CN116299305A - Multi-feature SAR oil spill detection method - Google Patents
Multi-feature SAR oil spill detection method Download PDFInfo
- Publication number
- CN116299305A CN116299305A CN202310575912.0A CN202310575912A CN116299305A CN 116299305 A CN116299305 A CN 116299305A CN 202310575912 A CN202310575912 A CN 202310575912A CN 116299305 A CN116299305 A CN 116299305A
- Authority
- CN
- China
- Prior art keywords
- oil
- oil spill
- area
- sar image
- region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000003305 oil spill Substances 0.000 title claims abstract description 127
- 238000001514 detection method Methods 0.000 title claims abstract description 58
- 230000010287 polarization Effects 0.000 claims abstract description 84
- 230000011218 segmentation Effects 0.000 claims abstract description 47
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000000605 extraction Methods 0.000 claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims description 36
- 238000001228 spectrum Methods 0.000 claims description 10
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 239000013598 vector Substances 0.000 claims description 9
- 238000011478 gradient descent method Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 5
- 238000012986 modification Methods 0.000 claims 1
- 230000004048 modification Effects 0.000 claims 1
- 238000012937 correction Methods 0.000 abstract description 14
- 230000007246 mechanism Effects 0.000 abstract description 12
- 238000013016 damping Methods 0.000 abstract description 7
- 239000011800 void material Substances 0.000 abstract description 3
- 239000003921 oil Substances 0.000 description 78
- 230000000694 effects Effects 0.000 description 5
- 239000013535 sea water Substances 0.000 description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 230000000295 complement effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000007667 floating Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 101710173681 Sulfate adenylyltransferase subunit 2 2 Proteins 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 239000012075 bio-oil Substances 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 230000005574 cross-species transmission Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011158 quantitative evaluation Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003746 surface roughness Effects 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
- G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
- G01S13/9004—SAR image acquisition techniques
- G01S13/9005—SAR image acquisition techniques with optical processing of the SAR signals
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A20/00—Water conservation; Efficient water supply; Efficient water use
- Y02A20/20—Controlling water pollution; Waste water treatment
- Y02A20/204—Keeping clear the surface of open water from oil spills
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Electromagnetism (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a multi-feature SAR oil spill detection method, which is characterized in that firstly, a novel method is designed by considering the physical deviation of a scattering mechanism of Bragg scattering
The polarization characteristic combination of the oil film and the oil-like film is effectively distinguished, and the influence of noise is effectively inhibited; on the basis, a detection result of the oil spill region is obtained based on the oil spill region extraction model; at this time, in order to further overcome the phenomena of unsmooth boundary and internal 'void' existing in the result, a contour regularization boundary correction strategy based on rough segmentation guiding is provided, so that a final oil spill detection result is obtained. The invention solves the problem that oil film interference factors such as weak damping analogues and biological oil films existing in complex sea surface environments can cause interference to detection of an oil spilling area, and effectively improves the detection accuracy.
Description
Technical Field
The invention belongs to the technical field of ocean, and particularly relates to a multi-feature SAR oil spill detection method.
Background
With the rapid development of world economy, oil spills such as collision and leakage of oil tankers, cracking of oil pipelines and the like occur endlessly and are difficult to clean, and huge harm is caused to marine environment. Therefore, effective detection of oil spill areas has become a research feature in the fields of geographic information and remote sensing. Synthetic Aperture Radar (SAR) has the advantages of being capable of being used all the time, all weather, wide in coverage range and the like, and is an important data source for developing ocean oil spill detection. In practical application, in order to timely perform oil spill disaster evaluation and diffusion control, accurate detection of an oil spill area in an SAR image is required. Nevertheless, there are many oil-film-like regions on the sea surface that have similar characteristics to spilled oil, such as bio-oil films, low wind speed ocean surfaces, floating ice regions, etc. Therefore, developing oil spill detection in SAR images remains a very challenging task.
For this reason, scholars have conducted extensive research work and have achieved a series of research results. The existing methods can be mainly divided into unipolar and hololarized SAR image detection methods according to the difference of the adopted data sources. The method mainly extracts the oil spill area based on the backward scattering intensity or texture features of the single-polarized SAR image, and has the advantages of mature data application, low calculation complexity and the like. Nevertheless, such methods describe the scattering properties of spilled oil from only a single angle, and it is difficult to effectively eliminate interference in oil-like film regions in the detection results.
In recent years, with the increasing abundance of full-polarization SAR data, polarization SAR-based methods have become the mainstream and development direction of oil spill detection technologies. Compared with a single-polarized SAR method, the full-polarized SAR can provide richer target scattering characteristics, and is beneficial to remarkably improving the accuracy of oil spill detection. Current research is focused mainly on extracting more representative and discriminating polarization features or feature combinations to enhance the separability of spilled oil and oil-like films. In the prior art, the method based on polarization scattering entropy is proposedHAnd various improvements in anisotropy
Polarization characteristic combination->
Great potential in oil spill detection applications. Nevertheless, the +>
The false alarm rate is still higher when the false alarm rate is faced with weak Bragg scattering generated by a biological oil film; on the other hand, oil film interference factors such as weak damping analogues and biological oil films existing in complex sea surface environments still can cause interference to detection of oil spilling areas.
Disclosure of Invention
The invention provides a multi-feature SAR oil spill detection method, which solves the problem that oil film interference factors such as weak damping analogues and biological oil films existing in complex sea surface environments can interfere with detection of an oil spill region.
In order to solve the technical problems, the invention adopts the following technical scheme:
the multi-feature SAR oil spill detection method comprises the following steps of, aiming at a target sea area sea surface SAR image, obtaining an oil spill region in the target sea area sea surface SAR image:
step 1: performing Cloude decomposition on a polarization coherence matrix corresponding to the sea surface SAR image of the target sea area to obtain a characteristic value of the polarization coherence matrix;
step 2: based on the eigenvalue of the polarization coherence matrix, combining the constructed polarization characteristic combination to obtain a polarization characteristic combination spectrum of the sea surface SAR image of the target sea area;
step 3: based on a polarization feature combination spectrum of a target sea area sea surface SAR image, a pre-trained sea area SAR image is utilized, the polarization feature combination spectrum of the sea area SAR image is combined as input, the sea area SAR image marked with the oil spilling area is taken as an output oil spilling area extraction model, the target sea area sea surface SAR image marked with the oil spilling area is obtained, and then the oil spilling area in the target sea area sea surface SAR image is obtained.
In the step 1, the eigenvalue of the polarization coherence matrix is obtained by the following formula as a preferred technical scheme of the present invention:
in the formula ,
representing a polarization coherence matrix corresponding to the sea surface SAR image of the target sea area; />
、/>
、/>
Three eigenvalues, & lt, respectively, of the polarization coherence matrix>
；/>
Representing a 3x3 special unitary matrix SU (3), wherein +_>
Representation->
Corresponding feature vector, ">
Representation->
Corresponding feature vectors; />
Representation->
Corresponding feature vectors.
As a preferred embodiment of the present invention, the polarization feature combination constructed in the step 2
The following is shown:
wherein ,
，/>
，
in the formula ,
three eigenvalues representing the polarization coherence matrix are based on ranking the first eigenvalue in a rank from large to small; />
Three eigenvalues representing the polarization coherence matrix are based on ranking the second eigenvalue in a rank from large to small;
three eigenvalues representing the polarization coherence matrix are ordered third based on the order from big to small; />
Indicating an improved degree of anisotropy; />
Indicating the base height.
As a preferable technical scheme of the invention, the oil spill region extraction model adopts an SVM model with a kernel function being a linear kernel function.
As a preferable technical scheme of the invention, the method further comprises a step 4 of correcting the oil spilling region by a contour regularization correction strategy under rough segmentation guidance based on the target sea area sea surface SAR image marked with the oil spilling region obtained in the step 3, and updating the oil spilling region in the target sea area sea surface SAR image:
step 4.1: aiming at the target sea area sea surface SAR image marked with the oil spill area, removing isolated points by adopting a related filtering method, and updating the target sea area sea surface SAR image marked with the oil spill area;
step 4.2: based on the target sea area sea surface SAR image marked with the oil spilling region, an 8-connection method is adopted for the oil spilling region to obtain a connected domain set Y corresponding to the oil spilling region in the image,
，/>
representing a subset of connected domains corresponding to the nth oil spill region, wherein N represents the total number of the oil spill regions;
step 4.3: based on the connected domain set Y corresponding to the oil region, aiming at each connected domain subset respectively
Constructing external rectangles corresponding to the connected domain subsets respectively, and obtaining external rectangular areas corresponding to the connected domain subsets respectively;
step 4.4: and respectively aiming at the circumscribed rectangular areas corresponding to the sub-sets of each connected domain, taking the minimum of the segmentation energy function as a target, combining a standard gradient descent method to iterate and segment the circumscribed rectangular areas to obtain an optimal oil spill segmentation contour, taking the oil spill area contained in the optimal oil spill segmentation contour as an updated oil spill area, and updating the oil spill area in the sea surface SAR image of the target sea area.
As a preferred technical solution of the present invention, in step 4.4, the following steps are executed for the circumscribed rectangular areas corresponding to the respective connected domain subsets, respectively:
step 4.4.1: constructing an image pixel intensity probability distribution model corresponding to a circumscribed rectangular area aiming at the circumscribed rectangular area corresponding to the connected domain subset;
step 4.4.2: based on the image pixel intensity probability distribution model, the following segmentation energy function is adopted
：
in the formula ,
when->
Indicates the oil spill area, when->
Indicating a non-spilled oil region; />
and />
All represent preset positive balance parameters; />
Representation->
The regional image obeys radar cross-section components obtained by probability distribution; />
Representing pixel points; />
Expressed in pixels->
In the preset vicinity as the center +.>
Pixel points of the region; />
Representing a level set function; />
Representing an energy function represented by a level set function; />
Representing the length of the zero level set profile; />
Representation regularization;
step 4.4.3: based on the segmentation energy function, taking the minimization of the segmentation energy function as a target, and adopting a standard gradient descent method to update alternately and iteratively
and />
Obtaining the corresponding optimal +.>
Thereby obtaining the best->
The corresponding optimal oil spill segmentation profile; and taking the oil spill region contained in the optimal oil spill segmentation contour as the updated oil spill region, and updating the oil spill region in the sea surface SAR image of the target sea area.
The beneficial effects of the invention are as follows: the invention provides a multi-feature SAR oil spill detection method, which is characterized in that firstly, a novel method is designed by considering the physical deviation of a scattering mechanism of Bragg scattering
The polarization characteristic combination of the oil film and the oil-like film is effectively distinguished, and the influence of noise is effectively inhibited; on the basis, a detection result of the oil spill region is obtained based on the oil spill region extraction model; at this time, in order to further overcome the phenomena of unsmooth boundary and internal 'void' existing in the result, a contour regularization boundary correction strategy based on rough segmentation guiding is provided, so that a final oil spill detection result is obtained. The invention solves the problem that oil film interference factors such as weak damping analogues and biological oil films existing in complex sea surface environments can cause interference to detection of an oil spilling area, and effectively improves the detection accuracy.
Drawings
FIG. 1 is a flowchart of a multi-feature SAR oil spill detection method in an embodiment of the present disclosure;
fig. 2 is an SAR oil spill image provided in an embodiment of the present invention;
FIG. 3 is a real image of oil spill produced by visual interpretation for SAR oil spill image in an embodiment of the present invention;
FIG. 4 is a comparison chart of oil spill detection results based on the corresponding polarization characteristics in the embodiment of the invention;
FIG. 5 is a linear comparison graph of MA indicators based on each polarization feature in an embodiment of the invention;
fig. 6 is a graph showing CRRS effectiveness analysis in the embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples will provide those skilled in the art with a more complete understanding of the invention, but are not intended to limit the invention in any way.
As shown in fig. 1, in the multi-feature SAR oil spill detection method, for a target sea area sea surface SAR image, the following steps are performed to obtain an oil spill region in the target sea area sea surface SAR image.
Step 1: and performing Cloude decomposition on a polarization coherence matrix corresponding to the sea surface SAR image of the target sea area to obtain a characteristic value of the polarization coherence matrix.
In the step 1, the eigenvalue of the polarization coherence matrix is obtained by the following formula:
in the formula ,
representing a polarization coherence matrix corresponding to the sea surface SAR image of the target sea area; />
、/>
、/>
Three eigenvalues, & lt, respectively, of the polarization coherence matrix>
；/>
Representing a 3x3 special unitary matrix SU (3), wherein +_>
Representation->
Corresponding feature vector, ">
Representation->
Corresponding feature vectors; />
Representation->
Corresponding feature vectors. -1 represents matrix inversion, ++>
、/>
、/>
The three characteristic values respectively represent a main scattering mechanism, a secondary scattering mechanism and a least-scattering mechanism from large to small.
In addition, interference of radar waves may generate coherent speckle noise on the SAR image, and the speckle may cause difficulty in accurately estimating characteristics of the backscatter body, which may easily cause erroneous detection. Therefore, before Cloude decomposition is performed, the embodiment performs data preprocessing on the sea surface SAR image of the target sea area, and firstly performs denoising processing on the original SAR image by adopting a Refind-Lee filtering method, and parameters are automatically set according to the suggestion of software. In addition, the radar SAT-2 image data is adopted in the experiment in the embodiment, the condition of uneven SAR image pixel value distribution can occur due to the distance between an imaging area and a satellite platform, and the geometric distortion of partial features can be generated due to the change of topography, so that the interpretation and judgment of images are not facilitated, and the requirements of sea surface oil spill recognition can not be met. For this purpose, the present embodiment uses geocoding and radiometric calibration methods to quantitatively describe the relationship between radar backscatter intensity and target backscatter cross-section to eliminate gray value differences and perform geometric corrections.
Step 2: based on the eigenvalue of the polarization coherence matrix, combining the constructed polarization characteristic combination to obtain the polarization characteristic combination spectrum of the sea surface SAR image of the target sea area.
Construction for computation
The polarization features combine the various features required, first performing a Cloude decomposition on the SAR image. The Cloude decomposition theory is based on covariance matrix and phaseThe idea behind the development of the dry matrix is to decompose the target correlation matrix into three weighted sums of mutually orthogonal correlation matrices, which represent three mutually independent and uncorrelated scattering processes: unidirectional scattering, bidirectional scattering, and cross scattering. This decomposition has been demonstrated to better describe the variability and randomness exhibited by oil spill scattering. For this purpose, the eigenvalues of the polarization coherence matrix in the output result of the Cloude decomposition are used as basis for the present text and improved features and feature combinations are further built up on this basis>
。
The polarization characteristic combination constructed in the step 2
The following is shown:
wherein ,
，/>
，
in the formula ,
three eigenvalues representing the polarization coherence matrix are based on ranking the first eigenvalue in a rank from large to small; />
Three eigenvalues representing the polarization coherence matrix are based on ranking the second eigenvalue in a rank from large to small;
three eigenvalues representing the polarization coherence matrix are ordered third based on the order from big to small; />
Indicating an improved degree of anisotropy; />
Indicating the base height.
On the basis of eigenvalue decomposition of a polarization coherence matrix, the embodiment firstly extracts an improved anisotropy degree
As one of the basic units of the proposed feature combination, the reason is that: bragg scattering dominates due to the large value of sea surface anisotropy. Therefore, the oil spill region is +.>
The value is lower. Improved anisotropy->
The two largest eigenvalues are used to calculate, unlike the traditional anisotropy, to prevent the second and third eigenvalues from being severely affected by noise, which helps to distinguish sea surface roughness. In addition, the base height is selected here>
This is because the base height can be very well distinguished from the oil-covered surface when handling the weak damping oil-bearing surface. In the case of a non-floating sea surface and a weakly damped floating sea surface->
The predicted values are very low, whereas on petroleum covered sea surfaces +.>
The expected value is much larger. Thus, it can be seen that +.>
Reflects the main scattering mechanism and the sub-scattering mechanism, whereas the most sub-scattering mechanism is omitted and +.>
The lowest scattering mechanism can be represented by the ratio of the minimum eigenvalue to the maximum eigenvalue, increasing the contrast of seawater and spilled oil in value. The two are therefore very complementary, which is why the two features are chosen here as basic units to construct a feature combination. On the basis of this, an improved anisotropy is combined>
And base height +>
A new set of polarization feature combinations is designed>
。
The polarization characteristic combination is formed by four factors together into a polarization characteristic combination spectrum, wherein,
the random scattering process is characterized, the numerical difference between spilled oil and sea water is enhanced due to the effect of factor multiplication, and further false alarm clutter information with low randomness such as background sea water, oil film with weak damping effect and the like can be effectively restrained, but larger noise is introduced, and the boundary is fuzzy; to further consider the separability of oil and water, we introduce +.>
and />
The oil-water difference is further enlarged, so that the boundary of an oil film is clearer and separable, and the oil-water contrast ratio is improved; to further reduce the false alarm rate of oil spill detection, we introduced +.>
The oil-water contrast ratio is enlarged. Therefore, the four formulas of the feature combination are helpful for distinguishing the spilled oil and the seawater from different angles, so that an organic complementary whole is formed, and the spilled oil identification precision is remarkably improved.
Step 3: based on a polarization feature combination spectrum of a target sea area sea surface SAR image, a pre-trained sea area SAR image is utilized, the polarization feature combination spectrum of the sea area SAR image is combined as input, the sea area SAR image marked with the oil spilling area is taken as an output oil spilling area extraction model, the target sea area sea surface SAR image marked with the oil spilling area is obtained, and then the oil spilling area in the target sea area sea surface SAR image is obtained.
And the oil spilling region is extracted into a model, and an SVM model with a kernel function being a linear kernel function is adopted. In the present embodiment, based on
After the feature space is constructed, the embodiment adopts a support vector machine (Support Vector Machine, SVM) method to roughly extract the oil spilling region. Among them, the SVM is a classification model whose basic model is a linear classifier defined at the maximum interval in a feature space, and has been widely used in various fields in recent years. In addition, the linear kernel function is adopted in the text, and the linear kernel function has the advantages of few parameters, simplicity, easiness in use, high calculation speed, ideal classification effect and the like under the condition of linear separable. And mapping the data into the high-altitude dimension by using a kernel function, and finding the optimal hyperplane, thereby obtaining the extraction result of the oil spill region.
Based on the extraction result of the oil spilling region obtained in the step 3, as the SVM-based extraction is a method of coarsely extracting the pixel level, the boundary of the oil spilling region is not smooth, and the phenomenon of 'hollowness' such as isolated points existing in the region is inevitably generated in the detection result, so that the problem of inconsistent appearance with the actual oil spilling region is caused. For this reason, the present embodiment proposes a contour regularization correction strategy CRRS (Contour regularization under rough segmentation guidelines) under rough segmentation guidance.
Therefore, in this embodiment, the method further includes step 4, based on the target sea area sea surface SAR image marked with the oil spill area obtained in step 3, of correcting the oil spill area through a contour regularization correction strategy under rough segmentation guidance, including steps 4.1-4.4, and updating the oil spill area in the target sea area sea surface SAR image, so as to obtain an oil spill identification map after contour regularization correction.
Step 4.1: aiming at the target sea area sea surface SAR image marked with the oil spill area, removing isolated points by adopting a related filtering method, and updating the target sea area sea surface SAR image marked with the oil spill area; considering that the oil spilling area usually has a certain area, the relevant filtering method is used for eliminating isolated points, namely eliminating pixel points, so that the influence of noise on a detection result is reduced.
Step 4.2: based on the target sea area sea surface SAR image marked with the oil spilling region, an 8-connection method is adopted for the oil spilling region to obtain a connected domain set Y corresponding to the oil spilling region in the image,
，/>
indicating a subset of connected domains corresponding to the nth oil spill region, and N indicating the total number of the oil spill regions.
Step 4.3: based on the connected domain set Y corresponding to the oil region, aiming at each connected domain subset respectively
Constructing external rectangle corresponding to each connected domain subset respectively, and obtaining external rectangle regions corresponding to each connected domain subset respectively;
step 4.4: and respectively aiming at the circumscribed rectangular areas corresponding to the sub-sets of each connected domain, taking the minimum of the segmentation energy function as a target, combining a standard gradient descent method to iterate and segment the circumscribed rectangular areas to obtain an optimal oil spill segmentation contour, taking the oil spill area contained in the optimal oil spill segmentation contour as an updated oil spill area, and updating the oil spill area in the sea surface SAR image of the target sea area. Specifically, the segmentation energy function value obtained in each iteration is compared, and when the segmentation energy function value is larger than the segmentation energy function value obtained in the previous iteration, the oil spill segmentation contour obtained in the previous iteration is the optimal oil spill segmentation contour.
In the step 4.4, the following steps are executed for the circumscribed rectangular areas corresponding to the connected domain subsets respectively:
step 4.4.1: and constructing an image pixel intensity probability distribution model corresponding to the circumscribed rectangular region aiming at the circumscribed rectangular region corresponding to the connected region subset.
Step 4.4.2: based on the image pixel intensity probability distribution model, the following segmentation energy function is adopted
：
in the formula ,
when->
Indicates the oil spill area, when->
Indicating a non-spilled oil region; />
and />
All represent preset positive balance parameters, 1 and 0.00007 are taken in this embodiment respectively; />
Representation->
The regional image obeys radar cross-section components obtained by probability distribution; />
Representing pixel points; />
Expressed in pixels->
In the preset vicinity as the center +.>
Pixel points of the region; />
Representing a level set function; />
Representing an energy function represented by a level set function; />
Representing the length of the zero level set profile; />
Representing regularization.
Step 4.4.3: based on the segmentation energy function, taking the minimization of the segmentation energy function as a target, and adopting a standard gradient descent method to update alternately and iteratively
and />
Obtaining the corresponding optimal +.>
Thereby obtaining the best->
The corresponding optimal oil spill segmentation profile; and taking the oil spill region contained in the optimal oil spill segmentation contour as the updated oil spill region, and updating the oil spill region in the sea surface SAR image of the target sea area.
Because the edge pixel intensity of the SAR image accords with the exponential distribution, a SAR image pixel intensity probability distribution model is constructed based on the original SAR image
In the embodiment, probability distribution obeys exponential distribution, the model is more accurate in pixel intensity description, more image edge detail information is contained, and accurate characterization of pixel intensity characteristics of an oil spill image is facilitated.
To characterize the performance of the segmentation operation, a level set function is used
To quantify the whole image domain
Is provided. Whole image domain
Can be expressed as
，
. Specifically, in the level set method, contours
By a Lipschitz function:
is represented by a zero level set of (2). In this case, a level set function is utilized
To represent the segmentation of the oil leakage image and the exterior and interior of the contour V, level set function
Taking positive and negative values, respectively. The energy function expressed by the level set function is as follows (subscript L means level)
in the formula ,
representing a preset positive equilibrium constant; />
Representing +.>
Pixel point +.>
Kernel function at offset>
；/>
Presetting a scale parameter larger than zero; />
Is pixel dot +.>
Corresponding preset adjacent area->
Middle pixel +.>
The strength of the material to be treated is,
，/>
for detecting system constants, wherein ∈ ->
，/>
，/>
As a Heaviside function, +.>
Indicating->
Smooth positive smooth parameters +.>
。
Subsequently, for calculation of
Is introduced into the length of the zero level set profile, a profile regularization term:
meanwhile, in order to maintain regularization of level set evolution so as to realize accurate calculation and stable evolution, a regularization term is introduced:
finally, the energy function proposed above is represented by a level set energy term to represent the suitability of the oil leakage, a regularization term about the contour of the oil leakage contour, and an updated regularization term for accurate computation and stable level set evolution. These terms are combined to obtain the overall segmentation energy function. The smaller the integral segmentation energy function value is, the initial contour is more close to the actual oil spill contour, otherwise, the segmentation effect is not ideal and is far away from the actual oil spill contour.
For the segmentation energy function, a novel segmentation energy function E is defined by introducing a contour regularization term, and a standard gradient descent method is used for the segmentation energy function
and />
Alternate iterative updating is carried out to push the contour to shrink continuously, and the result when the E minimum value is taken as the highest valueAnd (5) final result. All pixels in the final result closure profile are then marked as spill pixels, thereby obtaining the final result. Because the oil film edge is generally thin and is interlaced with the oil-like film, the pixels at the oil film edge generally do not conform to the polarization characteristics exhibited by the typical oil film region, resulting in the above pixel level detection methods being prone to false detection and false omission. In practice, the oil film boundary usually has a characteristic that the brightness change is obvious, the gray value and surrounding pixels are stepped, and the pixel distribution of the boundary usually meets the exponential distribution. Therefore, an image pixel intensity probability distribution model needs to be introduced, an updated contour regularization term is introduced on the basis of the image pixel intensity probability distribution model to control the motion speed of the movable contour model, so that the purposes of preserving weak edges and removing small isolated areas in the curve evolution process are achieved, and the boundary correction is carried out on the spillover oil detection result based on the polarization feature combination by the correction strategy.
In this embodiment, as shown in fig. 2, an SAR oil spill image provided in the case of this embodiment is shown; on this basis, we have made a true map of oil spills using visual interpretation, as shown in fig. 3. In fig. 3, 4 representative regions are selected and labeled with letters a through D for ease of subsequent detailed comparison.
As shown in FIG. 4, except for that proposed in this embodiment
Besides the polarization characteristic combination, oil spill detection results with several polarization characteristics are selected for comparison experiments: the first is +.>
Is a combination of polarization features of (a); the second is based on new polarization characteristics
The method comprises the steps of carrying out a first treatment on the surface of the Third is based on reduced polarization +.>
Construction of polarization feature combinations. The purpose of selecting these three polarization feature combinations is to: and->
Can explore +.>
An advancing role in oil spill detection; and->
The contrast of the polarization features can embody the advantages of the polarization feature combination relative to the single polarization feature; and->
Comparison of the combination of reduced polarization features allows one to explore the effectiveness of reduced full polarization versus reduced polarization. The comparison results are shown in FIG. 4, wherein (a) in FIG. 4 is a true map of spilled oil and (b) is based on +.>
The oil spill detection result; (c) Is based on->
The oil spill detection result; (d) Is based on->
The oil spill detection result; (e) Is based on->
Is a result of the oil spill detection.
As can be seen from the view of figure 4,
the combined oil spill recognition result is relatively close to the oil spill real image at the oil spill detail; />
The combined oil spill recognition result has the phenomenon of oil spill pixel point omission, and part of oil spill is not recognized; new polarization characteristics->
The oil spilling pixel point at the detail part is not detected, and the part is not identifiedAn oil spill area; reduced polarization->
The serious oil spilling pixel point omission phenomenon exists, and a serious pixel cavity exists at the main oil spilling region, and meanwhile, part of the oil spilling region is not identified.
For further quantitative analysis, in this embodiment, three indexes including an accuracy rate (Acc), a Recall rate (Recall) and a Kappa coefficient are selected for accuracy evaluation, and the values of the three indexes are all between 0 and 1, and the closer the value is to 1, the better the classification performance is represented. As shown in table 1 based on the quantitative accuracy evaluation table of each polarization characteristic,
TABLE 1
As can be seen from the table above,
the Acc, recall, and Kappa coefficients of the polarization feature combination are highest among the four polarization feature-corresponding methods, meaning that the accuracy of spill identification from the polarization feature combination is greater than the other three methods. This is because +.>
Compared with +.>
Introduces->
The contrast ratio of the seawater and the oil film can be further enlarged, and false alarm pixel points are reduced, so that the classification precision is remarkably improved; and new polarization characteristics->
Compared with the prior art, the combination of multiple characteristics forms an organic complementary whole to improve the oil spill recognition precision; compared to the reduced polarization feature combination +.>
The full polarization has higher recognition sensitivity to oil spill, thereby greatly reducing the situation of false leakage detection and leading the oil spill detection result to be closer to the actual oil spill scene. Thus, it can be confirmed by visual interpretation and quantitative analysis that we propose a basis for +.>
The method for combining polarization characteristics is obviously superior to the other 3 methods, and the effectiveness and reliability of the method are proved.
To further demonstrate the effectiveness of the constructed polarization feature combinations, one would
And->
、/>
、/>
、/>
、
And surface scattering fraction->
6 classical polarization features were compared. In order to objectively compare the merits of polarization feature combinations and single polarization features in oil spill recognition, we selected the intra-class pixel average accuracy (MA) for quantitative evaluation, as shown in FIG. 5 +.>
The MA index of (c) is significantly better than other unipolar features, demonstrating the effectiveness of the combination of polar features presented herein in oil-water detection as compared to the unipolar features. Furthermore, only +.>
The accuracy of (a) is better than that of other individual features than the method of the present embodiment, say +.>
The non-Bragg scattering mechanism of the oil spilling region can be reflected in the oil spilling detection field, and the identification force on the oil spilling region is stronger. But at->
Foundation constructed->
Then further introduce->
The method is used for representing the first scattering mechanism and has better detection effect.
To explore the effectiveness of CRRS in boundary correction on SVM rough segmentation results, we selected the 4 representative regions of a through D in fig. 3 for detailed analysis. As shown in fig. 6, (a) represents a column of the oil spill true map corresponding to the 4 block areas a to D, (b) represents a column of the SVM extraction oil spill map corresponding to the 4 block areas a to D, (c) represents a column of the CRRS correction map corresponding to the 4 block areas a to D, the rough segmentation map has a boundary non-smooth problem, and the pixel point of the internal oil spill area is more serious. After regularized boundary correction, the boundary of the oil spilling outline is smoother, the internal oil spilling area is more homogeneous, the pixel point cavity phenomenon is avoided, and the actual situation of the oil spilling outline is more met. Therefore, the regularization correction strategy presented herein is viable and effective.
The invention designs a multi-characteristic SAR oil spill detection method, firstly, a novel method is designed by considering the physical deviation of a scattering mechanism of Bragg scattering
The polarization characteristic combination of the oil film and the oil-like film is effectively distinguished, and the influence of noise is effectively inhibited; based on the overflowThe oil region extraction model obtains the detection result of the oil spilling region; at this time, in order to further overcome the boundary non-smoothness and internal "void" phenomenon existing in the result, a contour regularization boundary correction strategy (CRRS) based on rough segmentation guidance is provided to correct the oil spill region, so as to obtain the final oil spill detection result. The invention solves the problem that oil film interference factors such as weak damping analogues and biological oil films existing in complex sea surface environments can cause interference to detection of an oil spilling area, and effectively improves the detection accuracy.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that the foregoing embodiments may be modified or equivalents substituted for some of the features thereof. All equivalent structures made by the content of the specification and the drawings of the invention are directly or indirectly applied to other related technical fields, and are also within the scope of the invention.
Claims (6)
1. The multi-feature SAR oil spill detection method is characterized in that aiming at a target sea area sea surface SAR image, the following steps are executed to obtain an oil spill area in the target sea area sea surface SAR image:
step 1: performing Cloude decomposition on a polarization coherence matrix corresponding to the sea surface SAR image of the target sea area to obtain a characteristic value of the polarization coherence matrix;
step 2: based on the eigenvalue of the polarization coherence matrix, combining the constructed polarization characteristic combination to obtain a polarization characteristic combination spectrum of the sea surface SAR image of the target sea area;
step 3: based on a polarization feature combination spectrum of a target sea area sea surface SAR image, a pre-trained sea area SAR image is utilized, the polarization feature combination spectrum of the sea area SAR image is combined as input, the sea area SAR image marked with the oil spilling area is taken as an output oil spilling area extraction model, the target sea area sea surface SAR image marked with the oil spilling area is obtained, and then the oil spilling area in the target sea area sea surface SAR image is obtained.
2. The multi-feature SAR oil spill detection method according to claim 1, wherein in step 1, the feature value of the polarization coherence matrix is obtained by the following formula:
in the formula ,
representing a polarization coherence matrix corresponding to the sea surface SAR image of the target sea area; />
、/>
、/>
Three eigenvalues, & lt, respectively, of the polarization coherence matrix>
；/>
Representing a 3x3 special unitary matrix SU (3), wherein +_>
Representation->
Corresponding feature vector, ">
Representation->
Corresponding feature vector；/>
Representation of
Corresponding feature vectors.
3. The multi-feature SAR oil spill detection method according to claim 1, wherein said polarization feature combination constructed in step 2
The following is shown:
wherein ,
，/>
，
in the formula ,
three eigenvalues representing the polarization coherence matrix are based on ranking the first eigenvalue in a rank from large to small; />
Three eigenvalues representing the polarization coherence matrix are based on ranking the second eigenvalue in a rank from large to small;
three eigenvalues representing the polarization coherence matrix are ordered third based on the order from big to small; />
Indicating an improved degree of anisotropy；/>
Indicating the base height.
4. The multi-feature SAR oil spill detection method according to claim 1, wherein the oil spill region extraction model uses an SVM model with a kernel function that is a linear kernel function.
5. The multi-feature SAR oil spill detection method according to claim 1, further comprising the step of 4, based on the target sea area SAR image marked with the oil spill region obtained in the step of 3, updating the oil spill region in the target sea area sea surface SAR image by modifying the oil spill region through a contour regularization modification strategy under rough segmentation guidance, wherein the method comprises the following steps:
step 4.1: aiming at the target sea area sea surface SAR image marked with the oil spill area, removing isolated points by adopting a related filtering method, and updating the target sea area sea surface SAR image marked with the oil spill area;
step 4.2: based on the target sea area sea surface SAR image marked with the oil spilling region, an 8-connection method is adopted for the oil spilling region to obtain a connected domain set Y corresponding to the oil spilling region in the image,
，/>
representing a subset of connected domains corresponding to the nth oil spill region, wherein N represents the total number of the oil spill regions;
step 4.3: based on the connected domain set Y corresponding to the oil region, aiming at each connected domain subset respectively
Constructing external rectangles corresponding to the connected domain subsets respectively, and obtaining external rectangular areas corresponding to the connected domain subsets respectively;
step 4.4: and respectively aiming at the circumscribed rectangular areas corresponding to the sub-sets of each connected domain, taking the minimum of the segmentation energy function as a target, combining a standard gradient descent method to iterate and segment the circumscribed rectangular areas to obtain an optimal oil spill segmentation contour, taking the oil spill area contained in the optimal oil spill segmentation contour as an updated oil spill area, and updating the oil spill area in the sea surface SAR image of the target sea area.
6. The multi-feature SAR oil spill detection method according to claim 5, wherein in step 4.4, the following steps are performed for the circumscribed rectangular areas corresponding to the respective connected domain subsets:
step 4.4.1: constructing an image pixel intensity probability distribution model corresponding to a circumscribed rectangular area aiming at the circumscribed rectangular area corresponding to the connected domain subset;
step 4.4.2: based on the image pixel intensity probability distribution model, the following segmentation energy function is adopted
：
in the formula ,
when->
Indicates the oil spill area, when->
Indicating a non-spilled oil region; />
and />
All represent preset positive balance parameters; />
Representation->
The regional image obeys radar cross-section components obtained by probability distribution; />
Representing pixel points; />
Expressed in pixels->
In the preset vicinity as the center +.>
Pixel points of the region; />
Representing a level set function; />
Representing an energy function represented by a level set function; />
Representing the length of the zero level set profile; />
Representation regularization;
step 4.4.3: based on the segmentation energy function, taking the minimization of the segmentation energy function as a target, and adopting a standard gradient descent method to update alternately and iteratively
and />
Obtaining the corresponding optimal +.>
Thereby obtaining the optimum
The corresponding optimal oil spill segmentation profile; and taking the oil spill region contained in the optimal oil spill segmentation contour as the updated oil spill region, and updating the oil spill region in the sea surface SAR image of the target sea area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310575912.0A CN116299305B (en) | 2023-05-22 | 2023-05-22 | Multi-feature SAR oil spill detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310575912.0A CN116299305B (en) | 2023-05-22 | 2023-05-22 | Multi-feature SAR oil spill detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116299305A true CN116299305A (en) | 2023-06-23 |
| CN116299305B CN116299305B (en) | 2023-08-15 |
Family
ID=86820783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310575912.0A Active CN116299305B (en) | 2023-05-22 | 2023-05-22 | Multi-feature SAR oil spill detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116299305B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116894999A (en) * | 2023-07-18 | 2023-10-17 | 中国石油大学(华东) | Method and device for detecting oil spill polarization SAR based on condition parameterized convolution |
Citations (3)
| 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 |
| KR20190133442A (en) * | 2018-05-23 | 2019-12-03 | 한국해양과학기술원 | Method for detecting oil spills on satellite sar images using artificial neural network |
| CN115861247A (en) * | 2022-12-09 | 2023-03-28 | 山东科技大学 | High-resolution remote sensing image contour multistage regularization method, system and application |
-
2023
- 2023-05-22 CN CN202310575912.0A patent/CN116299305B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20190133442A (en) * | 2018-05-23 | 2019-12-03 | 한국해양과학기술원 | Method for detecting oil spills on satellite sar images using artificial neural network |
| 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 |
| CN115861247A (en) * | 2022-12-09 | 2023-03-28 | 山东科技大学 | High-resolution remote sensing image contour multistage regularization method, system and application |
Non-Patent Citations (4)
| Title |
|---|
| 李冠男: "基于星载极化SAR图像的海上溢油检测与分析研究", 中国博士学位论文全文数据库 工程科技Ⅰ辑（月刊）, no. 04, pages 027 - 66 * |
| 李颖 等: "基于星载SAR的海上溢油检测研究进展", 海洋通报, vol. 36, no. 3, pages 241 - 249 * |
| 甄宗晋: "基于深度学习的极化SAR海上溢油检测方法研究", 中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑（月刊）, no. 02, pages 019 - 69 * |
| 郑洪磊;张彦敏;王运华;: "基于极化特征SERD的SAR溢油检测", 海洋湖沼通报, no. 04 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116894999A (en) * | 2023-07-18 | 2023-10-17 | 中国石油大学(华东) | Method and device for detecting oil spill polarization SAR based on condition parameterized convolution |
| CN116894999B (en) * | 2023-07-18 | 2024-05-03 | 中国石油大学(华东) | Method and device for detecting oil spill polarization SAR based on condition parameterized convolution |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116299305B (en) | 2023-08-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109816641B (en) | Multi-scale morphological fusion-based weighted local entropy infrared small target detection method | |
| CN112395987B (en) | SAR image target detection method based on unsupervised domain adaptive CNN | |
| CN103473764B (en) | A kind of remote sensing image object variations detection method | |
| CN103345757A (en) | Optical image and SAR image automatic registration method within multilevel multi-feature constraint | |
| CN116299305B (en) | Multi-feature SAR oil spill detection method | |
| CN111008664B (en) | Hyperspectral sea ice detection method based on space-spectrum combined characteristics | |
| Shaoqing et al. | The comparative study of three methods of remote sensing image change detection | |
| CN113536963B (en) | SAR image airplane target detection method based on lightweight YOLO network | |
| CN108805057B (en) | SAR image reservoir area detection method based on joint significance analysis | |
| Hou et al. | End-to-end automatic ship detection and recognition in high-resolution Gaofen-3 spaceborne SAR images | |
| Chen et al. | Multi-scale patch-GAN with edge detection for image inpainting | |
| Zhang et al. | Saliency-driven oil tank detection based on multidimensional feature vector clustering for SAR images | |
| KR20180066669A (en) | Apparatus for classifying target of inverse synthetic aperture radar image using trace transform and method thereof | |
| CN116343113A (en) | Method and system for detecting oil spill based on polarized SAR characteristics and coding and decoding network | |
| CN114821358A (en) | Optical remote sensing image marine ship target extraction and identification method | |
| CN106056115B (en) | A kind of infrared small target detection method under non-homogeneous background | |
| CN113222898B (en) | Double-navigation SAR image trace detection method based on multi-element statistics and deep learning | |
| Zhang et al. | A review of recent advance of ship detection in single-channel SAR images | |
| CN116894999B (en) | Method and device for detecting oil spill polarization SAR based on condition parameterized convolution | |
| CN113505712A (en) | Novel loss function based sea surface oil spill detection method of convolutional neural network | |
| Haigang et al. | A novel ship detection method for large-scale optical satellite images based on visual LBP feature and visual attention model | |
| Dey et al. | A robust FLIR target detection employing an auto-convergent pulse coupled neural network | |
| Su et al. | Change detection in synthetic aperture radar images based on non-local means with ratio similarity measurement | |
| CN112686222B (en) | Method and system for detecting ship target by satellite-borne visible light detector | |
| Zhou et al. | Automatic extraction of lakes on the Qinghai-Tibet Plateau from Sentinel-1 SAR images |
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 |