes.
A method of detecting image changes based on a context covariance matrix, the method comprising:
acquiring remote sensing images of different time phases; the different time phases refer to first time phase information and second time phase information;
extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the upper covariance matrix and the lower covariance matrix comprise a plurality of elements;
constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix, and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information;
performing eigenvalue decomposition on the image change matrix to obtain a plurality of eigenvalues;
respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by utilizing the characteristic values to obtain a plurality of initial change detection amounts;
comparing the initial change detection amounts, and selecting the largest initial change detection amount as the final change detection amount;
and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map.
In one embodiment, extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood includes:
extracting rectangular neighborhood with the size of (2n+1) x (2n+1) from the pixel point corresponding to the remote sensing image as a center; wherein n represents one natural number in the positive natural number set;
and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood.
In one embodiment, calculating a context covariance matrix of the remote sensing image from the rectangular neighborhood comprises:
constructing upper and lower Wen Sanshe vectors from rectangular adjacent domains, and calculating upper and lower Wen Xie variance matrices as by using upper and lower Wen Sanshe vectorsWherein N represents the number of contextual scattering vectors, superscript H Represents conjugate transpose, k i ＝[s 1 s 2 s 3 ] T Representing upper and lower Wen Sanshe vectors, s 2 Is the center pixel, s 1 、s 3 Is two neighborhood pixels symmetrical about the center pixel.
In one embodiment, constructing a plurality of feature quantities using a plurality of elements in a context covariance matrix, and constructing an image change matrix according to the feature quantities of first phase information and second phase information, includes:
constructing a plurality of feature quantities as using a plurality of elements in a context covariance matrix
And->Wherein c 11 、c 13 、c 31 、c 33 All represent elements in the context covariance matrix, c ij Representing the elements of the ith row and jth column of the context covariance matrix.
In one embodiment, an image change matrix is constructed based on the feature amounts of the first phase information and the second phase information asWherein->And->Two different characteristic quantities representing the first phase information, +.>And->Two different feature amounts representing the second phase information.
In one embodiment, calculating a bidirectional change detection amount between the first time phase information and the second time phase information by using the characteristic value to obtain a plurality of initial change detection amounts includes:
calculating the change detection amount from the first time phase information to the second time phase information by using the characteristic value to obtain the forward initial change detection amount asWherein lambda is 1 And lambda (lambda) 2 Representing the characteristic value;
calculating the change detection amount from the second time phase information to the first time phase information by using the characteristic value to obtain the reverse initial change detection amount as
In one embodiment, if the context covariance matrix is a directional context covariance matrix, constructing a context Wen Sanshe vector from the rectangular neighborhood;
the context scattering vector is according to the neighborhood pixel s 1 、s 3 The area is divided into 4 groups;
respectively calculating by using four groups of upper and lower Wen Sanshe vectors to obtain four directional context covariance matrixes;
respectively constructing four corresponding change matrixes for the four directional context covariance matrixes, and carrying out eigenvalue decomposition and calculation according to the four obtained change matrixes to obtain four change detection quantities;
and carrying out average calculation on the four change detection amounts to obtain initial change detection amounts of the first time phase information and the second time phase information.
An image change detection apparatus based on a context covariance matrix, the apparatus comprising:
the context covariance matrix module is used for acquiring remote sensing images of different time phases; the different time phases refer to first time phase information and second time phase information; extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the upper covariance matrix and the lower covariance matrix comprise a plurality of elements;
the image change matrix constructing module is used for constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information;
the bidirectional processing module is used for decomposing the characteristic values of the image change matrix to obtain a plurality of characteristic values; respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by utilizing the characteristic values to obtain a plurality of initial change detection amounts;
the normalization module is used for comparing a plurality of initial change detection amounts and selecting the largest initial change detection amount as a final change detection amount; and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map.
A computer device comprising a memory storing a computer program and a processor which when executing the computer program performs the steps of:
acquiring remote sensing images of different time phases; the different time phases refer to first time phase information and second time phase information;
extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the upper covariance matrix and the lower covariance matrix comprise a plurality of elements;
constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix, and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information;
performing eigenvalue decomposition on the image change matrix to obtain a plurality of eigenvalues;
respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by utilizing the characteristic values to obtain a plurality of initial change detection amounts;
comparing the initial change detection amounts, and selecting the largest initial change detection amount as the final change detection amount;
and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:
acquiring remote sensing images of different time phases; the different time phases refer to first time phase information and second time phase information;
extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the upper covariance matrix and the lower covariance matrix comprise a plurality of elements;
constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix, and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information;
performing eigenvalue decomposition on the image change matrix to obtain a plurality of eigenvalues;
respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by utilizing the characteristic values to obtain a plurality of initial change detection amounts;
comparing the initial change detection amounts, and selecting the largest initial change detection amount as the final change detection amount;
and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map.
According to the image change detection method, device, computer equipment and storage medium based on the context covariance matrix, the context Wen Xie variance matrix is calculated for the same remote sensing image in different time phases respectively, the intensity information and the texture information of the image are extracted and combined, then the context covariance matrix is utilized to construct a change matrix, then the change matrix is subjected to eigenvalue decomposition, the change detection quantity is calculated according to the obtained two eigenvalues, the change detection quantity is subjected to bidirectional processing, median filtering and normalization processing, and then the change detection difference map of fusion space information is obtained, so that a change region can be accurately identified, and the detection effect on the texture change region is good.
Drawings
FIG. 1 is a flow chart of a method for detecting image changes based on a context covariance matrix according to an embodiment;
FIG. 2 is a schematic diagram of an upper and lower Wen Sanshe vector construction in one embodiment;
FIG. 3 is a schematic diagram of directional context covariance matrix construction in one embodiment;
FIG. 4 is a schematic diagram of three images acquired by an on-board SAR UAVSAR in 2012 on 6 th month 17 th, 7 th month 3 th day, and 7 th month 17 th day;
FIG. 5 is a difference plot of data sets acquired from three images acquired by an on-board synthetic aperture radar UAVSAR on 6, 17, 7, 3, and 7, 17 days 2012 in another embodiment;
FIG. 6 is a schematic representation of an ROI analysis of three unchanged areas in another embodiment;
FIG. 7 is a schematic representation of ROI analysis of three regions of variation in another embodiment;
FIG. 8 is an intent of a difference graph contrast analysis obtained by the method of the present application with a difference operator and a log ratio operator in another embodiment;
FIG. 9 is a block diagram of an image change detection device based on a context covariance matrix according to one embodiment;
fig. 10 is an internal structural view of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
In one embodiment, as shown in fig. 1, there is provided an image change detection method based on a context covariance matrix, comprising the steps of:
102, acquiring remote sensing images of different time phases; the different time phases refer to first time phase information and second time phase information; extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the upper and lower covariance matrices include a plurality of elements.
The method comprises the steps of obtaining remote sensing images of different time phases, wherein the time phases refer to a certain time, extracting rectangular neighborhood from the remote sensing images, and then calculating a context covariance matrix of the time phases, wherein the context covariance matrix comprises the intensity information and texture information of the remote sensing images of the time phases, and the remote sensing image change areas are further identified through extracting the remote sensing image change information of two different time phases.
And 104, constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix, and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information.
The characteristic quantity comprises characteristic information of the remote sensing image, and an image change matrix can be constructed according to the difference between the characteristic quantities of different time phases, wherein the image change matrix comprises the characteristic information of image change.
Step 106, decomposing the characteristic values of the image change matrix to obtain a plurality of characteristic values; and respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by using the characteristic values to obtain a plurality of initial change detection amounts.
The extraction of two pieces of image change information is realized by performing eigenvalue decomposition on the image change matrix, and in order to avoid errors caused by the extraction direction, the initial change detection amount from the first time phase to the second time phase (forward direction) and the initial change detection amount from the second time phase to the first time phase (reverse direction) are calculated at the same time.
Step 108, comparing the plurality of initial change detection amounts, and selecting the largest initial change detection amount as a final change detection amount; and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map.
Comparing a plurality of initial change detection amounts, selecting the largest initial change detection amount as a final change detection amount, wherein the largest initial change detection amount contains all characteristic information of image change, and using the largest initial change detection amount as the final change detection amount can improve the accuracy of image change detection, filtering and suppressing noise on the final change detection amount by adopting median filtering, and normalizing the filtered final change detection amount to [0,1] to obtain a change detection difference map with more accurate fusion space information. The change detection difference map includes an unchanged region and a region which has changed, and the brighter the difference map is, the more the region changes significantly.
According to the image change detection method, the device, the computer equipment and the storage medium based on the context covariance matrix, the change area can be accurately identified by respectively calculating the context Wen Xie variance matrix for the same remote sensing image in different time phases, extracting the intensity information and the texture information of the combined image, constructing the change matrix by using elements in the context covariance matrix, decomposing the characteristic value of the change matrix, calculating the change detection amount according to the obtained two characteristic values, performing bidirectional processing, median filtering and normalization processing on the change detection amount, and obtaining the change detection difference map of the fused space information.
In one embodiment, extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood includes:
extracting rectangular neighborhood with the size of (2n+1) x (2n+1) from the pixel point corresponding to the remote sensing image as a center; wherein n represents one natural number in the positive natural number set;
and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood.
In one embodiment, calculating a context covariance matrix of the remote sensing image from the rectangular neighborhood comprises:
constructing upper and lower Wen Sanshe vectors from rectangular adjacent domains, and calculating upper and lower Wen Xie variance matrices as by using upper and lower Wen Sanshe vectorsWherein N represents the number of contextual scattering vectors, superscript H Represents conjugate transpose, k i ＝[s 1 s 2 s 3 ] T Representing upper and lower Wen Sanshe vectors, s 2 Is the center pixel, s 1 、s 3 Is two neighborhood pixels symmetrical about the center pixel.
As shown in fig. 2, a process of extracting contextual scattering vectors in a 3 x 3 neighborhood of an image is illustrated. Two pixels symmetrical about the center pixel are selected from the neighborhood, and together with the center pixel, a top-bottom Wen Sanshe vector is formed. Operating in this way, a total of four upper and lower Wen Sanshe vectors are available in the figure:
k 1 ＝[I 1,1 I 2,2 I 3,3 ] T
k 2 ＝[I 1,2 I 2,2 I 3,2 ] T
k 3 ＝[I 1,3 I 2,2 I 3,1 ] T
k 4 ＝[I 2,3 I 2,2 I 2,1 ] T
wherein I is i,j Representing pixel values located in the ith row and jth column of the neighborhood.
In one embodiment, constructing a plurality of feature quantities using a plurality of elements in a context covariance matrix, and constructing an image change matrix according to the feature quantities of first phase information and second phase information, includes:
constructing a plurality of feature quantities as using a plurality of elements in a context covariance matrix
And->Wherein c 11 、c 13 、c 31 、c 33 All represent elements in the context covariance matrix, c ij Representing the elements of the ith row and jth column of the context covariance matrix.
In one embodiment, an image change matrix is constructed based on the feature amounts of the first phase information and the second phase information asWherein->And->Two different characteristic quantities representing the first phase information, +.>And->Two different feature amounts representing the second phase information.
In one embodiment, the eigenvalue decomposition is performed on the image change matrix to obtain eigenvalue λ 1 And lambda (lambda) 2 。
Calculating bidirectional change detection amounts between the first time phase information and the second time phase information by using the characteristic values respectively to obtain a plurality of initial change detection amounts, wherein the method comprises the following steps:
calculating the change detection amount from the first time phase information to the second time phase information by using the characteristic value to obtain the forward initial change detection amount asWherein lambda is 1 And lambda (lambda) 2 Representing the characteristic value;
calculating the change detection amount from the second time phase information to the first time phase information by using the characteristic value to obtain the reverse initial change detection amount as
In one embodiment, if the context covariance matrix is a directional context covariance matrix, constructing a context Wen Sanshe vector from the rectangular neighborhood;
the context scattering vector is according to the neighborhood pixel s 1 、s 3 The area is divided into 4 groups;
respectively calculating by using four groups of upper and lower Wen Sanshe vectors to obtain four directional context covariance matrixes;
respectively constructing four corresponding change matrixes for the four directional context covariance matrixes, and carrying out eigenvalue decomposition and calculation according to the four obtained change matrixes to obtain four change detection quantities;
and carrying out average calculation on the four change detection amounts to obtain initial change detection amounts of the first time phase information and the second time phase information.
The construction method of the directional context covariance matrix is as follows:
constructing upper and lower Wen Sanshe vectors k from neighborhood i ＝[s 1 s 2 s 3 ] T ；
The context scattering vector is according to the neighborhood pixel s 1 、s 3 The area (direction relative to the center pixel) where the pixel is located is divided into 4 groups;
for each pixel point on the image, the directional context covariance matrix is characterized by four covariance matrices.
Therefore, the upper and lower Wen Xie variance matrices are calculated for each set of upper and lower Wen Sanshe vectorsA total of four directions of the context covariance matrix is obtained. Where d represents the direction (d=0°,45 °,90 °,135 °);
respectively constructing four corresponding change matrixes for the four directional context covariance matrixes, and carrying out eigenvalue decomposition and calculation according to the four obtained change matrixes to obtain four change detection quantities p d ，(d＝0°,45°,90°,135°)；
Finally, the change detection amounts in four directions are averaged, namelyAn initial change detection amount of the first phase information and the second phase information is obtained.
In another embodiment, if the input remote sensing image is a multi-channel image, the manner in which the contextual scattering vector is constructed will change accordingly. Taking three-channel images as an example, the contextual scattering vectors areWherein the superscripts I, II and III denote pixel values of which channels and the subscript denotes the position (s 2 Is the center pixel, s 1 、s 3 Are two neighborhood pixels symmetrical about the center pixel). The context covariance matrix can then be constructed in the same way.
A schematic diagram of a 7×7 neighborhood construction directional context covariance matrix is shown in fig. 3. The context scattering vector formed by the neighborhood pixels of the middle column region is 0 DEG direction, and the covariance matrix C of the direction can be constructed by the context scattering vector 0° . The covariance matrix construction method in other directions is the same. Thus, for one pixel, a total of 4 matrices C can be obtained for the directional context covariance matrix 0° 、C 45° 、C 90° 、C 135° 。
The image of HH polarization (labeled 0617, 0703, and 0717, respectively) of UAVSAR dataset 6 months 17, 7 months 3, and 7 months 17, in order from left to right in fig. 4. From the figure it can be seen that the images of the different phases have a significant variation.
The difference diagrams of 0617-0703 and 0617-0717 are obtained in order from left to right in fig. 5. As can be seen by comparing fig. 4, the portion near black on the difference map is hardly changed on the original image. Whereas for the brighter parts of the disparity map, the images of the two phases can be seen to be different. Therefore, the difference map obtained by the method can reflect the change condition of the image.
FIG. 6 is an analysis of ROIs 1-3. The intensity and Pauli plots for the two phases are shown. Comparison of each line of images shows that there is little difference between the intensity and Pauli plots for the two phases, and the corresponding difference plot is also black, indicating that there is substantially no change in this region.
FIG. 7 is an analysis of ROI 4-6. Also comparing the images of each row, the intensity map and Pauli map of the two phases were found to exhibit a distinct difference, and the corresponding difference map was shown to be a brighter hue, indicating that the region was significantly changed.
FIG. 8 is a comparison experiment result of the present application with a difference operator and a log ratio operator. Fig. 8 (a) is a detection effect diagram of detecting a difference diagram of 0617-0703 by using a difference operator, fig. 8 (b) is a detection effect diagram of detecting a difference diagram of 0617-0703 by using a logarithmic ratio operator, fig. 8 (c) is a detection effect diagram of detecting a difference diagram of 0617-0703 by using the present application, fig. 8 (d) is a detection effect diagram of detecting a difference diagram of 0617-0717 by using a difference operator, fig. 8 (e) is a detection effect diagram of detecting a difference diagram of 0617-0717 by using a logarithmic ratio operator, and positions marked by gray frames in the diagrams are areas where changes occur on some original diagrams, which are better detected by the present invention, but the comparison method fails to completely detect.
It should be understood that, although the steps in the flowchart of fig. 1 are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in fig. 1 may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor do the order in which the sub-steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of other steps or sub-steps of other steps.
In one embodiment, as shown in fig. 9, there is provided an image change detection apparatus based on a context covariance matrix, comprising: a context covariance matrix module 902, a build image change matrix module 904, a bi-directional processing module 906, and a normalization module 908, wherein:
a context covariance matrix module 902, configured to obtain remote sensing images in different time phases; the different time phases refer to first time phase information and second time phase information; extracting a rectangular neighborhood of the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the upper covariance matrix and the lower covariance matrix comprise a plurality of elements;
a construction image change matrix module 904, configured to construct a plurality of feature quantities by using a plurality of elements in the context covariance matrix, and construct an image change matrix according to the feature quantities of the first phase information and the second phase information;
the bidirectional processing module 906 is configured to perform eigenvalue decomposition on the image change matrix to obtain a plurality of eigenvalues; respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by utilizing the characteristic values to obtain a plurality of initial change detection amounts;
a normalization module 908, configured to compare the plurality of initial change detection amounts, and select the largest initial change detection amount as the final change detection amount; and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map.
In one embodiment, the context covariance matrix module 902 is further configured to extract a rectangular neighborhood of the remote sensing image, and calculate a context covariance matrix of the remote sensing image according to the rectangular neighborhood, including:
extracting rectangular neighborhood with the size of (2n+1) x (2n+1) from the pixel point corresponding to the remote sensing image as a center; wherein n represents one natural number in the positive natural number set;
and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood.
In one embodiment, the context covariance matrix module 902 is further configured to calculate a context covariance matrix of the remote sensing image according to the rectangular neighborhood, including:
constructing upper and lower Wen Sanshe vectors from rectangular adjacent domains, and calculating upper and lower Wen Xie variance matrices as by using upper and lower Wen Sanshe vectorsWherein N represents the number of contextual scattering vectors, superscript H Represents conjugate transpose, k i ＝[s 1 s 2 s 3 ] T Representing upper and lower Wen Sanshe vectors, s 2 Is the center pixel, s 1 、s 3 Is two neighborhood pixels symmetrical about the center pixel.
In one embodiment, the image change matrix constructing module 904 is further configured to construct a plurality of feature quantities using a plurality of elements in the context covariance matrix, and construct an image change matrix according to the feature quantities of the first phase information and the second phase information, including:
constructing a plurality of feature quantities as using a plurality of elements in a context covariance matrix
And->Wherein c 11 、c 13 、c 31 、c 33 All represent elements in the context covariance matrix, c ij Representing the elements of the ith row and jth column of the context covariance matrix.
In one embodiment, the image change matrix constructing module 904 is further configured to construct an image change matrix as a set of feature values of the first phase information and the second phase informationWherein->And->Two different characteristic quantities representing the first phase information, +.>And->Two different feature amounts representing the second phase information.
In one embodiment, the bidirectional processing module 906 is further configured to calculate bidirectional change detection amounts between the first phase information and the second phase information by using the feature values, to obtain a plurality of initial change detection amounts, including:
calculating the change detection amount from the first time phase information to the second time phase information by using the characteristic value to obtain the forward initial change detection amount asWherein lambda is 1 And lambda (lambda) 2 Representing the characteristic value;
calculating the change detection amount from the second time phase information to the first time phase information by using the characteristic value to obtain the reverse initial change detection amount as
In one embodiment, if the context covariance matrix is a directional context covariance matrix, constructing a context Wen Sanshe vector from the rectangular neighborhood;
the context scattering vector is according to the neighborhood pixel s 1 、s 3 The area is divided into 4 groups;
respectively calculating by using four groups of upper and lower Wen Sanshe vectors to obtain four directional context covariance matrixes;
respectively constructing four corresponding change matrixes for the four directional context covariance matrixes, and carrying out eigenvalue decomposition and calculation according to the four obtained change matrixes to obtain four change detection quantities;
and carrying out average calculation on the four change detection amounts to obtain initial change detection amounts of the first time phase information and the second time phase information.
For a specific definition of an image change detection device based on a context covariance matrix, reference may be made to the definition of an image change detection method based on a context covariance matrix hereinabove, and the description thereof will not be repeated. The above-described respective modules in an image change detection apparatus based on a context covariance matrix may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a terminal, and an internal structure diagram thereof may be as shown in fig. 10. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of detecting image variations based on a context covariance matrix. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structure shown in fig. 10 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In an embodiment a computer device is provided comprising a memory storing a computer program and a processor implementing the steps of the method of the above embodiments when the computer program is executed.
In one embodiment, a computer storage medium is provided, on which a computer program is stored which, when executed by a processor, implements the steps of the method of the above embodiments.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples merely represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (3)
1. A method for detecting image changes based on a context covariance matrix, the method comprising:
acquiring remote sensing images of different time phases; the different time phases refer to first time phase information and second time phase information;
extracting a rectangular neighborhood from the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the context covariance matrix comprises a plurality of elements;
constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix, and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information;
performing eigenvalue decomposition on the image change matrix to obtain a plurality of eigenvalues;
respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by utilizing the characteristic values to obtain a plurality of initial change detection amounts;
comparing the initial change detection amounts, and selecting the largest initial change detection amount as a final change detection amount;
filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference diagram;
calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood, wherein the method comprises the following steps:
constructing upper and lower Wen Sanshe vectors from the rectangular neighborhood, calculating upper and lower using the upper and lower Wen Sanshe vectorsWen Xie variance matrix isWherein N represents the number of contextual scattering vectors, the superscript H represents the conjugate transpose, k i ＝[s 1 s 2 s 3 ] T Represent either up or down Wen Sanshe vector, s 2 Is the center pixel, s 1 、s 3 Two neighborhood pixels symmetrical about the center pixel;
constructing a plurality of feature quantities by using a plurality of elements in the context covariance matrix, and constructing an image change matrix according to the feature quantities of the first time phase information and the second time phase information, wherein the method comprises the following steps:
constructing a plurality of feature quantities by using a plurality of elements in the context covariance matrix
And->Wherein c 11 、c 13 、c 31 、c 33 All represent elements in the context covariance matrix, c ij Elements representing the ith row and jth column in the context covariance matrix;
constructing an image change matrix according to the characteristic quantity of the first time phase information and the second time phase information asWherein->And->Two different characteristic quantities representing the first phase information, +.>And->Two different feature amounts representing second phase information;
calculating bidirectional change detection amounts between the first time phase information and the second time phase information by using the characteristic values respectively to obtain a plurality of initial change detection amounts, wherein the method comprises the following steps:
calculating the change detection amount from the first time phase information to the second time phase information by using the characteristic value to obtain a forward initial change detection amount asWherein lambda is 1 And lambda (lambda) 2 Representing the characteristic value;
calculating the change detection amount from the second time phase information to the first time phase information by using the characteristic value to obtain a reverse initial change detection amount as
If the context covariance matrix is a directional context covariance matrix, constructing a context Wen Sanshe vector from the rectangular neighborhood;
the contextual scattering vector is according to the neighborhood pixel s 1 、s 3 The area is divided into 4 groups;
respectively calculating by using four groups of upper and lower Wen Sanshe vectors to obtain four directional context covariance matrixes;
respectively constructing four corresponding change matrixes for the four directional context covariance matrixes, and carrying out eigenvalue decomposition and calculation according to the four obtained change matrixes to obtain four change detection quantities;
and carrying out average calculation on the four change detection amounts to obtain initial change detection amounts of the first time phase information and the second time phase information.
2. The method of claim 1, wherein extracting the rectangular neighborhood from the remote sensing image, and calculating the context covariance matrix of the remote sensing image from the rectangular neighborhood, comprises:
extracting rectangular neighborhood with the size of (2n+1) x (2n+1) from the pixel point corresponding to the remote sensing image by taking the pixel point as a center; wherein n represents one natural number in the positive natural number set;
and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood.
3. An image change detection apparatus based on a context covariance matrix, the apparatus comprising:
the context covariance matrix module is used for acquiring remote sensing images of different time phases; the different time phases refer to first time phase information and second time phase information; extracting a rectangular neighborhood from the remote sensing image, and calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood; the context covariance matrix comprises a plurality of elements; calculating a context covariance matrix of the remote sensing image according to the rectangular neighborhood, wherein the method comprises the following steps:
constructing upper and lower Wen Sanshe vectors from the rectangular neighborhood, calculating upper and lower Wen Xie variance matrices as using the upper and lower Wen Sanshe vectorsWherein N represents the number of contextual scattering vectors, the superscript H represents the conjugate transpose, k i ＝[s 1 s 2 s 3 ] T Represent either up or down Wen Sanshe vector, s 2 Is the center pixel, s 1 、s 3 Two neighborhood pixels symmetrical about the center pixel;
the image change matrix constructing module is used for constructing a plurality of characteristic quantities by utilizing a plurality of elements in the context covariance matrix and constructing an image change matrix according to the characteristic quantities of the first time phase information and the second time phase information; constructing a plurality of feature quantities by using a plurality of elements in the context covariance matrix
And->Wherein c 11 、c 13 、c 31 、c 33 All represent elements in the context covariance matrix, c ij Elements representing the ith row and jth column in the context covariance matrix; constructing an image change matrix to +.>Wherein->And->Two different characteristic quantities representing the first phase information, +.>And->Two different feature amounts representing second phase information;
the bidirectional processing module is used for carrying out eigenvalue decomposition on the image change matrix to obtain a plurality of eigenvalues; respectively calculating bidirectional change detection amounts between the first time phase information and the second time phase information by utilizing the characteristic values to obtain a plurality of initial change detection amounts; calculating the change detection amount from the first time phase information to the second time phase information by using the characteristic value to obtain a forward initial change detection amount asWherein lambda is 1 And lambda (lambda) 2 Representing the characteristic value;
calculating the change detection amount from the second time phase information to the first time phase information by using the characteristic value to obtain a reverse initial change detection amount asIf the context covariance matrix is a directional context covariance matrix, constructing a context Wen Sanshe vector from the rectangular neighborhood;
the contextual scattering vector is according to the neighborhood pixel s 1 、s 3 The area is divided into 4 groups;
respectively calculating by using four groups of upper and lower Wen Sanshe vectors to obtain four directional context covariance matrixes;
respectively constructing four corresponding change matrixes for the four directional context covariance matrixes, and carrying out eigenvalue decomposition and calculation according to the four obtained change matrixes to obtain four change detection quantities;
average calculation is carried out on the four change detection amounts to obtain initial change detection amounts of the first time phase information and the second time phase information;
the normalization module is used for comparing the plurality of initial change detection amounts and selecting the largest initial change detection amount as a final change detection amount; and filtering the final change detection quantity by adopting median filtering, and normalizing the obtained filtering result to obtain a change detection difference map.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210014121.6A CN114359232B (en) | 2022-01-06 | 2022-01-06 | Image change detection method and device based on context covariance matrix |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210014121.6A CN114359232B (en) | 2022-01-06 | 2022-01-06 | Image change detection method and device based on context covariance matrix |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114359232A CN114359232A (en) | 2022-04-15 |
| CN114359232B true CN114359232B (en) | 2024-02-02 |
Family
ID=81106865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210014121.6A Active CN114359232B (en) | 2022-01-06 | 2022-01-06 | Image change detection method and device based on context covariance matrix |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114359232B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117310705B (en) * | 2023-11-28 | 2024-02-09 | 中国石油大学(华东) | Flood disaster rapid detection method based on dual-polarized SAR image |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111199188A (en) * | 2019-12-18 | 2020-05-26 | 星际空间（天津）科技发展有限公司 | Pixel processing method and device for remote sensing image difference map, storage medium and equipment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019049324A1 (en) * | 2017-09-08 | 2019-03-14 | Nec Corporation | Image processing device, image processing method and storage medium |
-
2022
- 2022-01-06 CN CN202210014121.6A patent/CN114359232B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111199188A (en) * | 2019-12-18 | 2020-05-26 | 星际空间（天津）科技发展有限公司 | Pixel processing method and device for remote sensing image difference map, storage medium and equipment |
Non-Patent Citations (1)
| Title |
|---|
| 基于稀疏表示及上下文信息的PolSAR图像分类;孙良洁;中国优秀硕士学位论文全文数据库信息科技辑;I136-788 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114359232A (en) | 2022-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Feng et al. | Robust registration for remote sensing images by combining and localizing feature-and area-based methods | |
| Zheng et al. | Spatial-spectral-temporal connective tensor network decomposition for thick cloud removal | |
| Moghimi et al. | Tensor-based keypoint detection and switching regression model for relative radiometric normalization of bitemporal multispectral images | |
| CN113705532B (en) | Target detection method, device and equipment based on medium-low resolution remote sensing image | |
| CN116883466A (en) | Optical and SAR image registration method, device and equipment based on position sensing | |
| CN114359232B (en) | Image change detection method and device based on context covariance matrix | |
| AU2021102192A4 (en) | Remote sensing classification and identification method, device, equipment and storage medium for soil types | |
| Zhang et al. | Image super-resolution via a novel cascaded convolutional neural network framework | |
| Ferraris et al. | Robust fusion algorithms for unsupervised change detection between multi-band optical images—A comprehensive case study | |
| Wang et al. | Fast spatial-spectral random forests for thick cloud removal of hyperspectral images | |
| CN114004998A (en) | Unsupervised polarization SAR image terrain classification method based on multi-view tensor product diffusion | |
| CN117291790B (en) | SAR image registration method, SAR image registration device, SAR image registration equipment and SAR image registration medium | |
| Han et al. | Unsupervised multimodal change detection based on adaptive optimization of structured graph | |
| Wang et al. | CD-GAN: A robust fusion-based generative adversarial network for unsupervised remote sensing change detection with heterogeneous sensors | |
| CN113284153A (en) | Satellite cloud layer image processing method and device, computer equipment and storage medium | |
| CN116485858B (en) | Heterogeneous image registration method and device based on multi-scale cross-modal neighborhood descriptor | |
| Anfinsen et al. | Unsupervised heterogeneous change detection in radar images by cross-domain affinity matching | |
| CN116311120A (en) | Video annotation model training method, video annotation method, device and equipment | |
| CN117291957A (en) | Cross-modal light-SAR image registration method, device, equipment and medium | |
| CN113435384B (en) | Target detection method, device and equipment for medium-low resolution optical remote sensing image | |
| Cengiz et al. | The Effect of Super Resolution Method on Classification Performance of Satellite Images | |
| Zhang et al. | Shared contents alignment across multiple granularities for robust SAR-optical image matching | |
| Chen et al. | Urban damage estimation using statistical processing of satellite images | |
| CN116310899A (en) | YOLOv 5-based improved target detection method and device and training method | |
| CN116659520B (en) | Matching positioning method, device and equipment based on bionic polarization vision enhancement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |