阅读说明：本技术 一种基于sar图像全局-局部特征的舰船目标检测方法 (Ship target detection method based on global-local features of SAR image ) 是由 李刚 王学谦 刘瑜 何友 于 2021-02-05 设计创作，主要内容包括：本发明提出一种基于SAR图像全局-局部特征的舰船目标检测方法,属于雷达图像处理领域。该方法首先对SAR图像建立对应的混合高斯模型,并对该图像进行超像素分割；然后分别计算每个超像素的费雪向量,以及每个费雪向量的全局密度对比度、全局密度距离对比度以及局部对比度；对全局密度对比度、全局密度距离对比度和局部对比度求积得到每个超像素的检验统计量；对每个检验统计量进行判定,最终得到SAR图像的舰船检测结果。本发明在SAR图像舰船检测中额外考虑了费雪向量的全局对比特征,有助于降低检测结果中的虚警率。(The invention provides a ship target detection method based on global-local features of an SAR image, and belongs to the field of radar image processing. Firstly, establishing a corresponding Gaussian mixture model for an SAR image, and performing superpixel segmentation on the image; then, respectively calculating the Fisher-snow vector of each super-pixel, and the global density contrast, the global density distance contrast and the local contrast of each Fisher-snow vector; performing product on the global density contrast, the global density distance contrast and the local contrast to obtain the test statistic of each super pixel; and judging each test statistic to finally obtain a ship detection result of the SAR image. The invention additionally considers the global contrast characteristic of the snow-wasting vector in SAR image ship detection, and is beneficial to reducing the false alarm rate in the detection result.)

1. A ship target detection method based on SAR image global-local characteristics is characterized in that the method comprises the steps of firstly obtaining an SAR image, establishing a corresponding Gaussian mixture model for the SAR image, and performing superpixel segmentation on the image; then, respectively calculating the Fisher-snow vector of each super-pixel, and the global density contrast, the global density distance contrast and the local contrast of each Fisher-snow vector; and (4) performing product calculation on the global density contrast, the global density distance contrast and the local contrast to obtain test statistics of each superpixel, judging each test statistic, and finally obtaining a ship detection result of the SAR image.

2. A method as claimed in claim 1, characterized in that the method comprises the following steps:

1) obtaining an SAR image, wherein the number of pixels of the image is N; setting the super pixel size S, the number of super pixels in the image is Represents rounding up; setting a regularization parameter lambda to be more than 0;

2) establishing a mixed Gaussian model GMM corresponding to the SAR image, wherein the expression is as follows:

wherein, ω is_q,μ_q,σ_qRespectively representing the weight, the mean value and the standard deviation of the qth Gaussian distribution in the GMM; q>0 is the order of the mixed Gaussian model; x represents the gray scale of a pixel in the image, f (x) represents the probability density function of the GMM,representing the qth gaussian distribution in the GMM;

3) super-pixel segmentation;

taking the size S of the super-pixel, a regularization parameter lambda and the SAR image as input, and obtaining all the super-pixels in the SAR image by using a simple linear iterative clustering SLIC algorithm;

4) calculating a Ferusse vector α for each superpixel_l：

Wherein the content of the first and second substances,

wherein L is the index of the super pixel, L is 1,2, …, L;representing the Ferusse vector alpha_lThe information of the zeroth order in (c),representing the Ferusse vector alpha_lThe first order information in (1) is,representing the Ferusse vector alpha_lL1, 2, …, L, Q1, 2, …, Q, P1, 2, …, P_l，P_lRepresenting the number of pixels, x, contained in the ith super-pixel_l,pRepresents the p-th pixel in the l-th super pixel;

β_qandfor intermediate variables, the expressions are as follows:

let alpha_l←sign(α_l)|α_l|^1/2And 2 norm normalization is carried out; sign () represents a sign function, and if the function input is positive, sign () is 1, otherwise-1;

5) calculating each of the snow vectorsGlobal density contrast k_l：

κ_l＝1-Nom(ρ_l) (6)

Where ρ is_lRepresenting the ferow vector density of the ith super pixel,denotes the l and the secondThe fern-vector distance of a super pixel, an index representing a super pixel other than the ith super pixel in the image; d_c> 0 for the truncation distance, nom (a) for a function normalizing a;

6) calculating the global density distance contrast psi for each Fisher-Tropsch vector_l：

Wherein the content of the first and second substances,a set of superpixels representing a lower Fisher vector density than the ith superpixel;

7) calculating local contrast of each Fisher-Tropsch vector

Where med { } denotes the median of the data in the selected set, G (l) denotes the neighborhood superpixel of the l-th superpixel, an index representing a neighborhood of the ith super pixel,andrespectively represent the l < th > and the l < th >Mean value of the gray level of the pixels inside the super pixels;

8) acquiring a ship detection result of the SAR image; the method comprises the following specific steps:

8-1) calculating a test statistic T for each superpixel_l：

8-2) calculating a decision threshold τ:

which is characterized in that the material is a mixture of,andrespectively represent { T_l1,2, …, L } and ξ is the scale factor;

8-3) separately for each test statistic T using_lAnd (4) judging:

all T_lAnd finally obtaining a ship detection result of the SAR image after the judgment is finished.

Technical Field

The invention belongs to the field of radar image processing, and particularly relates to a ship target detection method based on global-local features of an SAR image, which can be particularly used for ship detection in a synthetic aperture radar image.

Background

Synthetic Aperture Radar (SAR) is an active imaging device that can provide high resolution imaging results of sea surface vessel targets. Compared with optical and infrared sensors, SAR imaging is hardly affected by illumination and weather, and the SAR imaging sensor is a sensor with all-weather and all-time working capability. The detection of the ship target in the SAR image is a hotspot research problem in the fields of academia and national defense at present, and has important application in the aspects of military sea defense, civil ship monitoring, sustainable fishery and the like.

The Fisher vector is an image multi-order feature and contains zero-order, first-order and second-order information rich in superpixels in the SAR image. However, most of the existing SAR image Ship Detection methods based on the snow-wasting vector are based on Local comparison of the snow-wasting vector, wherein a method based on Local comparison of the snow-wasting vector is proposed in an article Ship Detection in SAR Images via Local Contrast of Fisher Vectors published in IEEE Transactions on Geoscience and Remote Sensing in 2020, and the method mainly comprises the following steps: firstly, establishing a corresponding Gaussian mixture model for an SAR image, and performing superpixel segmentation on the image; then, respectively calculating the Fisher-snow vector of each super pixel and the local contrast of each Fisher-snow vector, and taking the local contrast as test statistic; and judging each test statistic to finally obtain a ship detection result of the SAR image. Therefore, the conventional method lacks consideration of global snow-wasting vector comparison, so that more false alarm targets exist in a detection result, and the detection performance is reduced.

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provides a ship target detection method based on SAR image global-local characteristics. The invention additionally considers the global contrast characteristic of the snow-wasting vector in SAR image ship detection, and is beneficial to reducing the false alarm rate in the detection result.

The invention provides a ship target detection method based on SAR image global-local characteristics, which is characterized in that the method comprises the steps of firstly obtaining an SAR image, establishing a corresponding Gaussian mixture model for the SAR image, and performing superpixel segmentation on the image; then, respectively calculating the Fisher-snow vector of each super-pixel, and the global density contrast, the global density distance contrast and the local contrast of each Fisher-snow vector; and (4) performing product calculation on the global density contrast, the global density distance contrast and the local contrast to obtain test statistics of each superpixel, judging each test statistic, and finally obtaining a ship detection result of the SAR image. The method comprises the following steps:

1) obtaining an SAR image, wherein the number of pixels of the image is N; setting the super pixel size S, the number of super pixels in the image is Represents rounding up; setting a regularization parameter lambda to be more than 0;

2) establishing a mixed Gaussian model GMM corresponding to the SAR image, wherein the expression is as follows:

wherein, ω is_q,μ_q,σ_qRespectively representing the weight, the mean value and the standard deviation of the qth Gaussian distribution in the GMM; q>0 is the order of the mixed Gaussian model; x represents the gray scale of a pixel in the image, f (x) represents the probability density function of the GMM,denotes the qth high in GMMA distribution of Si;

3) super-pixel segmentation;

taking the size S of the super-pixel, a regularization parameter lambda and the SAR image as input, and obtaining all the super-pixels in the SAR image by using a simple linear iterative clustering SLIC algorithm;

4) calculating a Ferusse vector α for each superpixel_l：

Wherein the content of the first and second substances,

wherein L is the index of the super pixel, L is 1,2, …, L;representing the Ferusse vector alpha_lThe information of the zeroth order in (c),representing the Ferusse vector alpha_lThe first order information in (1) is,representing the Ferusse vector alpha_lL1, 2, …, L, Q1, 2, …, Q, P1, 2, …, P_l，P_lRepresenting the number of pixels, x, contained in the ith super-pixel_l,pRepresents the p-th pixel in the l-th super pixel;

β_qandfor intermediate variables, the expressions are as follows:

let alpha_l←sign(α_l)|α_l|^1/2And 2 norm normalization is carried out; sign () represents a sign function, and if the function input is positive, sign () is 1, otherwise-1;

5) calculating the global density contrast k for each Fisher-Tropsch vector_l：

κ_l＝1-Nom(ρ_l) (6)

Where ρ is_lRepresenting the ferow vector density of the ith super pixel,denotes the l and the secondThe fern-vector distance of a super pixel, an index representing a super pixel other than the ith super pixel in the image; d_c> 0 for the truncation distance, nom (a) for a function normalizing a;

6) calculating the global density distance contrast psi for each Fisher-Tropsch vector_l：

Wherein the content of the first and second substances,a set of superpixels representing a lower Fisher vector density than the ith superpixel;

7) calculating local contrast of each Fisher-Tropsch vector

Where med { } denotes the median of the data in the selected set, G (l) denotes the neighborhood superpixel of the l-th superpixel, an index representing a neighborhood of the ith super pixel,andrespectively represent the l < th > and the l < th >Mean value of the gray level of the pixels inside the super pixels;

8) acquiring a ship detection result of the SAR image; the method comprises the following specific steps:

8-1) calculating a test statistic T for each superpixel_l：

8-2) calculating a decision threshold τ:

which is characterized in that the material is a mixture of,andrespectively represent { T_l1,2, …, L } and ξ is the scale factor;

8-3) separately for each test statistic T using_lAnd (4) judging:

all T_lAnd finally obtaining a ship detection result of the SAR image after the judgment is finished.

The invention has the characteristics and beneficial effects that:

the existing SAR image ship detection method based on the Fisher-snow vector considers the local contrast characteristic of the Fisher-snow vector and does not consider the global contrast characteristic of the Fisher-snow vector, so that more false alarms in a detection result are easily caused. The SAR image ship detection method based on the Fisher-snow vector global and local contrast features is provided, the Fisher-snow vector global contrast features are further introduced, the global suppression effect on clutter is enhanced, and the reduction of the false alarm rate in the detection result is facilitated.

Drawings

FIG. 1 is an overall flow chart of the method of the present invention.

Detailed Description

The invention provides a ship target detection method based on global-local characteristics of an SAR image, which is further described in detail in the following by combining the attached drawings and specific embodiments.

The invention provides a ship target detection method based on global-local features of SAR images, the overall process is shown in figure 1, and the method comprises the following steps:

1) acquiring an SAR image, wherein the number of pixels of the image is N (N is 3000 multiplied by 3000 in the embodiment); setting a superpixel size S (S can be set to be 25% of the number of pixels occupied by ships, S is generally 10-100, and S is 30 in the embodiment), and then the number of superpixels in the image is Represents rounding up; setting a regularization parameter lambda to be more than 0 and used for controlling the spatial difference and the intensity difference of the superpixels, wherein lambda is generally 0.4-0.8; (in this example, λ is 0.4).

2) Modeling SAR images as containing a parameter { omega }_q,μ_q,σ_qQ is a Gaussian Mixture Model (GMM) of 1,2, …, Q, and the expression:

wherein, ω is_q,μ_q,σ_qRespectively representing the weight, mean and standard deviation of the qth gaussian distribution in the GMM. Q is the order of the mixed Gaussian model, and is generally 7-10; x tableShowing the gray scale of the pixels in the image, f (x) representing the probability density function of GMM,representing the qth gaussian distribution in the GMM.

In the present invention, the parameter { omega }_q,μ_q,σ_qQ ═ 1,2, …, Q } can be obtained based on The expectation-maximization (EM) algorithm published by t.k.

3) Super-pixel segmentation;

taking the size S of the super-pixel, the regularization parameter lambda and the SAR image as input, and obtaining all the super-pixels in the SAR image according to a Simple Linear Iterative Clustering (SLIC) algorithm in the Ship Detection With super-Level Fisher Vector in High-Resolution SAR Images published by the IEEE Geoscience and Remote Sensing rules.

4) Calculating a Ferusse vector α for each superpixel_l：

Wherein the content of the first and second substances,

wherein L is the index of the superpixel, L is 1,2, …, and L is the number of the superpixels in the SAR image;representing the Ferusse vector alpha_lThe information of the zeroth order in (c),representing the Ferusse vector alpha_lThe first order information in (1) is,representing the Ferusse vector alpha_lL1, 2, …, L, Q1, 2, …, Q, P1, 2, …, P_l，P_lRepresenting the number of pixels, x, contained in the ith super-pixel_l,pRepresenting the p-th pixel in the l-th super-pixel.

Symbol β in formulae (3) to (5)_qAndtemporary symbols for convenience, the expressions are respectively as follows:

let alpha_l←sign(α_l)|α_l|^1/2And 2 norm normalization is carried out; sign () represents a sign function, which is 1 if the function input is positive, and-1 otherwise.

5) Calculating the global density contrast k for each Fisher-Tropsch vector_l；

κ_l＝1-Nom(ρ_l) (6)

Where ρ is_lRepresenting the ferow vector density of the ith super pixel,denotes the l and the secondThe fern-vector distance of a super pixel, an index representing a super pixel other than the ith super pixel in the image; d_c> 0 denotes the truncation distance, which is a constant, D_cCan be generally arranged as30% of the medium maximum, nom (a) represents a function normalizing a.

6) Calculating the global density distance contrast psi for each Fisher-Tropsch vector_l：

Wherein the content of the first and second substances,representing a set of superpixels having a lower ficoll vector density than the current ith superpixel.

7) Calculating local contrast of each Fisher-Tropsch vector

Where med { } denotes the median of the data in the selected set, G (l) denotes the neighborhood superpixel of the l-th superpixel, an index representing a neighborhood of the ith super pixel,andrespectively represent the l < th > and the l < th >Mean value of the pixel gray levels inside the super-pixels.

8) Acquiring a ship detection result of the SAR image; the method comprises the following specific steps:

8-1) calculating a test statistic T for each superpixel_l：

Wherein the test statistic T_lIs the global density contrast k_lGlobal density distance contrast psi_lLocal contrast ratioThe product of (a).

8-2) calculating a decision threshold τ:

it is composed ofAndrespectively represent { T_lWhere L is the mean and standard deviation of 1,2, …, L, and ξ is the input scale factor, it may be generally ∈ 3,14]In this embodiment, ξ ═ 5.

8-3) separately for each test statistic T using_lAnd (4) judging:

all T_lAnd finally obtaining a ship detection result of the SAR image after the judgment is finished.