CN110853050B - SAR image river segmentation method, device and medium - Google Patents

SAR image river segmentation method, device and medium Download PDF

Info

Publication number
CN110853050B
CN110853050B CN201910998432.9A CN201910998432A CN110853050B CN 110853050 B CN110853050 B CN 110853050B CN 201910998432 A CN201910998432 A CN 201910998432A CN 110853050 B CN110853050 B CN 110853050B
Authority
CN
China
Prior art keywords
image
river
region
spectrum
segmentation
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
Application number
CN201910998432.9A
Other languages
Chinese (zh)
Other versions
CN110853050A (en
Inventor
张桂铭
张弓
王建涛
罗李焱
姚明明
黄珍梅
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CETC 29 Research Institute
Original Assignee
CETC 29 Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by CETC 29 Research Institute filed Critical CETC 29 Research Institute
Priority to CN201910998432.9A priority Critical patent/CN110853050B/en
Publication of CN110853050A publication Critical patent/CN110853050A/en
Application granted granted Critical
Publication of CN110853050B publication Critical patent/CN110853050B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10032—Satellite or aerial image; Remote sensing

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a SAR image river segmentation method, which comprises the following steps: removing image noise spots and retaining fine information of the image; (2) binarization coarse segmentation: performing binarization segmentation on the image by adopting a threshold value; (3) Extracting a saliency map of the binarized rough segmented image by using a visual attention calculation model; (4) On the basis of the saliency map, selecting seed points in a river salient region through manual interaction; (5) Dividing and growing a river region by adopting a region growing method, and stopping the algorithm if the river region is divided; if the river region is not completely segmented, jumping to the step (4) until the river region is completely segmented. Compared with a threshold method, the method has higher precision and better comprehensive segmentation performance; the method has the advantages of higher precision, lower alarm rate and better comprehensive segmentation performance, and also effectively solves the problem that the segmentation result of the region growing method has holes.

Description

SAR image river segmentation method, device and medium
Technical Field
The invention relates to the technical field of remote sensing image processing, in particular to a SAR image river segmentation method, device and medium based on a visual attention mechanism and region growth.
Background
The river channel segmentation has very important significance in the aspects of water conservancy evaluation, water and soil conservation, transportation, flood disaster prevention and control and the like. Synthetic aperture radar (Synthetic Aperture Radar, SAR) is used as an imaging radar working in a microwave frequency band, and river target detection technology based on SAR images is attracting more and more attention due to the advantages of full time, all weather, strong penetrating power, wide monitoring range and the like. In general, compared with surrounding adjacent ground objects, the backscattering coefficient of the river surface is relatively smaller, and the echo signal is weaker, so that the gray value of the river surface in the SAR image is lower and black, and the gray value of the land in the image shows obvious difference, and therefore, the river region and the background are often segmented based on the gray information of the water body pixels.
Studies have shown that the human visual system can quickly screen important targets and ignore insignificant information in the face of complex scenes, and this selective and proactive physiological and psychological activity is called the visual attention mechanism. In the SAR image, the river region and the background show obvious gray scale difference, so that the river region is obvious relative to the background, and the river region range is quickly segmented by applying the visual saliency model, thereby being beneficial to improving the information processing efficiency and the detection precision of the whole method. The currently common visual attention calculation models are: ITTI visual attention calculation model, AIM visual attention calculation model, and spectral residual (Spectral Residual, SR) visual attention calculation model. Compared with the first two visual attention calculation models, the spectrum residual error model is based on frequency domain processing, is simple and quick in calculation, and is beneficial to rapidly extracting a saliency map of a wide SAR image and constructing a real-time system.
In the SAR image river segmentation problem, a threshold method and a region growing method are two segmentation methods which are widely applied and based on water body pixel information. The threshold method has very high segmentation efficiency, low omission factor of the segmentation result, but poor noise resistance, and can not solve the problem of 'same spectrum foreign matters', so that the segmentation result has low precision; the region growing method has the characteristics of simple method, high operation efficiency, higher precision of a segmentation result and the like, but the problem of high false alarm rate caused by the cavity and the void of the segmentation result due to the uneven noise and gray scale.
In view of the above drawbacks, it is necessary to provide a method for segmenting a river in an SAR image, which has the advantages of higher precision and lower false alarm rate, and has higher operation segmentation efficiency.
Disclosure of Invention
The invention aims to provide a SAR image river segmentation method, a device and a medium based on a visual attention mechanism and region growth, wherein the segmentation operation efficiency is equivalent to that of the traditional region growth method, the problem of void in the segmentation result of the traditional region growth method is effectively solved, and compared with the segmentation result of a threshold method and the segmentation result of the region growth method, the SAR image river segmentation method, the device and the medium have higher precision and better comprehensive segmentation performance.
The aim of the invention is realized by the following technical scheme:
in order to overcome the defects of the existing SAR image river segmentation method of mainstream application, the invention provides an SAR image river segmentation method based on a visual attention mechanism and region growth, which comprises the following steps:
(1) Pretreatment: removing image noise spots and reserving fine information of the image to obtain a filtered image;
(2) Binarization rough segmentation: after preprocessing is completed, performing binarization segmentation on the image by adopting a threshold value, and setting the pixel value to be 0 when the pixel value in the image is smaller than or equal to the threshold value; when the pixel value in the image is greater than the threshold value, the pixel value is set to 255;
(3) Extracting a saliency map of the binarized rough segmented image by using a visual attention calculation model;
(4) On the basis of the saliency map, selecting seed points in a river salient region through manual interaction;
(5) Dividing and growing a river region by adopting a region growing method, and stopping the algorithm if the river region is divided; if the river region is not completely segmented, jumping to the step (4) until the river region is completely segmented.
In the preprocessing, a Lee filter or an enhanced Lee filter is adopted to filter the SAR original image, so that a filtered image is obtained.
Preferably, the preprocessing step further comprises performing piecewise linear stretching transformation on the filtered image, improving the contrast between the river region and the background, setting a proportion of pixel values in the image with gray values lower than a given gray value threshold to 0, setting the same proportion of pixel values with gray values higher than the given gray value threshold to 255, and expanding the rest pixels to a range of 0-255 through linear transformation.
Preferably, the linear stretching transformation is 2% linear stretching transformation, namely, 2% pixel values of which the gray values are lower than a given gray value threshold value in an image are set to 0, 2% pixel values of which the gray values are higher than the given gray value threshold value are set to 255, and the rest pixels are expanded to a range of 0-255 through linear transformation.
Preferably, the binary rough segmentation threshold value is obtained by an empirical method, and the range of the threshold value is 30-130.
Preferably, the visual attention calculation model is a spectrum residual error (Spectral Residual, SR) calculation model, an image I (x) is given, and the image is decomposed into an amplitude spectrum A (f) and a phase spectrum P (f) after Fourier transformation, namely
A(f)＝Amp{FFT[I(x)]}
P(f)＝Pha{FFT[I(x)]}
The amplitude spectrum is then logarithmically transformed as shown in the following equation:
L(f)＝log[A(f)]
and subtracting the log-magnitude spectrum smoothed by convolution with the log-magnitude spectrum by a low-pass filter to obtain a spectrum residual, wherein the spectrum residual is as shown in the following formula:
R(f)＝L(f)-h n (f)*L(f)
and finally, carrying out inverse Fourier transform on the spectrum residual error and the initial phase spectrum to obtain a saliency map of the original image, and carrying out smoothing treatment on the saliency map through a Gaussian filter to obtain a better visual effect, wherein the following formula is shown:
S(x)＝g(x)*|FFT -1 {exp|R(f)+i·P(f)]}| 2
in the above, FFT and FFT -1 Respectively representing fourier transforms and inverse transforms thereof; a (f), P (f), L (f), R (f) and S (x) respectively represent the amplitude spectrum, the phase spectrum, the logarithmic amplitude spectrum, the spectrum residual and the saliency map of the original image; h is a n (f) Representing a low pass filter in the frequency domain; g (x) represents emptyA low pass filter in the domain.
Preferably, the apparatus comprises:
and a pretreatment module: the method comprises the steps of removing noise spots of an image and keeping fine information of the image to obtain a filtered image;
binarization rough segmentation module: after preprocessing is completed, performing binarization segmentation on the image by adopting a threshold value, and setting the pixel value to be 0 when the pixel value in the image is smaller than or equal to the threshold value; when the pixel value in the image is greater than the threshold value, the pixel value is set to 255;
the saliency map extraction module: extracting a saliency map of the binarized rough segmented image by using a visual attention calculation model;
and a manual interaction module: on the basis of the saliency map, selecting seed points in a river salient region through manual interaction;
region growth segmentation module: dividing a river region grown by adopting a region growing method, and stopping running if the river region is divided; and if the river region is not completely segmented, reselecting the seed point through the manual interaction module until the river region is completely segmented.
In a preferred mode, in the preprocessing module, a Lee filter or an enhanced Lee filter is adopted to filter the SAR original image, so as to obtain a filtered image.
Preferably, the preprocessing module performs piecewise linear stretching transformation on the filtered image, improves the contrast between the river region and the background, sets a certain proportion of pixel values in the image with gray values lower than a given gray value threshold to 0, sets the same proportion of pixel values with gray values higher than the given gray value threshold to 255, and expands the rest pixels to a range of 0-255 through linear transformation.
Preferably, the computer program is executed by a processor by a method according to any of claims 1-6.
The beneficial effects of the invention are as follows: the SAR image river segmentation based on the visual attention mechanism and the regional growth has at least the following advantages: compared with a threshold method, the method has higher precision and better comprehensive segmentation performance; the method has the advantages of higher precision, lower alarm rate and better comprehensive segmentation performance, and also effectively solves the problem that the segmentation result of the region growing method has holes.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some examples of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flow chart of a method for segmenting a river in an SAR image based on a visual attention mechanism and region growing according to the invention;
fig. 2 shows an SAR original image and an intermediate process output image according to an embodiment of the present invention, where (a) is the SAR original image, (b) is the preprocessed image, (c) is the binarized segmented image, and (d) is the saliency map;
fig. 3 is a comparison chart of segmentation results of each algorithm of the SAR image according to an embodiment of the present invention, where (a) is a manual segmentation result, (b) is a thresholding segmentation result, (c) is a region growing segmentation result, and (d) is a segmentation result according to the present invention.
Detailed Description
The technical solution of the present invention will be described in further detail with reference to the accompanying drawings, but the scope of the present invention is not limited to the following description.
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, based on the embodiments of the invention, which are apparent to those of ordinary skill in the art without inventive faculty, are intended to be within the scope of the invention. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, based on the embodiments of the invention, which are apparent to those of ordinary skill in the art without inventive faculty, are intended to be within the scope of the invention.
Example 1
As shown in fig. 1, in order to overcome the drawbacks of the existing mainstream application SAR image river segmentation method, the present invention provides a SAR image river segmentation method based on visual attention mechanism and region growth, comprising the following steps:
(1) Pretreatment: removing image noise spots and reserving fine information of the image to obtain a filtered image;
(2) Binarization rough segmentation: after preprocessing is completed, performing binarization segmentation on the image by adopting a threshold value, and setting the pixel value to be 0 when the pixel value in the image is smaller than or equal to the threshold value; when the pixel value in the image is greater than the threshold value, the pixel value is set to 255;
(3) Extracting a saliency map of the binarized rough segmented image by using a visual attention calculation model; fig. 2 shows an SAR original image and an intermediate process output image according to an embodiment of the present invention, where (a) is the SAR original image, (b) is the preprocessed image, (c) is the binarized segmented image, and (d) is the saliency map.
(4) On the basis of the saliency map, selecting seed points in a river salient region through manual interaction;
(5) Dividing and growing a river region by adopting a region growing method, and stopping the algorithm if the river region is divided; if the river region is not completely segmented, jumping to the step (4) until the river region is completely segmented.
In preprocessing, a Lee filter or an enhanced Lee filter is adopted to filter an SAR original image, so that a filtered image is obtained. In the preprocessing step, piecewise linear stretching transformation is carried out on the filtered image, the contrast between the river region and the background is improved, a certain proportion of pixel values with gray values lower than a given gray value threshold in the image are set to be 0, the same proportion of pixel values with gray values higher than the given gray value threshold are set to be 255, and the rest pixels are expanded to be in a range of 0-255 through linear transformation.
The linear stretching transformation is 2% linear stretching transformation, namely, 2% pixel values with gray values lower than a given gray value threshold value in an image are set to 0, 2% pixel values with gray values higher than the given gray value threshold value are set to 255, and the rest pixels are expanded to a range of 0-255 through linear transformation.
The binarization rough segmentation threshold value is obtained by an empirical method, and the range of the threshold value is 30-130.
The visual attention calculation model is a spectral residual (Spectral Residual, SR) calculation model, and the logarithmic magnitude spectrum of each natural image has an approximate trend, so that the change part, i.e., the salient region, in the image is caused by the spectral residual. Given an image I (x), the image is decomposed into an amplitude spectrum A (f) and a phase spectrum P (f) after Fourier transformation, namely
A(f)＝Amp{FFT[I(x)]}
P(f)＝Pha{FFT[I(x)]}
The amplitude spectrum is then logarithmically transformed as shown in the following equation:
L(f)＝log[A(f)]
and subtracting the log-magnitude spectrum smoothed by convolution with the log-magnitude spectrum by a low-pass filter to obtain a spectrum residual, wherein the spectrum residual is as shown in the following formula:
R(f)＝L(f)-h n (f)*L(f)
and finally, carrying out inverse Fourier transform on the spectrum residual error and the initial phase spectrum to obtain a saliency map of the original image, and carrying out smoothing treatment on the saliency map through a Gaussian filter to obtain a better visual effect, wherein the following formula is shown:
S(x)＝g(x)*|FFr -1 {exp[R(f)+i·P(f)]}| 2
in the above, FFT and FFT- 1 Respectively representing fourier transforms and inverse transforms thereof; a (f), P (f), L (f), R (f) and S (x) respectively represent the amplitude spectrum, the phase spectrum, the logarithmic amplitude spectrum, the spectrum residual and the saliency map of the original image; h is a n (f) Representing a low pass filter in the frequency domain; g (x) represents emptyA low pass filter in the domain.
Example two
Since the apparatus described in this embodiment is an apparatus for implementing a method for dividing a river in an SAR image in this embodiment, the method described in this embodiment of the present invention, those skilled in the art can understand the specific implementation and various modifications of the apparatus in this embodiment, so how the apparatus implements the method in this embodiment of the present invention will not be described in detail herein. As long as the person skilled in the art uses the apparatus for implementing the method according to the embodiments of the present invention, it is within the scope of protection of the present invention.
A SAR image river segmentation apparatus, the apparatus comprising:
and a pretreatment module: the method comprises the steps of removing noise spots of an image and keeping fine information of the image to obtain a filtered image;
binarization rough segmentation module: after preprocessing is completed, performing binarization segmentation on the image by adopting a threshold value, and setting the pixel value to be 0 when the pixel value in the image is smaller than or equal to the threshold value; when the pixel value in the image is greater than the threshold value, the pixel value is set to 255;
the saliency map extraction module: extracting a saliency map of the binarized rough segmented image by using a visual attention calculation model;
and a manual interaction module: on the basis of the saliency map, selecting seed points in a river salient region through manual interaction;
region growth segmentation module: dividing a river region grown by adopting a region growing method, and stopping running if the river region is divided; and if the river region is not completely segmented, reselecting the seed point through the manual interaction module until the river region is completely segmented.
And in the preprocessing module, filtering the SAR original image by adopting a Lee filter or an enhanced Lee filter to obtain a filtered image.
The preprocessing module performs piecewise linear stretching transformation on the filtered image, improves the contrast between a river region and a background, sets a certain proportion of pixel values with gray values lower than a given gray value threshold value in the image as 0, sets the same proportion of pixel values with gray values higher than the given gray value threshold value as 255, and expands the rest pixels to be in a range of 0-255 through linear transformation.
The binary rough segmentation threshold value is obtained by an empirical method, and the range of the threshold value is 30-130.
The visual attention calculation model is a spectrum residual calculation model, an image I (x) is given, and the image is decomposed into an amplitude spectrum A (f) and a phase spectrum P (f) after Fourier transformation, namely:
A(f)＝Amp{FFT[I(x)]}
P(f)＝Pha{FFT[I(x)]}
the amplitude spectrum is then logarithmically transformed as shown in the following equation:
L(f)＝log[A(f)]
and subtracting the log-magnitude spectrum smoothed by convolution with the log-magnitude spectrum by a low-pass filter to obtain a spectrum residual, wherein the spectrum residual is as shown in the following formula:
R(f)＝L(f)-h n (f)*L(f)
and finally, carrying out inverse Fourier transform on the spectrum residual error and the initial phase spectrum to obtain a saliency map of the original image, and carrying out smoothing treatment on the saliency map through a Gaussian filter to obtain a better visual effect, wherein the following formula is shown:
S(x)＝g(x)*|FFT -1 {exp|R(f)+i·P(f)]}| 2
in the above, FFT and FFT -1 Respectively representing fourier transforms and inverse transforms thereof; a (f), P (f), L (f), R (f) and S (x) respectively represent the amplitude spectrum, the phase spectrum, the logarithmic amplitude spectrum, the spectrum residual and the saliency map of the original image; h is a n (f) Representing a low pass filter in the frequency domain; g (x) denotes a low-pass filter in the space domain.
Example III
The present embodiment provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method embodiments.
Example IV
The invention provides a SAR image river segmentation method, device and medium based on a visual attention mechanism and region growth, and relates to the technical problems of binarization river region rough segmentation processing, spectrum residual error visual saliency calculation model-based saliency map extraction, river segmentation based on a region growth method and the like.
1. Experimental conditions
The hardware environment for implementation is: intel (R) Xeon (R) CPU E5-2603 v4@1.70GHz processor, 32.0GB memory and 4.0GB video memory, and the running software environment is: matlab R2017b and Windows7.
2. Experimental content and results analysis
The present invention uses SAR images with pixel sizes of 480 x 341 for test experiments, as shown with reference to fig. 2. Fig. 2 (a) shows an original SAR image, from which it can be seen that the river region is significant with respect to the background, and that there is a lot of speckle noise in the mechanical image due to coherent imaging; FIG. 2 (b) shows a pre-processed image in which the pre-processed speckle noise is effectively suppressed, the contrast between the river region and the background is enhanced, and the filter in this embodiment employs an enhanced Lee filter to suppress the speckle noise (where the filter size is 3×3), and the piecewise linear stretching transformation is specifically 2% linear stretching; fig. 2 (c) shows an image after the binarization rough segmentation process, and the segmented river region is more remarkable than the background, and the binarization rough segmentation threshold value is 60 in this embodiment; fig. 2 (d) shows a saliency map extracted by using an SR model, and in contrast to fig. 2 (c), it can be seen that the uniformly connected region in the later map is a complete and bounded closed region on the saliency map, and it is suitable to further precisely divide the river region by using a region growing method, where in this embodiment, the low-pass filter of the SR model in the frequency domain is selected as a mean filter (where the filter size is 3×3), and the low-pass filter in the space domain is selected as a gaussian smoothing filter (where the filter size is 10×10, and the standard deviation of the filter is 0.3).
FIG. 3 shows river segmentation results obtained by the artificial marking, thresholding, region growing methods, and methods according to the present invention. As can be seen from fig. 3 (b), the threshold method cannot solve the problem of "homospectral foreign matter", and the river segmentation accuracy is low; as can be seen from fig. 3 (c), the accuracy of dividing the river region by the region growing method is higher, but there are many voids in the division result, so that the omission rate of the division result is higher; fig. 3 (d) shows the division result of the present invention, and it is apparent from the figure that the river division accuracy is high, and that no significant void is found in the river region after division. The method quantitatively analyzes the advantages and disadvantages of the three river segmentation methods by respectively calculating the precision, the false alarm rate, the F value (evaluating the comprehensive segmentation performance) and the operation time, and the calculation results are shown in the table 1, so that the method provided by the invention has the highest precision and comprehensive segmentation performance in the three methods, and has equivalent operation efficiency compared with the region growing method, but the false alarm rate is reduced while the problem of void in the river region segmentation result is effectively solved. In conclusion, the introduction of the visual attention mechanism detection theory effectively improves the information processing efficiency and the detection precision of the method.
Table 13 performance comparisons of the methods
Figure GDA0004102656280000111
/>
It will be apparent to those skilled in the art that embodiments of the present invention may be a method, an apparatus, or a 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 employ a computer program product embodied on one or more storage media (including disk storage, CD-ROM, optical storage) having computer program code embodied therein.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
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 once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the invention. The foregoing description of the preferred embodiment of the invention is not intended to be limiting, but rather to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

Claims (9)

1. The SAR image river segmentation method is characterized by comprising the following steps of:
(1) Pretreatment: removing image noise spots and reserving fine information of the image to obtain a filtered image;
(2) Binarization rough segmentation: after preprocessing is completed, performing binarization segmentation on the image by adopting a threshold value, and setting the pixel value to be 0 when the pixel value in the image is smaller than or equal to the threshold value; when the pixel value in the image is greater than the threshold value, the pixel value is set to 255;
(3) Extracting a saliency map of the binarized rough segmented image by using a visual attention calculation model;
(4) On the basis of the saliency map, selecting seed points in a river salient region through manual interaction;
(5) Dividing and growing a river region by adopting a region growing method, and stopping the algorithm if the river region is divided; if the river region is not completely segmented, jumping to the step (4) until the river region is completely segmented;
the visual attention calculation model is a spectrum residual calculation model, an image I (x) is given, and the image is decomposed into an amplitude spectrum A (f) and a phase spectrum P (f) after Fourier transformation, namely
A(f)＝Amp{FFT[I(x)]}
P(f)＝Pha{FFT[I(x)]}
The amplitude spectrum is then logarithmically transformed as shown in the following equation:
L(f)＝log[A(f)]
and subtracting the log-magnitude spectrum smoothed by convolution with the log-magnitude spectrum by a low-pass filter to obtain a spectrum residual, wherein the spectrum residual is as shown in the following formula:
R(f)＝L(f)-h n (f)*L(f)
and finally, carrying out inverse Fourier transform on the spectrum residual error and the initial phase spectrum to obtain a saliency map of the original image, and carrying out smoothing treatment on the saliency map through a Gaussian filter to obtain a better visual effect, wherein the following formula is shown:
S(x)＝g(x)*|FFT -1 (exp[R(f)+i·P(f)]}| 2
in the above, FFT and FFT -1 Respectively representing fourier transforms and inverse transforms thereof; a (f), P (f), L (f), R (f) and S (x) respectively represent the amplitude spectrum, the phase spectrum, the logarithmic amplitude spectrum, the spectrum residual and the saliency map of the original image; h is a n (f) Representing a low pass filter in the frequency domain; g (x) denotes a low-pass filter in the space domain.
2. The SAR image river splitting method of claim 1, wherein: in preprocessing, a Lee filter or an enhanced Lee filter is adopted to filter an SAR original image, so that a filtered image is obtained.
3. The SAR image river splitting method of claim 1, wherein: in the preprocessing step, piecewise linear stretching transformation is carried out on the filtered image, the contrast between the river region and the background is improved, a certain proportion of pixel values with gray values lower than a given gray value threshold in the image are set to be 0, the same proportion of pixel values with gray values higher than the given gray value threshold are set to be 255, and the rest pixels are expanded to be in a range of 0-255 through linear transformation.
4. A SAR image river splitting method according to claim 3, wherein: the linear stretching transformation is 2% linear stretching transformation, namely, 2% pixel values of gray values lower than a given gray value threshold value in an image are set to be 0, 2% pixel values of gray values higher than the given gray value threshold value are set to be 255, and the rest pixels are expanded to be in a range of 0-255 through linear transformation.
5. The SAR image river splitting method of claim 1, wherein: the binary rough segmentation threshold value is obtained by an empirical method, and the range of the threshold value is 30-130.
6. A SAR image river segmentation apparatus, comprising:
and a pretreatment module: the method comprises the steps of removing noise spots of an image and keeping fine information of the image to obtain a filtered image;
binarization rough segmentation module: after preprocessing is completed, performing binarization segmentation on the image by adopting a threshold value, and setting the pixel value to be 0 when the pixel value in the image is smaller than or equal to the threshold value; when the pixel value in the image is greater than the threshold value, the pixel value is set to 255;
the saliency map extraction module: extracting a saliency map of the binarized rough segmented image by using a visual attention calculation model;
and a manual interaction module: on the basis of the saliency map, selecting seed points in a river salient region through manual interaction;
region growth segmentation module: dividing a river region grown by adopting a region growing method, and stopping running if the river region is divided; if the river region is not completely segmented, reselecting seed points through a manual interaction module until the river region is completely segmented;
the visual attention calculation model is a spectrum residual calculation model, an image I (x) is given, and the image is decomposed into an amplitude spectrum A (f) and a phase spectrum P (f) after Fourier transformation, namely
A(f)＝Amp{FFT[I(x)]}
P(f)＝Pha{FFT[I(x)]}
The amplitude spectrum is then logarithmically transformed as shown in the following equation:
L(f)＝log[A(f)]
and subtracting the log-magnitude spectrum smoothed by convolution with the log-magnitude spectrum by a low-pass filter to obtain a spectrum residual, wherein the spectrum residual is as shown in the following formula:
R(f)＝L(f)-h n (f)*L(f)
and finally, carrying out inverse Fourier transform on the spectrum residual error and the initial phase spectrum to obtain a saliency map of the original image, and carrying out smoothing treatment on the saliency map through a Gaussian filter to obtain a better visual effect, wherein the following formula is shown:
S(x)＝g(x)*|FFT -1 (exp[R(f)+i·P(f)]}| 2
in the above, FFT and FFT -1 Respectively representing fourier transforms and inverse transforms thereof; a (f), P (f), L (f), R (f) and S (x) respectively represent the amplitude spectrum, the phase spectrum, the logarithmic amplitude spectrum, the spectrum residual and the saliency map of the original image; h is a n (f) Representing a low pass filter in the frequency domain; g (x) denotes a low-pass filter in the space domain.
7. The SAR image river splitting device of claim 6, wherein: and in the preprocessing module, filtering the SAR original image by adopting a Lee filter or an enhanced Lee filter to obtain a filtered image.
8. The SAR image river splitting device of claim 6, wherein: the preprocessing module performs piecewise linear stretching transformation on the filtered image, improves the contrast between a river region and a background, sets a certain proportion of pixel values with gray values lower than a given gray value threshold value in the image as 0, sets the same proportion of pixel values with gray values higher than the given gray value threshold value as 255, and expands the rest pixels to be in a range of 0-255 through linear transformation.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program is executed by a processor for performing the method according to any one of claims 1-5.
CN201910998432.9A 2019-10-21 2019-10-21 SAR image river segmentation method, device and medium Active CN110853050B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910998432.9A CN110853050B (en) 2019-10-21 2019-10-21 SAR image river segmentation method, device and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910998432.9A CN110853050B (en) 2019-10-21 2019-10-21 SAR image river segmentation method, device and medium

Publications (2)

Publication Number Publication Date
CN110853050A CN110853050A (en) 2020-02-28
CN110853050B true CN110853050B (en) 2023-05-26

Family

ID=69596601

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910998432.9A Active CN110853050B (en) 2019-10-21 2019-10-21 SAR image river segmentation method, device and medium

Country Status (1)

Country Link
CN (1) CN110853050B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111696054B (en) * 2020-05-21 2022-08-26 河海大学 Rubber dam body detection method based on full-polarization SAR image
CN112258522B (en) * 2020-10-19 2021-05-04 哈尔滨体育学院 Martial arts competition area segmentation method based on secondary area growth

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102129559A (en) * 2011-04-22 2011-07-20 西安电子科技大学 SAR (Synthetic Aperture Radar) image object detection method based on Primal Sketch algorithm
CN102509290A (en) * 2011-10-25 2012-06-20 西安电子科技大学 Saliency-based synthetic aperture radar (SAR) image airfield runway edge detection method
CN105354541A (en) * 2015-10-23 2016-02-24 西安电子科技大学 SAR (Synthetic Aperture Radar) image target detection method based on visual attention model and constant false alarm rate
CN107274401A (en) * 2017-06-22 2017-10-20 中国人民解放军海军航空工程学院 A kind of High Resolution SAR Images Ship Detection of view-based access control model attention mechanism
CN107784655A (en) * 2016-12-28 2018-03-09 中国测绘科学研究院 A kind of visual attention model SAR naval vessels detection algorithm of adaptive threshold
CN108961284A (en) * 2018-06-12 2018-12-07 中国电子科技集团公司第二十九研究所 SAR image building extracting method, equipment and the storage medium of side lobe effect pollution
CN109214439A (en) * 2018-08-22 2019-01-15 电子科技大学 A kind of infrared image icing River detection method based on multi-feature fusion

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020154833A1 (en) * 2001-03-08 2002-10-24 Christof Koch Computation of intrinsic perceptual saliency in visual environments, and applications

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102129559A (en) * 2011-04-22 2011-07-20 西安电子科技大学 SAR (Synthetic Aperture Radar) image object detection method based on Primal Sketch algorithm
CN102509290A (en) * 2011-10-25 2012-06-20 西安电子科技大学 Saliency-based synthetic aperture radar (SAR) image airfield runway edge detection method
CN105354541A (en) * 2015-10-23 2016-02-24 西安电子科技大学 SAR (Synthetic Aperture Radar) image target detection method based on visual attention model and constant false alarm rate
CN107784655A (en) * 2016-12-28 2018-03-09 中国测绘科学研究院 A kind of visual attention model SAR naval vessels detection algorithm of adaptive threshold
CN107274401A (en) * 2017-06-22 2017-10-20 中国人民解放军海军航空工程学院 A kind of High Resolution SAR Images Ship Detection of view-based access control model attention mechanism
CN108961284A (en) * 2018-06-12 2018-12-07 中国电子科技集团公司第二十九研究所 SAR image building extracting method, equipment and the storage medium of side lobe effect pollution
CN109214439A (en) * 2018-08-22 2019-01-15 电子科技大学 A kind of infrared image icing River detection method based on multi-feature fusion

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
杨伟.选择性视觉注意机制及其在图像处理中的应用.《中国优秀硕士学位论文全文数据库 信息科技辑》.2013,第I138-1101页. *
贺飞跃 等.合成孔径雷达图像水体目标的提取.《电光与控制》.2018,第25卷(第11期),第21-24、61页. *

Also Published As

Publication number Publication date
CN110853050A (en) 2020-02-28

Similar Documents

Publication Publication Date Title
CN110031843B (en) ROI (region of interest) -based SAR (synthetic Aperture Radar) image target positioning method, system and device
CN106709426B (en) Ship target detection method based on infrared remote sensing image
BRPI0808278A2 (en) "COMPUTER IMPLEMENTED METHOD FOR PROCESSING INTERFEROMETRIC OPEN SYNTHETIC RADAR (SAR) IMAGES AND COMPUTER SYSTEM FOR PROCESSING INTERFEROMETRIC SYNTHETIC OPEN RADAR (SAR) IMAGES"
CN103942803A (en) SAR (Synthetic Aperture Radar) image based automatic water area detection method
BRPI0808277A2 (en) "COMPUTER IMPLEMENTED METHOD FOR PROCESSING SYNTHETIC OPENING RADAR IMAGES (SAR) AND COMPUTER SYSTEM FOR PROCESSING SYNTHETIC OPENING RADAR IMAGES (SAR)
BRPI0808280A2 (en) "COMPUTER IMPLEMENTED METHOD FOR UNCOMPRESSING SYNTHETIC OPENING RADAR (SAR) IMAGES AND COMPUTER SYSTEM FOR DECOMPRESSING SYNTHETIC OPENING RADAR (SAR) IMAGES"
CN110853050B (en) SAR image river segmentation method, device and medium
CN111008585B (en) Ship target detection method based on self-adaptive layered high-resolution SAR image
BRPI0808279A2 (en) "COMPUTER IMPLEMENTED METHOD FOR COMPRESSING SYNTHETIC OPENING RADAR (SAR) IMAGES AND COMPUTER SYSTEM FOR COMPRESSING SYNTHETIC OPENING RADAR (SAR) IMAGES"
CN101661611A (en) Realization method based on bayesian non-local mean filter
CN101489034A (en) Method for video image noise estimation and elimination
CN110717922A (en) Image definition evaluation method and device
BRPI0808276A2 (en) "COMPUTER IMPLEMENTED METHOD FOR PROCESSING SYNTHETIC OPENING RADAR (SAR) IMAGES AND COMPUTER SYSTEM FOR PROCESSING COMPLEX SYNTHETIC OPENING RADAR (SAR) IMAGES"
Nair et al. Fuzzy logic-based automatic contrast enhancement of satellite images of ocean
CN103065299A (en) Ultrasonic image edge extraction method and ultrasonic image edge extraction device
CN111160260B (en) SAR image target detection method and system
Azhar et al. Detail-preserving switching algorithm for the removal of random-valued impulse noise
Li et al. A method to improve the accuracy of SAR image change detection by using an image enhancement method
CN112529853A (en) Method and device for detecting damage of netting of underwater aquaculture net cage
CN104616259A (en) Non-local mean image de-noising method with noise intensity self-adaptation function
El Hassani et al. Efficient image denoising method based on mathematical morphology reconstruction and the Non-Local Means filter for the MRI of the head
CN117218149A (en) Image reconstruction method and system based on self-coding neural network
CN107146230A (en) Merge the SAR image segmentation method of cost based on K S distances
Mustaghfirin et al. The comparison of iris detection using histogram equalization and adaptive histogram equalization methods
CN103530636A (en) Snakes model based method for extracting SAR (synthetic aperture radar) image target profile

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