CN105528787A - Polarimetric SAR image bridge detection method and device based on level set segmentation - Google Patents

Abstract

The present invention provides a polarimetric SAR image bridge detection method and device based on level set segmentation. The method comprises a step of carrying out level set segmentation on a polarimetric SAR image according to a regional statistical characteristic and obtaining a land and a water area, a step of extracting the characteristic point of the contour of the water area and determining an area of interest in the land through the distance of the characteristic point to be a suspected bridge area, a step of removing the false alarm in the suspected bridge area to realize bridge detection, and a step of carrying out constant false alarm detection on the suspected bridge area with the removal of the false alarm and distinguishing a strong scatter bridge. According to the method of the embodiment of the present invention, the accuracy of bridge detection can be improved.

Description

Polarized SAR image bridge detection method and device based on level set segmentation

Technical Field

The invention relates to the technical field of image processing, in particular to a method and a device for detecting a polarized SAR image bridge based on level set segmentation.

Background

The ability to detect bridges in images is of great importance to updates to geographic databases, natural disaster assessment, military planning, and the like. In the related art, the bridge detection can be performed on a polarized SAR (synthetic aperture radar) image by methods such as edge detection or Randon transform, but the detection accuracy is mostly low.

Disclosure of Invention

The present invention is directed to solving, at least to some extent, one of the technical problems in the related art. Therefore, an object of the present invention is to provide a method for detecting a bridge based on a level set segmentation polarized SAR image, which can improve the accuracy of bridge detection.

The second purpose of the invention is to provide a polarized SAR image bridge detection device based on level set segmentation.

According to the embodiment of the first aspect of the invention, the polarized SAR image bridge detection method based on level set segmentation comprises the following steps: carrying out level set segmentation on the polarized SAR image according to the regional statistical characteristics to obtain land and a water area; extracting characteristic points of the contour of the water area, and determining an interested area in the land according to the distance of the characteristic points to serve as a suspected bridge area; removing false alarms in the suspected bridge area to realize bridge detection; and carrying out constant false alarm detection on the suspected bridge area after the false alarms are eliminated, and distinguishing the strong scatterer bridge.

According to the polarized SAR image bridge detection method based on level set segmentation, disclosed by the embodiment of the invention, the level set segmentation is carried out on the polarized SAR image to obtain the land and the water area, the characteristic points of the outline of the water area are extracted, the suspected bridge area is determined according to the characteristic points, and then the constant false alarm detection is carried out on the suspected bridge area without the false alarm, so that the strong scatterer bridge is distinguished. Therefore, land and water areas can be obtained more accurately by a level set segmentation method, and strong scatterer bridges can be distinguished by combining the processes of false alarm rejection, constant false alarm detection and the like, so that the accuracy of bridge detection can be greatly improved.

In addition, the method for detecting a polarized SAR image bridge based on level set segmentation according to the above embodiment of the present invention may further have the following additional technical features:

according to one embodiment of the invention, the feature points of the contour of the water area are extracted by a digital curve splitting and merging algorithm.

According to one embodiment of the invention, the region of interest that is not connected is taken as the false alarm.

According to an embodiment of the present invention, the performing constant false alarm detection on the suspected bridge area after the false alarms are removed specifically includes: and taking the suspected bridge area after the false alarm is removed as a point target, representing the point target by an area average coherence matrix, and detecting the point target by a polarized whitening filter.

The polarized SAR image bridge detection device based on level set segmentation according to the embodiment of the second aspect of the invention comprises: the segmentation module is used for carrying out level set segmentation on the polarized SAR image according to the regional statistical characteristics to obtain land and a water area; the determining module is used for extracting characteristic points of the contour of the water area and determining an interested area in the land as a suspected bridge area according to the distance of the characteristic points; the rejecting module is used for rejecting false alarms in the suspected bridge area to realize bridge detection; and the detection module is used for carrying out constant false alarm detection on the suspected bridge area after the false alarms are eliminated so as to distinguish the strong scatterer bridge.

According to the polarized SAR image bridge detection device based on level set segmentation, disclosed by the embodiment of the invention, the level set segmentation is carried out on the polarized SAR image to obtain the land and the water area, the characteristic points of the outline of the water area are extracted, the suspected bridge area is determined according to the characteristic points, and then the constant false alarm detection is carried out on the suspected bridge area without the false alarm, so that the strong scatterer bridge is distinguished. Therefore, land and water areas can be obtained more accurately through level set segmentation, strong scatterer bridges can be distinguished by combining false alarm rejection, constant false alarm detection and the like, and the accuracy of bridge detection can be greatly improved.

In addition, the polarized SAR image bridge detection apparatus based on level set segmentation according to the above embodiment of the present invention may further have the following additional technical features:

according to one embodiment of the invention, the determination module extracts feature points of the contour of the water area by a digital curve splitting and merging algorithm.

According to one embodiment of the invention, the culling module treats the region of interest that is not connected as the false alarm.

According to an embodiment of the present invention, the detection module is specifically configured to: and taking the suspected bridge area after the false alarm is removed as a point target, representing the point target by an area average coherence matrix, and detecting the point target by a polarized whitening filter.

Drawings

FIG. 1 is a flowchart of a method for detecting a bridge based on a level set segmentation polarized SAR image according to an embodiment of the present invention;

FIG. 2 is a schematic illustration of characteristic points of a contour of a water area and a suspected bridge area according to one embodiment of the invention;

FIG. 3 is a pseudo-color image of a polarized SAR image of a Singapore region including a plurality of bridges according to one embodiment of the present invention;

FIG. 4 is a diagram illustrating the result of performing a bridge inspection on the image of FIG. 3 according to one embodiment of the present invention;

FIG. 5 is a diagram illustrating the result of distinguishing a strong scatterer bridge according to an embodiment of the present invention;

fig. 6 is a block diagram of a structure of a polarized SAR image bridge detection apparatus based on level set segmentation according to an embodiment of the present invention.

Detailed Description

Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.

The method and the device for detecting the polarized SAR image bridge based on level set segmentation in the embodiment of the invention are described below with reference to the accompanying drawings.

Fig. 1 is a flowchart of a method for detecting a bridge based on a level set segmentation polarized SAR image according to an embodiment of the present invention.

As shown in fig. 1, the method for detecting a polarized SAR image bridge based on level set segmentation in the embodiment of the present invention includes the following steps:

and S101, carrying out level set segmentation on the polarized SAR image according to the regional statistical characteristics to obtain the land and the water area.

In the embodiment of the present invention, a multi-view polarization SAR image is taken as an example for description. The scattering matrix of the multi-view polarization SAR image is subjected to complex Wishart distribution, and for a homogeneous region, the coherence matrix T obtained by vectorization of the scattering matrix is also subjected to the complex Wishart distribution. If the mean value of the coherence matrix of the homogeneous region is Σ, the view is L, and the number of polarization channels is p, it can be written as T to W (Σ, L, p). Thus, the probability density function of the correlation matrix is:

$f\left(T\right|\mathrm{\Σ},L,p)=\frac{L^{pL}{\left(\right|T\left|\right)}^{L-p}\exp \{-Ltr\left({\mathrm{\Σ}}^{-1}T\right)\}}{K(L,p){\left(\right|\mathrm{\Σ}\left|\right)}^L}---\left(1\right)$

wherein, $K(L,p)={\mathrm{\π}}^{\frac{p(p-1)}{2}}G\left(L\right)\mathrm{...}G(L-p+1),$ g (-) is a Gamma (Gamma) function.

In one embodiment of the invention, the image is represented by RPlane, I for a given polarized SAR image, P for segmentation, to represent the boundary between land and water, wherein a closed curve is obtained, and R is obtained₁And R₂Representing land and water respectively segmented by curves, the posterior probability of the polarized SAR image I is P (I | P (R) under segmentation P₁,R₂)). According to the Bayes criterion, the segmentation with the maximum posterior probability is the optimal segmentation, namely the conditions met by the optimal segmentation are as follows:

$P=\underset{R_1,R_2}{\max }p\left(I\right|P\left(R_1,R_2\right))---\left(2\right)$

and because P (I | P) octo P (P | I) P (P), if land R₁To the water area R₂Independent of each other, the conditional probability function of the region is f (I)_i|R_i) Where I is 1 or 2, representing the probability density function of the image I when divided into land or water areas, respectively. The segmentation prior probability p (P) can be defined as the contour line length function p (S) planand e^-ν||,ν>0, the energy function of the segmentation curve and the equivalent form of the segmentation model can be obtained:

$E(\mathrm{\Γ},\{R_1,R_2\left\}\right)=v\left|\mathrm{\Γ}\right|-\underset{R_1}{\∫}\log f\left(x\right|R_1)dx-\underset{R_2}{\∫}\log f\left(x\right|R_2)dx$

$\widehat{\mathrm{\Γ}}=\underset{\mathrm{\Γ}}{\min }\left\{E(\mathrm{\Γ},\{R_1,R_2\left\}\right)\right\}---\left(3\right)$

wherein v is a curve normalization parameter, u The | represents the length of the curve,is the optimal segmentation curve.

In one embodiment of the present invention, if a curve is represented by the level set function Φ (c) (t), then the zero level set corresponds to the curve (t) { c (t) | Φ (c (t), t) ═ 0 }. Meanwhile, obtaining the energy definition of a level set general function phi according to the curve energy definition:

$E\left(\mathrm{\Φ}\right)=-\underset{R}{\∫}\left(H\right(\mathrm{\Φ}\left)log\right(f\left(x|R_1\right))+(1-H\left(\mathrm{\Φ}\right)\left)log\right(f\left(x|R_2\right)\left)\right)dx+v\underset{R}{\∫}|\▿H\left(\mathrm{\Φ}\right)|dx---\left(4\right)$

wherein, H (phi) is a step function, when phi is not less than 0, H (phi) is 1, when phi is not less than 0<When 0, H (Φ) is 0. R₁Corresponding to a region of phi ≥ 0, R₂Corresponds to phi<And a region 0.

The partial differential equation for the level set general function Φ is:

$\frac{\&part;\mathrm{\&Phi;}}{\&part;t}=-\mathrm{\&delta;}\left(\mathrm{\&Phi;}\right)(v\mathrm{\&kappa;}+log\frac{f\left(x\right|R_2)}{f\left(x\right|R_1)})---\left(5\right)$

wherein (phi) is an impulse function,is a curved curvature. Then the zero level set function under the condition of minimum energy can be solved according to the formula (5) and by successive approximation along the negative gradient direction of the level set energy function through the variational method. In the inventionIn one embodiment, if the average values of the coherence matrices for the water and land are Σ respectively₁,Σ₂Then the level set evolution function is:

$\frac{\&part;\mathrm{\&Phi;}}{\&part;t}=-\mathrm{\&delta;}\left(\mathrm{\&Phi;}\right)(v\mathrm{\&kappa;}+L(log\left|{\mathrm{\&Sigma;}}_1\right|+tr\left({{\mathrm{\&Sigma;}}_1}^{-1}T\right))-L(log\left|{\mathrm{\&Sigma;}}_2\right|+tr\left({{\mathrm{\&Sigma;}}_2}^{-1}T\right)\left)\right)---\left(6\right)$

wherein, sigma₁,Σ₂The method can be obtained through likelihood estimation, and iteration is carried out by using the formula (6) under a set initialization curve and the corresponding phi and parameter value v until the zero level set function is not changed any more, so that level set segmentation can be realized.

After the level set division is implemented, the classification of the two regions can be determined based on the average scattering power of the two regions to be divided, and generally, the average scattering power of the water region is small. As the contour of each region segmented by the level set is continuously closed, all connected regions can be obtained by an 8-connected domain judgment algorithm for the land and water segmentation binary image.

S102, extracting characteristic points of the outline of the water area, and determining an interested area in the land according to the distance of the characteristic points to serve as a suspected bridge area.

After the preliminary segmentation process, some water areas and lands with small areas exist in the segmentation result, and for the condition that the bridge is not particularly dense, the small area areas are considered to be irrelevant to bridge detection and can be removed under a set pixel area threshold, and the threshold can be determined according to the size and the resolution of the image. Therefore, areas with lower scattering intensity of the water area which are segmented by mistake are obtained, and if the distance between the connection parts of the areas is small, suspected bridge areas are formed. Considering that the detection bridge is positioned above the main sea surface or river branch, merging the water areas according to the distance between the water area outlines, setting a reasonable threshold value according to the width of the bridge, and extracting the water area parts which are close to the main water area and the branches of the main water area.

Specifically, the feature points of the contour of the water area can be extracted by a digital curve splitting and merging algorithm. In an embodiment of the present invention, the DP (Douglas-Peucker) algorithm can be selected as the digital curve splitting and merging algorithm, and the specific process is as follows: selecting the head and tail ends of the contour curve segment of the water area as initial characteristic points; acquiring the linear distances from all points on the curve segment to the initial characteristic points; if the straight-line distances from all points on the curve segment to the initial characteristic point are smaller than the preset maximum tolerance, determining that the initial characteristic point is the characteristic point of the profile of the water area, otherwise, selecting the point with the maximum straight-line distance from the initial characteristic point as a dividing point, dividing the curve segment into two sections by the dividing point, and then recursively adopting the algorithm for the two divided curve segments respectively until all the characteristic points of the profile curve segment of the water area are determined.

Fig. 2 shows characteristic points of the water area profile, wherein the black area is the land, the white area is the water area, and the marked points at the "+" are the characteristic points of the water area profile. Assume that all feature points { u } of the contour of the A water area are determined according to the above algorithm_iAll feature points of the outline of the water and B water { v }_iCan calculate { u }_iAnd { v } and_ithe distance of each feature point between. In an embodiment of the present invention, two feature points { p ] having a distance lower than a preset distance may be selected_iAnd q_iAnd forming a characteristic point pair, and determining the end point of the water area bridge according to the coordinate relation of each characteristic point in the characteristic point pair. Further, the characteristic point pairs may be merged by setting a threshold value according to the size of the bridge, and for example, all the characteristic point pairs having a distance lower than the set threshold value may be merged, so that when a plurality of bridges exist between the a and B waters, bridges having different lengths can be distinguished. If the corresponding sets of the A and B waters are respectively { { S after merging₁},...,{S_N} and { { W₁},...,{W_N}, then { S }_iAnd { W }and { W_iA new pair of feature points can be formed. Thus, a new characteristic point pair { S } can be further specified_iAnd { W }_iAnd (5) determining an area of interest according to the end point of the water area bridge represented by the point. For example, as shown in fig. 2, if the end point of the water area bridge is four points, a quadrilateral region composed of the four points is an interesting region, and the interesting region is a suspected bridge region. In a specific embodiment of the present invention, the preset maximum tolerance, the preset distance and the bridge length threshold may be set according to the scale of the polarized SAR image and the specific detection requirement.Therefore, by forming the characteristic point pair by the two characteristic points with the distance lower than the preset distance, the excessive calculation amount and the large determined region of interest caused by calculating all the characteristic points can be prevented, and the speed and the precision of bridge detection can be improved to a certain extent.

S103, rejecting false alarms in the suspected bridge area to realize bridge detection.

In the water area obtained in step S101, a part of land-based water areas close to the subject water area and its branches still exist, and the land between these water areas and other water areas is erroneously classified as a suspected bridge area, which becomes a false alarm target. Usually, compared with a real bridge, the false alarm targets have irregular outline shapes, and the bridge body areas determined by the end points are not communicated, so that the non-communicated interested areas can be used as false alarms and can be eliminated.

And S104, performing constant false alarm detection on the suspected bridge area without the false alarm, and distinguishing the strong scatterer bridge.

Generally, there are metal guardrails around important bridges, the scattering strength of the metal guardrails is high, and the scattering of the bridge deck, the guardrails and the water area at the dihedral or trihedral angles can make the scattering strength of the bridge area higher than that of other bridge targets and false alarm targets. Thus, the bridge region may be distinguished from the suspected bridge region by the scattering intensity. In one embodiment of the invention, strong scatterer bridge discrimination may be performed using a point target constant false alarm detector. Specifically, the suspected bridge area from which the false alarm is removed can be used as a point target, the point target is represented by an area average coherence matrix, and the point target is detected by a polarization whitening filter, so that the bridge with high scattering intensity can be effectively detected.

More specifically, if the average coherence matrix of the suspected bridge area after the false alarm is rejected is C, the polarization whitening filter may be:

$\mathrm{\&Lambda;}={\&Sigma;}_N^{-1}C---\left(7\right)$

wherein, sigma_NFor other regions, the Λ obedience parameter is (L, σ)²) The gamma distribution of (a), namely:

$p\left(\mathrm{\&Lambda;}\right|H_0)=\frac{1}{\mathrm{\&Gamma;}\left(L\right)}{\left(\frac{L}{{\mathrm{\&sigma;}}^2}\right)}^L{\mathrm{\&Lambda;}}^{L-1}{e}^{-\frac{L\mathrm{\&Lambda;}}{{\mathrm{\&sigma;}}^2}},\mathrm{\&Lambda;}\&GreaterEqual;0---\left(8\right)$

wherein L is an equivalent visual number σ²Is the average power, H₀Representing imagesOther regions of (a).

In one embodiment of the invention, the detection threshold may be given as γ, and if Λ > γ, then a decision point target is present.

Fig. 3 is a pseudo-color image of a polarized SAR image of a singapore area including a plurality of bridges according to an embodiment of the present invention, wherein the marked areas No. 1 to 14 are bridges, the resolution of the image is 4.73 × 4.80 meters, and the image size is 5491 × 2156 pixels. And carrying out bridge detection on the image through the steps, wherein the curve regularization parameter can be set to be 0.2, the iteration number can be set to be 100, the set pixel area threshold can be 1000 pixels, the preset maximum tolerance can be set to be 10 pixels, and the threshold of the size of the bridge can be 30 pixels wide and 150 pixels long. Fig. 4 shows the result of bridge inspection performed on the image in fig. 3 according to steps S101-S103, and proposes an algorithm to correctly inspect 13 (nos. 1-10 and 12-14), false alarms 3 (nos. 15-17), and omission 1 (No. 11) for 14 bridge targets in fig. 3, according to an embodiment of the invention. Fig. 5 shows the strong scatterer bridges distinguished according to step S104, and as shown in fig. 5, No. 1, 3, 4, 8 and 12 strong scatterer bridges are finally distinguished.

According to the polarized SAR image bridge detection method based on level set segmentation, disclosed by the embodiment of the invention, the level set segmentation is carried out on the polarized SAR image to obtain the land and the water area, the characteristic points of the outline of the water area are extracted, the suspected bridge area is determined according to the characteristic points, and then the constant false alarm detection is carried out on the suspected bridge area without the false alarm, so that the strong scatterer bridge is distinguished. Therefore, land and water areas can be obtained more accurately by a level set segmentation method, and strong scatterer bridges can be distinguished by combining the processes of false alarm rejection, constant false alarm detection and the like, so that the accuracy of bridge detection can be greatly improved.

In order to realize the embodiment, the invention further provides a polarized SAR image bridge detection device based on level set segmentation.

Fig. 6 is a block diagram of a structure of a polarized SAR image bridge detection apparatus based on level set segmentation according to an embodiment of the present invention.

As shown in fig. 6, the apparatus for detecting a polarized SAR image bridge based on level set segmentation according to the embodiment of the present invention includes: a segmentation module 10, a determination module 20, a culling module 30 and a detection module 40.

The segmentation module 10 is configured to perform level set segmentation on the polarized SAR image according to the regional statistical characteristics to obtain a land area and a water area; the determining module 20 is configured to extract feature points of the contour of the water area, and determine an area of interest in the land as a suspected bridge area according to distances between the feature points; the rejecting module 30 is used for rejecting false alarms in suspected bridge areas to realize bridge detection; the detection module 40 is configured to perform constant false alarm detection on the suspected bridge area from which the false alarms are rejected, and distinguish a strong scatterer bridge.

In the embodiment of the present invention, a multi-view polarization SAR image is taken as an example for description. The scattering matrix of the multi-view polarization SAR image is subjected to complex Wishart distribution, and for a homogeneous region, the coherence matrix T obtained by vectorization of the scattering matrix is also subjected to the complex Wishart distribution. If the mean value of the coherence matrix of the homogeneous region is Σ, the view is L, and the number of polarization channels is p, it can be written as T to W (Σ, L, p). Thus, the probability density function of the correlation matrix is:

$f\left(T\right|\&Sigma;,L,p)=\frac{L^{pL}{\left(\right|T\left|\right)}^{L-p}\exp \{-Ltr\left({\&Sigma;}^{-1}T\right)\}}{K(L,p){\left(\right|\&Sigma;\left|\right)}^L}---\left(1\right)$

wherein, $K(L,p)={\mathrm{\&pi;}}^{\frac{p(p-1)}{2}}G\left(L\right)\mathrm{...}G(L-p+1),$ g (-) is a Gamma (Gamma) function.

In one embodiment of the invention, the image plane is represented by R, the given polarized SAR image is represented by I, the segmentation is represented by P to represent the boundary line of land and water, wherein the curve is closed, and simultaneously, the R is used₁And R₂Representing land and water respectively segmented by curves, the posterior probability of the polarized SAR image I is P (I | P (R) under segmentation P₁,R₂)). According to the Bayes criterion, the segmentation with the maximum posterior probability is the optimal segmentation, namely the conditions met by the optimal segmentation are as follows:

$P=\underset{R_1,R_2}{\max }p\left(I\right|P\left(R_1,R_2\right))---\left(2\right)$

and because P (I | P) octo P (P | I) P (P), if land R₁To the water area R₂Independent of each other, the conditional probability function of the region is f (I)_i|R_i) Where I is 1 or 2, representing the probability density function of the image I when divided into land or water areas, respectively. The segmentation prior probability p (P) can be defined as the contour line length function p (S) planand e^-ν||,ν>0, the energy function of the segmentation curve and the equivalent form of the segmentation model can be obtained:

$E(\mathrm{\&Gamma;},\{R_1,R_2\left\}\right)=v\left|\mathrm{\&Gamma;}\right|-\underset{R_1}{\&Integral;}\log f\left(x\right|R_1)dx-\underset{R_2}{\&Integral;}\log f\left(x\right|R_2)dx$

$\widehat{\mathrm{\&Gamma;}}=\underset{\mathrm{\&Gamma;}}{\min }\left\{E(\mathrm{\&Gamma;},\{R_1,R_2\left\}\right)\right\}---\left(3\right)$

wherein ν is a curve normalization parameter, | | | represents a curve length,is the optimal segmentation curve.

In one embodiment of the present invention, if a curve is represented by the level set function Φ (c) (t), then the zero level set corresponds to the curve (t) { c (t) | Φ (c (t), t) ═ 0 }. Meanwhile, obtaining the energy definition of a level set general function phi according to the curve energy definition:

$E\left(\mathrm{\&Phi;}\right)=-\underset{R}{\&Integral;}\left(H\right(\mathrm{\&Phi;}\left)log\right(f\left(x|R_1\right))+(1-H\left(\mathrm{\&Phi;}\right)\left)log\right(f\left(x|R_2\right)\left)\right)dx+v\underset{R}{\&Integral;}|\&dtri;H\left(\mathrm{\&Phi;}\right)|dx---\left(4\right)$

wherein, H (phi) is a step function, when phi is not less than 0, H (phi) is 1, when phi is not less than 0<At the time of 0, the number of the first,H(Φ)＝0。R₁corresponding to a region of phi ≥ 0, R₂Corresponds to phi<And a region 0.

The partial differential equation for the level set general function Φ is:

$\frac{\&part;\mathrm{\&Phi;}}{\&part;t}=-\mathrm{\&delta;}\left(\mathrm{\&Phi;}\right)(v\mathrm{\&kappa;}+log\frac{f\left(x\right|R_2)}{f\left(x\right|R_1)})---\left(5\right)$

wherein (phi) is an impulse function,is a curved curvature. Then the zero level set function under the condition of minimum energy can be solved according to the formula (5) and by successive approximation along the negative gradient direction of the level set energy function through the variational method. In one embodiment of the invention, if the average values of the coherence matrices for water and land are Σ respectively₁,Σ₂Then the level set evolution function is:

$\frac{\&part;\mathrm{\&Phi;}}{\&part;t}=-\mathrm{\&delta;}\left(\mathrm{\&Phi;}\right)(v\mathrm{\&kappa;}+L(log\left|{\mathrm{\&Sigma;}}_1\right|+tr\left({{\mathrm{\&Sigma;}}_1}^{-1}T\right))-L(log\left|{\mathrm{\&Sigma;}}_2\right|+tr\left({{\mathrm{\&Sigma;}}_2}^{-1}T\right)\left)\right)---\left(6\right)$

wherein, sigma₁,Σ₂The method can be obtained through likelihood estimation, and iteration is carried out by using the formula (6) under a set initialization curve and the corresponding phi and parameter value v until the zero level set function is not changed any more, so that level set segmentation can be realized.

After the level set division is implemented, the classification of the two regions can be determined based on the average scattering power of the two regions to be divided, and generally, the average scattering power of the water region is small. As the contour of each region segmented by the level set is continuously closed, all connected regions can be obtained by an 8-connected domain judgment algorithm for the land and water segmentation binary image.

After the preliminary segmentation, some water areas and lands with small areas exist in the segmentation result, and for the condition that the bridge is not particularly dense, the small area areas are considered to be irrelevant to bridge detection and can be removed under a set pixel area threshold, and the threshold can be determined according to the size and the resolution of the image. Therefore, areas with lower scattering intensity of the water area which are segmented by mistake are obtained, and if the distance between the connection parts of the areas is small, suspected bridge areas are formed. Considering that the detection bridge is positioned above the main sea surface or river branch, merging the water areas according to the distance between the water area outlines, setting a reasonable threshold value according to the width of the bridge, and extracting the water area parts which are close to the main water area and the branches of the main water area.

Specifically, the determination module 20 may extract the feature points of the contour of the water area through a digital curve splitting and merging algorithm. In an embodiment of the present invention, the DP (Douglas-Peucker) algorithm can be selected as the digital curve splitting and merging algorithm, and the specific process is as follows: selecting the head and tail ends of the contour curve segment of the water area as initial characteristic points; acquiring the linear distances from all points on the curve segment to the initial characteristic points; if the straight-line distances from all points on the curve segment to the initial characteristic point are smaller than the preset maximum tolerance, determining that the initial characteristic point is the characteristic point of the profile of the water area, otherwise, selecting the point with the maximum straight-line distance from the initial characteristic point as a dividing point, dividing the curve segment into two sections by the dividing point, and then recursively adopting the algorithm for the two divided curve segments respectively until all the characteristic points of the profile curve segment of the water area are determined.

Fig. 2 shows characteristic points of the water area profile, wherein the black area is the land, the white area is the water area, and the marked points at the "+" are the characteristic points of the water area profile. Assume that all feature points { u } of the contour of the A water area are determined according to the above algorithm_iAll feature points of the outline of the water and B water { v }_iCan calculate { u }_iAnd { v } and_ithe distance of each feature point between. In an embodiment of the present invention, two feature points { p ] having a distance lower than a preset distance may be selected_iAnd q_iAnd forming a characteristic point pair, and determining the end point of the water area bridge according to the coordinate relation of each characteristic point in the characteristic point pair. Further, the characteristic point pairs may be merged by setting a threshold value according to the size of the bridge, and for example, all the characteristic point pairs having a distance lower than the set threshold value may be merged, so that when a plurality of bridges exist between the a and B waters, bridges having different lengths can be distinguished. If the corresponding sets of the A and B waters are respectively { { S after merging₁},...,{S_N} and { { W₁},...,{W_N}, then { S }_iAnd { W }and { W_iA new pair of feature points can be formed. Thus, a new characteristic point pair { S } can be further specified_iAnd { W }_iAnd (5) determining an area of interest according to the end point of the water area bridge represented by the point. For example, as shown in fig. 2, if the end point of the water area bridge is four points, a quadrilateral region composed of the four points is an interesting region, and the interesting region is a suspected bridge region. In a specific embodiment of the present invention, the preset maximum tolerance, the preset distance and the bridge length threshold may be set according to the scale of the polarized SAR image and the specific detection requirement. Therefore, by forming the characteristic point pair by the two characteristic points with the distance lower than the preset distance, the excessive calculation amount and the large determined region of interest caused by calculating all the characteristic points can be prevented, and the speed and the precision of bridge detection can be improved to a certain extent.

The water areas obtained by the division module 10 still have land-based water areas with a short distance from the main water area and its branches, and the land areas between these water areas and other water areas are wrongly divided into suspected bridge areas, which become false alarm targets. Usually, these false alarm targets have irregular outlines compared to real bridges, and the bridge body regions determined by the end points are not connected, so the removing module 30 can take the disconnected interested regions as false alarms and remove them.

Generally, there are metal guardrails around important bridges, the scattering strength of the metal guardrails is high, and the scattering of the bridge deck, the guardrails and the water area at the dihedral or trihedral angles can make the scattering strength of the bridge area higher than that of other bridge targets and false alarm targets. Thus, the bridge region may be distinguished from the suspected bridge region by the scattering intensity. In one embodiment of the invention, the detection module 40 may be a point target constant false alarm detector. Specifically, the suspected bridge area from which the false alarm is removed can be used as a point target, the point target is represented by an area average coherence matrix, and the point target is detected by a polarization whitening filter, so that the bridge with high scattering intensity can be effectively detected.

More specifically, if the average coherence matrix of the suspected bridge area after the false alarm is rejected is C, the polarization whitening filter may be:

$\mathrm{\&Lambda;}={\&Sigma;}_N^{-1}C---\left(7\right)$

wherein, sigma_NFor other regions, the Λ obedience parameter is (L, σ)²) The gamma distribution of (a), namely:

$p\left(\mathrm{\&Lambda;}\right|H_0)=\frac{1}{\mathrm{\&Gamma;}\left(L\right)}{\left(\frac{L}{{\mathrm{\&sigma;}}^2}\right)}^L{\mathrm{\&Lambda;}}^{L-1}{e}^{-\frac{L\mathrm{\&Lambda;}}{{\mathrm{\&sigma;}}^2}},\mathrm{\&Lambda;}\&GreaterEqual;0---\left(8\right)$

wherein L is an equivalent visual number σ²Is the average power, H₀Representing other areas of the image.

In one embodiment of the invention, the detection threshold may be given as γ, and if Λ > γ, then a decision point target is present.

Fig. 3 is a pseudo-color image of a polarized SAR image of a singapore area including a plurality of bridges according to an embodiment of the present invention, wherein the marked areas No. 1 to 14 are bridges, the resolution of the image is 4.73 × 4.80 meters, and the image size is 5491 × 2156 pixels. And carrying out bridge detection on the image through the device, wherein a curve regularization parameter can be set to be 0.2, the iteration number can be set to be 100, a set pixel area threshold can be 1000 pixels, a preset maximum tolerance can be set to be 10 pixels, and a threshold of the size of the bridge can be 30 pixels wide and 150 pixels long. Fig. 4 shows the result of performing bridge inspection on the image in fig. 3 according to an embodiment of the present invention, and for 14 bridge targets in fig. 3, an algorithm is proposed to correctly inspect 13 (nos. 1-10 and 12-14), false alarms 3 (nos. 15-17), and missed inspections 1 (No. 11). Fig. 5 shows the strong scatterer bridges distinguished by the detection module 40, and as shown in fig. 5, the number 1, 3, 4, 8, and 12 strong scatterer bridges are finally distinguished.

According to the polarized SAR image bridge detection device based on level set segmentation, disclosed by the embodiment of the invention, the level set segmentation is carried out on the polarized SAR image to obtain the land and the water area, the characteristic points of the outline of the water area are extracted, the suspected bridge area is determined according to the characteristic points, and then the constant false alarm detection is carried out on the suspected bridge area without the false alarm, so that the strong scatterer bridge is distinguished. Therefore, land and water areas can be obtained more accurately through level set segmentation, strong scatterer bridges can be distinguished by combining false alarm rejection, constant false alarm detection and the like, and the accuracy of bridge detection can be greatly improved.

In the description of the present invention, it is to be understood that the terms "central," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," "circumferential," and the like are used in the orientations and positional relationships indicated in the drawings for convenience in describing the invention and to simplify the description, and are not intended to indicate or imply that the referenced device or element must have a particular orientation, be constructed and operated in a particular orientation, and are not to be considered limiting of the invention.

Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.

In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; either directly or indirectly through intervening media, either internally or in any other relationship. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.

In the present invention, unless otherwise expressly stated or limited, the first feature "on" or "under" the second feature may be directly contacting the first and second features or indirectly contacting the first and second features through an intermediate. Also, a first feature "on," "over," and "above" a second feature may be directly or diagonally above the second feature, or may simply indicate that the first feature is at a higher level than the second feature. A first feature being "under," "below," and "beneath" a second feature may be directly under or obliquely under the first feature, or may simply mean that the first feature is at a lesser elevation than the second feature.

In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.

Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.

Claims (8)

1. A polarized SAR image bridge detection method based on level set segmentation is characterized by comprising the following steps:

carrying out level set segmentation on the polarized SAR image according to the regional statistical characteristics to obtain land and a water area;

extracting characteristic points of the contour of the water area, and determining an interested area in the land according to the distance of the characteristic points to serve as a suspected bridge area;

removing false alarms in the suspected bridge area to realize bridge detection;

and carrying out constant false alarm detection on the suspected bridge area after the false alarms are eliminated, and distinguishing the strong scatterer bridge.

2. The method for detecting the polarized SAR image bridge based on the level set segmentation as claimed in claim 1, characterized in that the feature points of the contour of the water area are extracted by a digital curve splitting and merging algorithm.

3. The level set segmentation based polarized SAR image bridge detection method of claim 1, characterized in that the region of interest that is not connected is used as the false alarm.

4. The method for detecting a polarized SAR image bridge based on level set segmentation as claimed in claim 1, wherein the performing of the constant false alarm detection on the suspected bridge area after the false alarm is removed specifically comprises:

and taking the suspected bridge area after the false alarm is removed as a point target, representing the point target by an area average coherence matrix, and detecting the point target by a polarized whitening filter.

5. The utility model provides a polarize SAR image bridge detection device based on level set segmentation which characterized in that includes:

the segmentation module is used for carrying out level set segmentation on the polarized SAR image according to the regional statistical characteristics to obtain land and a water area;

the determining module is used for extracting characteristic points of the contour of the water area and determining an interested area in the land as a suspected bridge area according to the distance of the characteristic points;

the rejecting module is used for rejecting false alarms in the suspected bridge area to realize bridge detection;

and the detection module is used for carrying out constant false alarm detection on the suspected bridge area after the false alarms are eliminated so as to distinguish the strong scatterer bridge.

6. The apparatus of claim 5, wherein the determining module extracts feature points of the contour of the water area through a digital curve splitting and merging algorithm.

7. The level set segmentation-based polarized SAR image bridge detection device of claim 5, wherein the culling module takes the region of interest that is not connected as the false alarm.

8. The apparatus according to claim 5, wherein the detection module is specifically configured to:

and taking the suspected bridge area after the false alarm is removed as a point target, representing the point target by an area average coherence matrix, and detecting the point target by a polarized whitening filter.