CN117911253B - Polarized SAR image processing method and device - Google Patents
Polarized SAR image processing method and device Download PDFInfo
- Publication number
- CN117911253B CN117911253B CN202410319412.5A CN202410319412A CN117911253B CN 117911253 B CN117911253 B CN 117911253B CN 202410319412 A CN202410319412 A CN 202410319412A CN 117911253 B CN117911253 B CN 117911253B
- Authority
- CN
- China
- Prior art keywords
- sample
- representing
- characteristic
- optimal
- moving
- 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
- 238000003672 processing method Methods 0.000 title abstract description 3
- 238000001914 filtration Methods 0.000 claims abstract description 67
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 65
- 230000010287 polarization Effects 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 28
- 239000011159 matrix material Substances 0.000 claims abstract description 20
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 8
- 230000001427 coherent effect Effects 0.000 claims abstract description 6
- 238000004364 calculation method Methods 0.000 claims description 34
- 238000012545 processing Methods 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 15
- 230000000694 effects Effects 0.000 abstract description 19
- 230000014759 maintenance of location Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000011158 quantitative evaluation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention relates to a polarized SAR image processing method and a device, wherein the method carries out the polarized coherent matrix acquired from the polarized SAR image to be processedDecomposing to obtain corresponding characteristic parameters, calculating the characteristic weight of each characteristic parameter by adopting ReliefF algorithm to obtain the optimal characteristic parameter, and adopting MEAN SHIFT algorithm to realize speckle filtering based on the optimal characteristic parameter and the corresponding spatial domain information. Therefore, the invention fully considers the abundant feature information characteristics of the polarized SAR image and carries out the polarization coherence matrix of the polarized SAR image to be processedThe characteristic parameters obtained by decomposition can describe polarized SAR image information from different angles, accuracy and comprehensiveness of the obtained characteristic parameters are guaranteed, the optimal characteristic parameters and corresponding spatial domain information are obtained by a ReliefF algorithm while the MEAN SHIFT algorithm is utilized to obtain the optimal characteristic parameters, and the filtering effect of the speckle filtering is improved.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a device for processing polarized SAR images.
Background
Because of the inherent characteristics of the coherent imaging system, polarized SAR (SYNTHETIC APERTURE RADAR ) data is inevitably affected by speckle noise, so that speckle filtering is one of the main problems of data processing in polarized SAR images.
The early speckle filtering method for the polarized SAR is mainly a denoising method based on optical remote sensing, non-stationarity signals in data are corrected by improving and popularizing the method into an image of the polarized SAR, along with the development of an information technology of image processing, in order to filter out speckle noise and simultaneously keep image information, research on the filtering of the polarized SAR image is mainly focused on selection of pixels with similar scattering characteristics, for example, improved Lee filtering selects the most uniform window from eight edge aligned windows, intensity driving self-adaptive neighborhood filtering uses a region growing technology to generate self-adaptive neighborhood, in a Sigma filter, homogeneous pixels are selected in the improved Sigma range, the traditional filtering methods are used for realizing the speckle filtering by screening homogeneous pixels and then estimating central pixels by utilizing a minimum mean square error criterion, and the filtering effect of the speckle filtering is more dependent on the accuracy of the selection of the homogeneous pixels, but the filtering effect of the speckle filtering is to be optimized due to the fact that mixed pixels and complex ground object information exist in the polarized SAR image.
The main difference between the mean shift filtering and the traditional filtering method is that the mean shift algorithm uses a moving rather than fixed window to select homogeneous pixels, and takes spatial information and spectral information into consideration in two main speckle filtering steps, while the filtering effect is superior to that of the traditional filtering, the polarized SAR image contains abundant ground object information, and the speckle filtering is only carried out based on the two types of information, so that the filtering effect still needs to be improved.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the invention provides a method and a device for processing a polarized SAR image, which improve the filtering effect of speckle filtering of the polarized SAR image.
In order to solve the technical problems, the invention adopts the following technical scheme:
in a first aspect, the present invention provides a method for processing polarized SAR images, comprising:
Acquiring polarization coherence matrixes of the SAR images to be processed, and carrying out the processing on each polarization coherence matrix Decomposing to obtain characteristic parameters corresponding to each polarization coherence matrix;
calculating the feature weight of each feature parameter by adopting ReliefF algorithm, and obtaining the optimal feature parameter according to the feature weight;
and realizing speckle filtering by adopting MEAN SHIFT algorithm based on the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters.
The invention has the beneficial effects that: fully considers the abundant feature information characteristics of the polarized SAR image, so that the polarization coherence matrix of the polarized SAR image to be processed is carried outThe method comprises the steps of decomposing, namely enabling characteristic parameters obtained through decomposition to describe polarized SAR image information from different angles, enabling a polarized coherent matrix to be unchanged in rotation and have clearer mathematical derivation, therefore ensuring accuracy and comprehensiveness of the obtained characteristic parameters, calculating characteristic weights of each characteristic parameter through ReliefF algorithm to obtain optimal characteristic parameters before speckle filtering through MEAN SHIFT algorithm, comprehensively enabling space domain information corresponding to the optimal characteristic parameters to achieve speckle filtering through MEAN SHIFT algorithm, and taking abundant ground feature information of the polarized SAR image into consideration while utilizing MEAN SHIFT algorithm to use moving window characteristics, and improving filtering effect of the speckle filtering.
Optionally, the calculating the feature weight of each feature parameter by adopting ReliefF algorithm, and obtaining the optimal feature parameter according to the feature weight includes:
Randomly selecting a training set with a preset proportion from the polarized SAR image to be processed, and randomly selecting a first sample from the training set;
K first adjacent samples are selected from other samples with the same sample type as the first sample, k second adjacent samples are selected from samples with different sample types as the first sample, and the characteristic weight of each characteristic parameter is updated and calculated according to the k first adjacent samples, the k second adjacent samples and a characteristic weight formula until an updating frequency threshold is reached, so that the optimal characteristic parameter is obtained, wherein the characteristic weight formula is as follows:
(A)=(A)+
Wherein, (A) Representing the feature weight of the i-th update calculation feature parameter a, P (C) representing the probability that the sample type is class C, class (R) representing the type of the first sample R,) Representing the probability of the sample type of the first sample R, M j representing the jth second neighboring sample in sample type C being different from the sample type of the first sample R, H j representing the jth first neighboring sample in the other sample types being the same as the sample type of the first sample R,The difference between the first sample R and the other samples Q on the characteristic parameter a is represented, R represents the first sample, Q represents the other samples, m represents the threshold value of the update times, R (a) represents the value of the first sample R on the characteristic parameter a, and Q (a) represents the value of the other samples Q on the characteristic parameter a.
According to the description, the Relief algorithm calculates and measures the feature weight of each feature parameter, so as to judge the correlation between each feature parameter and the sample type, namely the distinguishing capability of different feature parameters, and ensure the accuracy and rationality of the screened optimal feature parameters.
Optionally, the implementing the speckle filtering based on the spatial domain information corresponding to the optimal characteristic parameter by adopting MEAN SHIFT algorithm includes:
Inputting the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters into a moving point calculation formula to calculate a target moving point, wherein the moving point calculation formula is as follows:
=-x
Wherein, The method comprises the steps that a target moving point which moves next time is represented, x r represents an optimal characteristic parameter, x s represents spatial domain information corresponding to the optimal characteristic parameter, h s represents a first broadband of a kernel function of a MEAN SHIFT algorithm, h r represents a second broadband of the kernel function of the MEAN SHIFT algorithm, x represents a joint domain vector, x i represents a random vector which is independently and uniformly distributed in d dimensions, x i,s represents spatial domain information of an ith pixel in a moving window of the MEAN SHIFT algorithm, x i,r represents characteristic domain information of the ith pixel in the moving window of the MEAN SHIFT algorithm, g s represents a spatial information kernel function, and g r represents a characteristic information kernel function;
Inputting the target moving point into a moving offset formula to calculate a moving offset, wherein the moving offset formula is as follows:
-
Wherein, The amount of shift offset is indicated and,A target movement point representing the current movement; if the moving offset exceeds the mean shift threshold, calculating a target moving point of the next movement by adopting a moving point calculation formula, and calculating the moving offset based on the target moving point of the next movement calculated each time until the calculated moving offset is lower than the mean shift threshold, so as to realize the speckle filtering.
According to the description, the first broadband of the kernel function and the second broadband of the kernel function in the moving point calculation formula have a smoothing effect, can influence the smoothness degree of the algorithm on noise of the speckle filtering, and have a good denoising effect and good information retention, so that the filtering effect of the speckle filtering is improved.
In a second aspect, the present invention provides an apparatus for polarized SAR image processing, comprising:
The decomposition module is used for acquiring polarization coherence matrixes of the SAR images to be processed and carrying out the processing on each polarization coherence matrix Decomposing to obtain characteristic parameters corresponding to each polarization coherence matrix;
the feature weight calculation module is used for calculating the feature weight of each feature parameter by adopting ReliefF algorithm and obtaining the optimal feature parameter according to the feature weight;
And the speckle filtering module is used for realizing speckle filtering by adopting MEAN SHIFT algorithm based on the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters.
The invention has the beneficial effects that: fully considers the abundant feature information characteristics of the polarized SAR image, so that the polarization coherence matrix of the polarized SAR image to be processed is carried outThe method comprises the steps of decomposing, namely enabling characteristic parameters obtained through decomposition to describe polarized SAR image information from different angles, enabling a polarized coherent matrix to be unchanged in rotation and have clearer mathematical derivation, therefore ensuring accuracy and comprehensiveness of the obtained characteristic parameters, calculating characteristic weights of each characteristic parameter through ReliefF algorithm to obtain optimal characteristic parameters before speckle filtering through MEAN SHIFT algorithm, comprehensively enabling space domain information corresponding to the optimal characteristic parameters to achieve speckle filtering through MEAN SHIFT algorithm, and taking abundant ground feature information of the polarized SAR image into consideration while utilizing MEAN SHIFT algorithm to use moving window characteristics, and improving filtering effect of the speckle filtering.
Optionally, the feature weight calculation module specifically includes:
Randomly selecting a training set with a preset proportion from the polarized SAR image to be processed, and randomly selecting a first sample from the training set;
K first adjacent samples are selected from other samples with the same sample type as the first sample, k second adjacent samples are selected from samples with different sample types as the first sample, and the characteristic weight of each characteristic parameter is updated and calculated according to the k first adjacent samples, the k second adjacent samples and a characteristic weight formula until an updating frequency threshold is reached, so that the optimal characteristic parameter is obtained, wherein the characteristic weight formula is as follows:
(A)=(A)+
Wherein, (A) Representing the feature weight of the i-th update calculation feature parameter a, P (C) representing the probability that the sample type is class C, class (R) representing the type of the first sample R,) Representing the probability of the sample type of the first sample R, M j representing the jth second neighboring sample in sample type C being different from the sample type of the first sample R, H j representing the jth first neighboring sample in the other sample types being the same as the sample type of the first sample R,The difference between the first sample R and the other samples Q on the characteristic parameter a is represented, R represents the first sample, Q represents the other samples, m represents the threshold value of the update times, R (a) represents the value of the first sample R on the characteristic parameter a, and Q (a) represents the value of the other samples Q on the characteristic parameter a.
According to the description, the Relief algorithm calculates and measures the feature weight of each feature parameter, so as to judge the correlation between each feature parameter and the sample type, namely the distinguishing capability of different feature parameters, and ensure the accuracy and rationality of the screened optimal feature parameters.
Optionally, the speckle filtering module specifically includes:
Inputting the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters into a moving point calculation formula to calculate a target moving point, wherein the moving point calculation formula is as follows:
=-x
Wherein, The method comprises the steps that a target moving point which moves next time is represented, x r represents an optimal characteristic parameter, x s represents spatial domain information corresponding to the optimal characteristic parameter, h s represents a first broadband of a kernel function of a MEAN SHIFT algorithm, h r represents a second broadband of the kernel function of the MEAN SHIFT algorithm, x represents a joint domain vector, x i represents a random vector which is independently and uniformly distributed in d dimensions, x i,s represents spatial domain information of an ith pixel in a moving window of the MEAN SHIFT algorithm, x i,r represents characteristic domain information of the ith pixel in the moving window of the MEAN SHIFT algorithm, g s represents a spatial information kernel function, and g r represents a characteristic information kernel function;
Inputting the target moving point into a moving offset formula to calculate a moving offset, wherein the moving offset formula is as follows:
-
Wherein, The amount of shift offset is indicated and,A target movement point representing the current movement;
If the moving offset exceeds the mean shift threshold, calculating a target moving point of the next movement by adopting a moving point calculation formula, and calculating the moving offset based on the target moving point of the next movement calculated each time until the calculated moving offset is lower than the mean shift threshold, so as to realize the speckle filtering.
According to the description, the first broadband of the kernel function and the second broadband of the kernel function in the moving point calculation formula have a smoothing effect, can influence the smoothness degree of the algorithm on noise of the speckle filtering, and have a good denoising effect and good information retention, so that the filtering effect of the speckle filtering is improved.
Drawings
Fig. 1 is a flowchart of a method for processing polarized SAR images according to an embodiment of the present invention;
fig. 2 is an overall flow diagram of a method for processing polarized SAR images according to an embodiment of the present invention;
Fig. 3 is a schematic view of polarization information retention of a polarized SAR image according to an embodiment of the present invention;
fig. 4 is a schematic diagram of denoising effect on a polarized SAR image according to an embodiment of the present invention;
Fig. 5 is a detailed information retention schematic diagram of a polarized SAR image according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a polarized SAR image processing apparatus according to an embodiment of the present invention.
[ Reference numerals description ]
1. A polarized SAR image processing device;
2. a decomposition module;
3. the feature weight calculation module;
4. And a speckle filtering module.
Detailed Description
In order that the above-described aspects may be better understood, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
Example 1
Referring to fig. 1 to 5, the present invention provides a method for processing polarized SAR images, comprising the steps of:
s1, acquiring polarization coherence matrixes of to-be-processed polarized SAR images, and performing each polarization coherence matrix Decomposing to obtain characteristic parameters corresponding to each polarization coherence matrix;
in this embodiment, as shown in fig. 2, a conventional polarization coherence matrix extraction method is adopted to obtain polarization coherence matrices from a to-be-processed polarization SAR image, and each polarization coherence matrix is performed Decomposing to obtain corresponding characteristic parameters, namely obtaining polarization scattering entropy H and scattering angleCharacteristic parameters such as polarization anisotropy A.
S2, calculating the characteristic weight of each characteristic parameter by adopting ReliefF algorithm, and obtaining the optimal characteristic parameter according to the characteristic weight;
At this time, the calculating the feature weight of each feature parameter by using the ReliefF algorithm in step S2, and obtaining the optimal feature parameter according to the feature weight includes:
s21, randomly selecting a training set with a preset proportion from the polarized SAR image to be processed, and randomly selecting a first sample from the training set;
in this embodiment, as shown in fig. 2, a training set with a preset proportion is randomly selected from the polarized SAR image to be processed, where the preset proportion is 20%, that is, 20% of the training set is selected, and a first sample is randomly selected from the training set, where the preset proportion can be adjusted and set according to the actual situation.
S22, selecting k first adjacent samples from other samples with the same sample type as the first samples, selecting k second adjacent samples from samples with different sample types as the first samples, and updating and calculating the characteristic weight of each characteristic parameter according to the k first adjacent samples, the k second adjacent samples and a characteristic weight formula until reaching an updating frequency threshold value to obtain the optimal characteristic parameter, wherein the characteristic weight formula is as follows:
(A)=(A)+
Wherein, (A) Representing the feature weight of the i-th update calculation feature parameter a, P (C) representing the probability that the sample type is class C, class (R) representing the type of the first sample R,) Representing the probability of the sample type of the first sample R, M j representing the jth second neighboring sample in sample type C being different from the sample type of the first sample R, H j representing the jth first neighboring sample in the other sample types being the same as the sample type of the first sample R,The difference between the first sample R and the other samples Q on the characteristic parameter a is represented, R represents the first sample, Q represents the other samples, m represents the threshold value of the update times, R (a) represents the value of the first sample R on the characteristic parameter a, and Q (a) represents the value of the other samples Q on the characteristic parameter a.
In this embodiment, as shown in fig. 2, k first neighboring samples are selected from other sample types with the same sample type as the first sample R selected in step S21, where the first neighboring samples refer to neighboring samples with the same sample type and closest to each other, and k second neighboring samples are selected from sample types C with different sample types as the first sample R, where the second neighboring samples refer to neighboring samples with different sample types and closest to each other, the feature weight of each feature parameter is updated according to a feature weight formula, and it is known from the feature weight formula that if the distance between the first neighboring sample and the first sample is greater than the distance between the second neighboring sample and the first sample on the feature parameter a, the feature weight of the feature parameter is reduced, and otherwise, the feature weight of the feature parameter is increased.
S3, realizing speckle filtering by adopting MEAN SHIFT algorithm based on the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters.
In this embodiment, as shown in fig. 2, speckle filtering is implemented by adopting MEAN SHIFT algorithm based on the optimal characteristic parameters and the corresponding spatial domain information obtained in step S2, where the spatial domain information refers to pixel position information.
At this time, the implementing the speckle filtering based on the spatial domain information corresponding to the optimal characteristic parameter and the optimal characteristic parameter in step S3 by adopting the MEAN SHIFT algorithm includes:
s31, inputting the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters into a moving point calculation formula to calculate a target moving point, wherein the moving point calculation formula is as follows:
=-x
Wherein, The method comprises the steps that a target moving point which moves next time is represented, x r represents an optimal characteristic parameter, x s represents spatial domain information corresponding to the optimal characteristic parameter, h s represents a first broadband of a kernel function of a MEAN SHIFT algorithm, h r represents a second broadband of the kernel function of the MEAN SHIFT algorithm, x represents a joint domain vector, x i represents a random vector which is independently and uniformly distributed in d dimensions, x i,s represents spatial domain information of an ith pixel in a moving window of the MEAN SHIFT algorithm, x i,r represents characteristic domain information of the ith pixel in the moving window of the MEAN SHIFT algorithm, g s represents a spatial information kernel function, and g r represents a characteristic information kernel function;
s32, inputting the target moving point into a moving offset formula to calculate a moving offset, wherein the moving offset formula is as follows:
-
Wherein, The amount of shift offset is indicated and,A target movement point representing the current movement; and S33, if the moving offset exceeds the mean value translation threshold, calculating a target moving point of the next movement by adopting a moving point calculation formula, and calculating the moving offset based on the target moving point of the next movement calculated each time until the calculated moving offset is lower than the mean value translation threshold, so as to realize speckle filtering.
In this embodiment, as shown in fig. 2, iterative loop calculation is performed by using the characteristic that MEAN SHIFT algorithm always moves in the direction of increasing probability density to the maximum, i.e. calculating convergence point, and the target movement point and the movement offset of the next movement of each time are calculated by iterative loop calculation, and the corresponding target movement point when the movement offset is lower than the mean shift threshold is used as convergence point, i.e. stopping iteration, so as to implement speckle filtering, where the target movement point at the time of the first iteration is=x。
In a specific embodiment, as shown in fig. 3 to 5, by selecting three 40 x 40 pixels to calculate an average value as a reference image to replace noise-free data for quantitative evaluation, analyzing the filtering effect obtained by performing speckle filtering on the polarized SAR image by the method provided by the embodiment of the present invention, respectively calculating the absolute value of the relative deviation of each scattering category, and simultaneously using the median value as an index for evaluating the information retention of the polarized SAR image to reduce the deviation possibly caused by the average value, the smaller the value of the anisotropy a indicates the smaller the deviation of the filtering effect from the reference image, the better the information retention, and the experiment uses the edge retention average ratio, i.e., the edge index EPD-ROA to evaluate the detail information retention, the edge index EPD-ROA is the ratio of the filtering data to the original data, so that the result is closer to 1, the smaller the deviation, the greater the equivalent view ENL, the better the denoising effect, wherein Lee in fig. 3 to 5 represents a Lee filter, sigma represents a Sigma filter, IDAN represents an IDAN filter, MS represents a conventional MEAN SHIFT algorithm, ours represents a method of speckle filtering adopted in the present embodiment, in fig. 3, the anisotropy a value corresponding to Ours is the lowest compared with other filters and a conventional MEAN SHIFT algorithm, that is, the deviation is the smallest, the polarization information retention is the best, the equivalent view ENL corresponding to Ours in fig. 4 is the largest compared with other filters and a conventional MEAN SHIFT algorithm, that is, the denoising effect is the best, and the edge index EPD-ROA corresponding to Ours in fig. 5 is the closest to 1 compared with other filters and a conventional MEAN SHIFT algorithm, that is, the detail information retention is the best.
Example two
Referring to fig. 6, the present invention provides a polarized SAR image processing apparatus 1, which includes a decomposition module 2, a feature weight calculation module 3, and a speckle filtering module 4.
The decomposition module 2 is configured to acquire polarization coherence matrixes of the to-be-processed polarized SAR images, and perform each polarization coherence matrixDecomposing to obtain characteristic parameters corresponding to each polarization coherence matrix;
The feature weight calculation module 3 is used for calculating the feature weight of each feature parameter by adopting ReliefF algorithm and obtaining the optimal feature parameter according to the feature weight;
And the speckle filtering module 4 is used for realizing speckle filtering by adopting MEAN SHIFT algorithm based on the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters.
Specifically, the feature weight calculation module 3 specifically includes:
Randomly selecting a training set with a preset proportion from the polarized SAR image to be processed, and randomly selecting a first sample from the training set;
K first adjacent samples are selected from other samples with the same sample type as the first sample, k second adjacent samples are selected from samples with different sample types as the first sample, and the characteristic weight of each characteristic parameter is updated and calculated according to the k first adjacent samples, the k second adjacent samples and a characteristic weight formula until an updating frequency threshold is reached, so that the optimal characteristic parameter is obtained, wherein the characteristic weight formula is as follows:
(A)=(A)+
Wherein, (A) Representing the feature weight of the i-th update calculation feature parameter a, P (C) representing the probability that the sample type is class C, class (R) representing the type of the first sample R,) Representing the probability of the sample type of the first sample R, M j representing the jth second neighboring sample in sample type C being different from the sample type of the first sample R, H j representing the jth first neighboring sample in the other sample types being the same as the sample type of the first sample R,The difference between the first sample R and the other samples Q on the characteristic parameter a is represented, R represents the first sample, Q represents the other samples, m represents the threshold value of the update times, R (a) represents the value of the first sample R on the characteristic parameter a, and Q (a) represents the value of the other samples Q on the characteristic parameter a.
Specifically, the speckle filtering module 4 specifically comprises:
Inputting the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters into a moving point calculation formula to calculate a target moving point, wherein the moving point calculation formula is as follows:
=-x
Wherein, The method comprises the steps that a target moving point which moves next time is represented, x r represents an optimal characteristic parameter, x s represents spatial domain information corresponding to the optimal characteristic parameter, h s represents a first broadband of a kernel function of a MEAN SHIFT algorithm, h r represents a second broadband of the kernel function of the MEAN SHIFT algorithm, x represents a joint domain vector, x i represents a random vector which is independently and uniformly distributed in d dimensions, x i,s represents spatial domain information of an ith pixel in a moving window of the MEAN SHIFT algorithm, x i,r represents characteristic domain information of the ith pixel in the moving window of the MEAN SHIFT algorithm, g s represents a spatial information kernel function, and g r represents a characteristic information kernel function;
Inputting the target moving point into a moving offset formula to calculate a moving offset, wherein the moving offset formula is as follows:
-
Wherein, The amount of shift offset is indicated and,A target movement point representing the current movement;
If the moving offset exceeds the mean shift threshold, calculating a target moving point of the next movement by adopting a moving point calculation formula, and calculating the moving offset based on the target moving point of the next movement calculated each time until the calculated moving offset is lower than the mean shift threshold, so as to realize the speckle filtering.
Since the system/device described in the foregoing embodiments of the present invention is a system/device used for implementing the method of the foregoing embodiments of the present invention, those skilled in the art will be able to understand the specific structure and modification of the system/device based on the method of the foregoing embodiments of the present invention, and thus will not be described in detail herein. All systems/devices used in the methods of the above embodiments of the present invention are within the scope of the present invention.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions.
It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the terms first, second, third, etc. are for convenience of description only and do not denote any order. These terms may be understood as part of the component name.
Furthermore, it should be noted that in the description of the present specification, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with the embodiment or example being included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, the different embodiments or examples described in this specification and the features of the different embodiments or examples may be combined and combined by those skilled in the art without contradiction.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art upon learning the basic inventive concepts. Therefore, the appended claims should be construed to include preferred embodiments and all such variations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, the present invention should also include such modifications and variations provided that they come within the scope of the following claims and their equivalents.
Claims (2)
1. A method of polarized SAR image processing, comprising:
Acquiring polarization coherence matrixes of the SAR images to be processed, and carrying out the processing on each polarization coherence matrix Decomposing to obtain characteristic parameters corresponding to each polarization coherence matrix;
calculating the feature weight of each feature parameter by adopting ReliefF algorithm, and obtaining the optimal feature parameter according to the feature weight;
based on the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters, adopting MEAN SHIFT algorithm to realize speckle filtering;
the characteristic parameters comprise H, a scattering angle alpha and A, wherein H represents polarization scattering entropy and A represents polarization anisotropy;
Calculating the feature weight of each feature parameter by adopting ReliefF algorithm, and obtaining the optimal feature parameter according to the feature weight comprises the following steps:
Randomly selecting a training set with a preset proportion from the polarized SAR image to be processed, and randomly selecting a first sample from the training set;
K first adjacent samples are selected from other samples with the same sample type as the first sample, k second adjacent samples are selected from samples with different sample types as the first sample, and the characteristic weight of each characteristic parameter is updated and calculated according to the k first adjacent samples, the k second adjacent samples and a characteristic weight formula until an updating frequency threshold is reached, so that the optimal characteristic parameter is obtained, wherein the characteristic weight formula is as follows:
(A)=(A)+
Wherein, (A) Representing the feature weight of the ith updated computed feature parameter a, P (C) representing the probability that the sample type is class C, class (R) representing the type of the first sample R, P (class (R)) representing the probability of the sample type of the first sample R, M j representing the jth second neighboring sample in sample type C that is different from the sample type of the first sample R, H j representing the jth first neighboring sample in other sample types that are the same as the sample type of the first sample R,Representing the difference between the first sample R and the other samples Q on the characteristic parameter a, R representing the first sample, Q representing the other samples, m representing the threshold number of updates, R (a) representing the value of the first sample R on the characteristic parameter a, Q (a) representing the value of the other samples Q on the characteristic parameter a;
The implementation of the speckle filtering based on the spatial domain information corresponding to the optimal characteristic parameter by adopting MEAN SHIFT algorithm comprises the following steps:
Inputting the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters into a moving point calculation formula to calculate a target moving point, wherein the moving point calculation formula is as follows:
=-x
Wherein, The method comprises the steps that a target moving point which moves next time is represented, x r represents an optimal characteristic parameter, x s represents spatial domain information corresponding to the optimal characteristic parameter, h s represents a first broadband of a kernel function of a MEAN SHIFT algorithm, h r represents a second broadband of the kernel function of the MEAN SHIFT algorithm, x represents a joint domain vector, x i represents a random vector which is independently and uniformly distributed in d dimensions, x i,s represents spatial domain information of an ith pixel in a moving window of the MEAN SHIFT algorithm, x i,r represents characteristic domain information of the ith pixel in the moving window of the MEAN SHIFT algorithm, g s represents a spatial information kernel function, and g r represents a characteristic information kernel function;
Inputting the target moving point into a moving offset formula to calculate a moving offset, wherein the moving offset formula is as follows:
-
Wherein, The amount of shift offset is indicated and,A target movement point representing the current movement; if the moving offset exceeds the mean shift threshold, calculating a target moving point of the next movement by adopting a moving point calculation formula, and calculating the moving offset based on the target moving point of the next movement calculated each time until the calculated moving offset is lower than the mean shift threshold, so as to realize the speckle filtering.
2. An apparatus for polarized SAR image processing, comprising:
The decomposition module is used for acquiring polarization coherence matrixes of the SAR images to be processed and carrying out the processing on each polarization coherence matrix Decomposing to obtain characteristic parameters corresponding to each polarization coherence matrix;
the characteristic parameters comprise H, a scattering angle alpha and A, wherein H represents polarization scattering entropy and A represents polarization anisotropy;
the feature weight calculation module is used for calculating the feature weight of each feature parameter by adopting ReliefF algorithm and obtaining the optimal feature parameter according to the feature weight;
The coherent spot filtering module is used for realizing coherent spot filtering by adopting MEAN SHIFT algorithm based on the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters;
The characteristic weight calculation module specifically comprises:
Randomly selecting a training set with a preset proportion from the polarized SAR image to be processed, and randomly selecting a first sample from the training set;
K first adjacent samples are selected from other samples with the same sample type as the first sample, k second adjacent samples are selected from samples with different sample types as the first sample, and the characteristic weight of each characteristic parameter is updated and calculated according to the k first adjacent samples, the k second adjacent samples and a characteristic weight formula until an updating frequency threshold is reached, so that the optimal characteristic parameter is obtained, wherein the characteristic weight formula is as follows:
(A)=(A)+
Wherein, (A) Representing the feature weight of the ith updated computed feature parameter a, P (C) representing the probability that the sample type is class C, class (R) representing the type of the first sample R, P (class (R)) representing the probability of the sample type of the first sample R, M j representing the jth second neighboring sample in sample type C that is different from the sample type of the first sample R, H j representing the jth first neighboring sample in other sample types that are the same as the sample type of the first sample R,Representing the difference between the first sample R and the other samples Q on the characteristic parameter a, R representing the first sample, Q representing the other samples, m representing the threshold number of updates, R (a) representing the value of the first sample R on the characteristic parameter a, Q (a) representing the value of the other samples Q on the characteristic parameter a;
The speckle filtering module specifically comprises:
Inputting the optimal characteristic parameters and the spatial domain information corresponding to the optimal characteristic parameters into a moving point calculation formula to calculate a target moving point, wherein the moving point calculation formula is as follows:
=-x
Wherein, The method comprises the steps that a target moving point which moves next time is represented, x r represents an optimal characteristic parameter, x s represents spatial domain information corresponding to the optimal characteristic parameter, h s represents a first broadband of a kernel function of a MEAN SHIFT algorithm, h r represents a second broadband of the kernel function of the MEAN SHIFT algorithm, x represents a joint domain vector, x i represents a random vector which is independently and uniformly distributed in d dimensions, x i,s represents spatial domain information of an ith pixel in a moving window of the MEAN SHIFT algorithm, x i,r represents characteristic domain information of the ith pixel in the moving window of the MEAN SHIFT algorithm, g s represents a spatial information kernel function, and g r represents a characteristic information kernel function;
Inputting the target moving point into a moving offset formula to calculate a moving offset, wherein the moving offset formula is as follows:
-
Wherein, The amount of shift offset is indicated and,A target movement point representing the current movement; if the moving offset exceeds the mean shift threshold, calculating a target moving point of the next movement by adopting a moving point calculation formula, and calculating the moving offset based on the target moving point of the next movement calculated each time until the calculated moving offset is lower than the mean shift threshold, so as to realize the speckle filtering.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410319412.5A CN117911253B (en) | 2024-03-20 | 2024-03-20 | Polarized SAR image processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410319412.5A CN117911253B (en) | 2024-03-20 | 2024-03-20 | Polarized SAR image processing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117911253A CN117911253A (en) | 2024-04-19 |
| CN117911253B true CN117911253B (en) | 2024-07-02 |
Family
ID=90687628
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410319412.5A Active CN117911253B (en) | 2024-03-20 | 2024-03-20 | Polarized SAR image processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117911253B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109490886A (en) * | 2018-11-28 | 2019-03-19 | 北京工业大学 | A kind of accurate extracting method in polarimetric synthetic aperture radar remote sensing offshore spilled oil region |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105261051B (en) * | 2015-09-25 | 2018-10-02 | 沈阳东软医疗系统有限公司 | A kind of method and device obtaining image mask |
| CN109523576A (en) * | 2018-10-19 | 2019-03-26 | 中国地质大学（武汉） | A kind of Mean Shift filtering method for taking characters of ground object into account |
| CN110109101A (en) * | 2019-04-04 | 2019-08-09 | 电子科技大学 | A kind of compressed sensing three-dimensional S AR imaging method based on adaptive threshold |
| CN110363105B (en) * | 2019-06-25 | 2021-12-10 | 电子科技大学 | Method for inhibiting speckle of fully-polarized SAR image |
-
2024
- 2024-03-20 CN CN202410319412.5A patent/CN117911253B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109490886A (en) * | 2018-11-28 | 2019-03-19 | 北京工业大学 | A kind of accurate extracting method in polarimetric synthetic aperture radar remote sensing offshore spilled oil region |
Non-Patent Citations (1)
| Title |
|---|
| 行晓黎等.《遥感学报》.2017,正文第1-7页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117911253A (en) | 2024-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Baur et al. | MelanoGANs: high resolution skin lesion synthesis with GANs | |
| CN108389188B (en) | Sparse hyperspectral abnormal target detection method | |
| CN102609904B (en) | Bivariate nonlocal average filtering de-noising method for X-ray image | |
| US8908989B2 (en) | Recursive conditional means image denoising | |
| KR101070981B1 (en) | Image Enhancement Method Using Neural Network Model Based on Edge Component Classification | |
| CN113313641A (en) | CT image denoising method with self-adaptive median filtering | |
| Dharejo et al. | A color enhancement scene estimation approach for single image haze removal | |
| CN116012364B (en) | SAR image change detection method and device | |
| Muniyappan et al. | Contrast enhancement of medical images through adaptive genetic algorithm (AGA) over genetic algorithm (GA) and particle swarm optimization (PSO) | |
| US11145032B2 (en) | Image processing apparatus, method and storage medium for reducing color noise and false color | |
| CN110830043B (en) | Image compressed sensing reconstruction method based on mixed weighted total variation and non-local low rank | |
| CN112150386B (en) | SAR image speckle non-local average inhibition method based on contrast mean value | |
| Kervrann et al. | Unsupervised patch-based image regularization and representation | |
| CN113256667A (en) | SAR image ship target detection method based on Otsu and joint distribution | |
| Ozdarici et al. | A comparison of SAR filtering techniques on agricultural area identification | |
| Bhardwaj et al. | A Novel Method for Despeckling of Ultrasound Images Using Cellular Automata-Based Despeckling Filter | |
| Wang et al. | Application of mixture regression for improved polarimetric SAR speckle filtering | |
| Li et al. | A method to improve the accuracy of SAR image change detection by using an image enhancement method | |
| Yang et al. | No-reference quality assessment of noise-distorted images based on frequency mapping | |
| CN106097292B (en) | Space-time neighborhood Gaussian weighted median filtering method for remote sensing image processing | |
| CN107451608B (en) | SAR image non-reference quality evaluation method based on multi-view amplitude statistical characteristics | |
| CN117911253B (en) | Polarized SAR image processing method and device | |
| Sustika et al. | Generative adversarial network with residual dense generator for remote sensing image super resolution | |
| Javed et al. | Multi-denoising based impulse noise removal from images using robust statistical features and genetic programming | |
| CN103886590A (en) | Method for automatic focusing of push-scanning type remote sensing camera based on wavelet packet energy spectrum |
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 |