CN112766287A - SAR image ship target detection acceleration method based on density examination - Google Patents

SAR image ship target detection acceleration method based on density examination Download PDF

Info

Publication number
CN112766287A
CN112766287A CN202110161176.5A CN202110161176A CN112766287A CN 112766287 A CN112766287 A CN 112766287A CN 202110161176 A CN202110161176 A CN 202110161176A CN 112766287 A CN112766287 A CN 112766287A
Authority
CN
China
Prior art keywords
superpixel
density
superpixels
clutter
cluster center
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110161176.5A
Other languages
Chinese (zh)
Other versions
CN112766287B (en
Inventor
李刚
王学谦
刘瑜
何友
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.)
Tsinghua University
Naval Aeronautical University
Original Assignee
Tsinghua University
Naval Aeronautical University
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 Tsinghua University, Naval Aeronautical University filed Critical Tsinghua University
Priority to CN202110161176.5A priority Critical patent/CN112766287B/en
Publication of CN112766287A publication Critical patent/CN112766287A/en
Application granted granted Critical
Publication of CN112766287B publication Critical patent/CN112766287B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G01S13/9021—SAR image post-processing techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Remote Sensing (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Electromagnetism (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

本发明提出一种基于密度审查的SAR图像舰船目标检测加速方法，属于合成孔径雷达图像处理领域。该方法首先对SAR图像进行超像素分割；分别计算每个超像素的密度特征和密度距离特征，筛选出该SAR图像作为目标聚类中心的超像素和作为杂波聚类中心的超像素；利用最近邻分类器比较每个超像素与杂波聚类中心的差异以及该超像素与目标聚类中心的差异，删除与目标聚类中心差异更大的超像素，将最终保留的超像素作为之后舰船目标检测方法的输入。本发明通过在检测之前先行快速删除大量海杂波超像素，将保留的超像素作为舰船目标检测方法的输入并进行精细检测，显著提升现有超像素检测方法的计算速度，提高了SAR图像中舰船目标检测的运行效率。

Figure 202110161176

The invention proposes an acceleration method for detecting a ship target in a SAR image based on density review, which belongs to the field of synthetic aperture radar image processing. The method firstly performs superpixel segmentation on the SAR image; calculates the density feature and density distance feature of each superpixel respectively, and selects the SAR image as the superpixel as the target clustering center and the superpixel as the clutter clustering center; The nearest neighbor classifier compares the difference between each superpixel and the clutter cluster center as well as the difference between the superpixel and the target cluster center, deletes the superpixel that is more different from the target cluster center, and uses the final retained superpixel as the later Input to the ship target detection method. The invention quickly deletes a large number of sea clutter superpixels before detection, uses the reserved superpixels as the input of the ship target detection method and performs fine detection, thereby significantly improving the calculation speed of the existing superpixel detection methods and improving the SAR image. Operational efficiency of ship target detection in China.

Figure 202110161176

Description

一种基于密度审查的SAR图像舰船目标检测加速方法An acceleration method for ship target detection in SAR images based on density review

技术领域technical field

本发明属于合成孔径雷达(SAR)图像处理领域，具体涉及一种基于密度审查的SAR图像舰船目标检测加速方法，可用于现有超像素舰船目标检测方法的快速实现。The invention belongs to the field of Synthetic Aperture Radar (SAR) image processing, and in particular relates to a density review-based SAR image ship target detection acceleration method, which can be used for rapid realization of the existing superpixel ship target detection method.

背景技术Background technique

合成孔径雷达(SAR)是一种主动成像装置，可提供海面舰船目标的高分辨成像结果。相比于光学、红外等传感器，SAR成像几乎不受光照和天气的影响，是一种具备全天候、全时段工作能力的传感器。SAR图像中的舰船目标检测在军事海防、民船监视、可持续渔业等方面有着重要的应用。Synthetic Aperture Radar (SAR) is an active imaging device that provides high-resolution imaging of surface ship targets. Compared with optical and infrared sensors, SAR imaging is almost unaffected by light and weather, and is a sensor capable of all-weather and all-time work. Ship target detection in SAR images has important applications in military coastal defense, civilian ship surveillance, and sustainable fisheries.

近年来，众多专家学者提出了多种超像素检测方法，在SAR图像舰船目标检测中取得了良好的效果。然而，现有的超像素检测方法需要滑窗精细处理SAR图像中的每一个超像素，运行速度很慢，使得检测过程的计算效率大大降低。In recent years, many experts and scholars have proposed a variety of superpixel detection methods, which have achieved good results in ship target detection in SAR images. However, the existing superpixel detection methods require a sliding window to process each superpixel in the SAR image, and the running speed is very slow, which greatly reduces the computational efficiency of the detection process.

发明内容SUMMARY OF THE INVENTION

本发明的目的是为克服已有技术的不足之处，提出一种基于密度审查的SAR图像舰船目标检测加速方法。本发明通过在检测之前先行快速删除大量海杂波超像素，将保留的超像素作为之后舰船目标检测方法的输入，对保留的超像素进行精细检测，显著提升了现有超像素检测方法的计算速度，提高了SAR图像中的舰船目标检测的运行效率。The purpose of the present invention is to overcome the shortcomings of the prior art and propose a method for accelerating detection of ship targets in SAR images based on density review. The invention quickly deletes a large number of sea clutter superpixels before detection, uses the retained superpixels as the input of the subsequent ship target detection method, and performs fine detection on the retained superpixels, thereby significantly improving the performance of the existing superpixel detection methods. The calculation speed improves the operational efficiency of ship target detection in SAR images.

本发明提出一种基于密度审查的SAR图像舰船目标检测加速方法，其特征在于，该方法首先对SAR图像进行超像素分割；分别计算每个超像素的密度特征和密度距离特征，筛选出该SAR图像作为目标聚类中心的超像素和作为杂波聚类中心的超像素；利用最近邻分类器比较每个超像素与杂波聚类中心的差异以及该超像素与目标聚类中心的差异，删除与目标聚类中心差异更大的超像素，将最终保留的超像素作为之后舰船目标检测方法的输入。该方法包括以下步骤：The present invention proposes a method for speeding up ship target detection in SAR images based on density review. The method is characterized in that, the method firstly performs superpixel segmentation on SAR images; respectively calculates the density feature and density distance feature of each superpixel, and selects the SAR images as superpixels as target cluster centers and as superpixels as clutter cluster centers; use nearest neighbor classifier to compare the difference between each superpixel and the clutter cluster center and the difference between the superpixel and the target cluster center , delete the superpixels that are more different from the target cluster center, and use the final retained superpixels as the input of the ship target detection method. The method includes the following steps:

1)获取一张SAR图像，该图像像素数量为N；设置超像素尺寸S,则该图像中超像素个数为

Figure BDA0002936728400000011
表示向上取整；设置正则化参数λ＞0；1) Obtain a SAR image, the number of pixels in the image is N; set the superpixel size S, then the number of superpixels in the image is
Figure BDA0002936728400000011
Indicates rounding up; set the regularization parameter λ>0;

2)超像素分割；2) Superpixel segmentation;

将超像素的尺寸S、正则化参数λ、以及SAR图像作为输入，利用简单线性迭代聚类SLIC算法获得该SAR图像中的所有超像素；Taking the size S of the superpixel, the regularization parameter λ, and the SAR image as input, use the simple linear iterative clustering SLIC algorithm to obtain all the superpixels in the SAR image;

3)计算每个超像素的密度特征ρi：3) Calculate the density feature ρ i of each superpixel:

Figure BDA0002936728400000021
Figure BDA0002936728400000021

其中，i＝1,2,…,I，i为超像素的索引；j表示图像中除了第i个像素外其他像素的索引，Di,j＝|μi-μj|表示第i个超像素的灰度均值μi和第j个超像素灰度均值μj的绝对值差异，μ表示超像素灰度均值，

Figure BDA0002936728400000022
表示软截断距离，α是输入的尺度因子，α∈(0,1)；Among them, i = 1, 2, . is the absolute value difference between the gray mean value μ i of the superpixel and the jth superpixel gray mean value μ j , where μ represents the gray mean value of the superpixel,
Figure BDA0002936728400000022
represents the soft truncation distance, α is the scale factor of the input, α∈(0,1);

4)计算每个超像素的密度距离特征ri：4) Calculate the density distance feature ri for each superpixel :

Figure BDA0002936728400000023
Figure BDA0002936728400000023

其中，Γi＝{j|ρj＜ρi,j＝1,2,...,I,j≠i}表示比第i个超像素密度低的超像素的集合，

Figure BDA0002936728400000024
表示空集；where Γ i ={j|ρ j ＜ρ i ,j=1,2,...,I,j≠i} represents the set of superpixels with a lower density than the ith superpixel,
Figure BDA0002936728400000024
represents the empty set;

5)分别计算SAR图像的目标聚类中心与杂波聚类中心；5) Calculate the target clustering center and the clutter clustering center of the SAR image respectively;

首先将各密度特征ρi和各密度距离特征ri分别进行归一化到区间[0,1]，得到归一化的密度特征

Figure BDA0002936728400000025
和归一化的密度距离特征
Figure BDA0002936728400000026
First, each density feature ρ i and each density distance feature ri are normalized to the interval [0, 1], and the normalized density feature is obtained.
Figure BDA0002936728400000025
and the normalized density distance feature
Figure BDA0002936728400000026

再利用归一化的密度特征

Figure BDA0002936728400000027
和归一化的密度距离特征
Figure BDA0002936728400000028
寻找该图像目标聚类中心i目标和杂波的聚类中心i杂波：Reuse normalized density features
Figure BDA0002936728400000027
and the normalized density distance feature
Figure BDA0002936728400000028
Find the image target cluster center i target and the cluster center i clutter of the clutter :

Figure BDA0002936728400000029
Figure BDA0002936728400000029

Figure BDA00029367284000000210
Figure BDA00029367284000000210

其中，C目标表示所有超像素中作为目标聚类中心的超像素索引，C杂波表示所有超像素中作为杂波聚类中心的超像素索引；Among them, C target represents the superpixel index as the target cluster center in all superpixels, and Cclutter represents the superpixel index as the clutter cluster center in all superpixels;

6)利用最近邻分类器删除SAR图像中的杂波超像素；6) Use the nearest neighbor classifier to remove clutter superpixels in SAR images;

对每个超像素，判定如下：For each superpixel, the decision is as follows:

Figure BDA0002936728400000031
Figure BDA0002936728400000031

其中，

Figure BDA0002936728400000032
表示第i个超像素与杂波聚类中心的差异；
Figure BDA0002936728400000033
表示当前超像素与目标聚类中心的差异；
Figure BDA0002936728400000034
表示图像中作为杂波聚类中心的超像素的密度特征值，
Figure BDA0002936728400000035
表示图像中作为目标聚类中心的超像素的密度特征值，
Figure BDA0002936728400000036
表示图像中作为杂波聚类中心的超像素的密度距离特征值，
Figure BDA0002936728400000037
表示图像中作为目标聚类中心的超像素的密度距离特征值；in,
Figure BDA0002936728400000032
represents the difference between the i-th superpixel and the clutter cluster center;
Figure BDA0002936728400000033
Represents the difference between the current superpixel and the target cluster center;
Figure BDA0002936728400000034
represents the density eigenvalues of the superpixels in the image that are the centers of clutter clusters,
Figure BDA0002936728400000035
represents the density feature value of the superpixel in the image as the target cluster center,
Figure BDA0002936728400000036
represents the density distance eigenvalues of the superpixels in the image that are the centers of clutter clusters,
Figure BDA0002936728400000037
Represents the density distance feature value of the superpixel in the image as the target cluster center;

7)输出SAR图像中经过步骤6)后保留的超像素。7) Output the superpixels retained after step 6) in the output SAR image.

本发明的特点及有益效果：Features and beneficial effects of the present invention:

现有的SAR图像超像素舰船目标检测方法需要精细检测图像中的每一个超像素，而精细检测每一个超像素往往需要矩阵求逆、参数迭代估计等复杂运算，导致超像素检测方法的计算运行效率较低。本发明基于SAR图像的密度特征，提出了一种基于密度审查的SAR图像舰船目标检测加速方法，在精细检测之前先快速滤除图像中大量的杂波超像素，将保留的超像素作为之后舰船目标检测方法的输入，对保留的超像素进行精细检测，显著加快了检测的运行速度，有望提升我国对舰船目标的快速响应能力。Existing SAR image superpixel ship target detection methods need to finely detect each superpixel in the image, and fine detection of each superpixel often requires complex operations such as matrix inversion, parameter iterative estimation, etc., which leads to the calculation of superpixel detection methods. Operational efficiency is low. Based on the density characteristics of SAR images, the present invention proposes a method for accelerating ship target detection in SAR images based on density review. The input of the ship target detection method, the fine detection of the reserved superpixels, significantly speeds up the detection speed, and is expected to improve my country's rapid response capability to ship targets.

附图说明Description of drawings

图1为本发明方法的整体流程图。Fig. 1 is the overall flow chart of the method of the present invention.

具体实施方式Detailed ways

本发明提出一种基于密度审查的SAR图像舰船目标检测加速方法，下面结合附图和具体实施例进一步详细说明如下。The present invention proposes a method for speeding up ship target detection in SAR images based on density review, which is further described in detail below with reference to the accompanying drawings and specific embodiments.

本发明提出一种基于密度审查的SAR图像舰船目标检测加速方法，整体流程如图1所示，包括以下步骤：The present invention proposes a method for speeding up ship target detection in SAR images based on density review. The overall process is shown in Figure 1, including the following steps:

1)获取一张SAR图像，该图像像素数量为N(例如N＝3000×3000)；设置超像素尺寸S(S可设置为舰船所占像素数量的25％，S一般取10～100，本实施例取30),则该图像中超像素个数为

Figure BDA0002936728400000041
表示向上取整；设置正则化参数λ＞0，用于超像素分割(本实施例取参数λ＝0.4)。1) Obtain a SAR image, the number of pixels in the image is N (for example, N=3000×3000); set the superpixel size S (S can be set to 25% of the number of pixels occupied by the ship, and S generally takes 10 to 100, This embodiment takes 30), then the number of superpixels in the image is
Figure BDA0002936728400000041
Indicates rounding up; set the regularization parameter λ>0, which is used for superpixel segmentation (parameter λ=0.4 in this embodiment).

2)超像素分割；2) Superpixel segmentation;

将超像素的尺寸S、正则化参数λ、以及SAR图像作为输入，根据林慧平发表在IEEEGeoscience and Remote Sensing Letters上的Ship Detection With Superpixel-LevelFisher Vector in High-Resolution SAR Images中的简单线性迭代聚类(SLIC)算法获得该SAR图像中的所有超像素。Taking the size S of the superpixel, the regularization parameter λ, and the SAR image as input, simple linear iterative clustering ( SLIC) algorithm to obtain all superpixels in this SAR image.

3)计算每个超像素的密度特征ρi；3) Calculate the density feature ρ i of each superpixel;

Figure BDA0002936728400000042
Figure BDA0002936728400000042

其中，i＝1,2,…,I，i为超像素的索引，I为SAR图像中超像素的数量；j表示图像中除了第i个像素外其他像素的索引，Di,j＝|μi-μj|表示第i个超像素的灰度均值μi和第j个超像素灰度均值μj的绝对值差异，μ表示超像素灰度均值，

Figure BDA0002936728400000043
表示软截断距离，是一个常数，α是输入的尺度因子，尺度因子α∈(0,1)，用于计算密度特征，这里取α＝0.3。Among them, i=1,2,...,I, i is the index of the superpixel, I is the number of superpixels in the SAR image; j is the index of other pixels except the i-th pixel in the image, D i,j =|μ i - μ j | represents the absolute difference between the gray mean value μ i of the ith superpixel and the gray mean value μ j of the j th superpixel, and μ represents the gray mean value of the superpixel,
Figure BDA0002936728400000043
Represents the soft truncation distance, which is a constant, α is the input scale factor, and the scale factor α∈(0,1) is used to calculate the density feature, where α=0.3.

4)计算每个超像素的密度距离特征ri：4) Calculate the density distance feature ri for each superpixel :

Figure BDA0002936728400000044
Figure BDA0002936728400000044

其中，Γi＝{j|ρj＜ρi,j＝1,2,...,I,j≠i}表示比第i个超像素密度低的超像素的集合。

Figure BDA0002936728400000045
表示空集。Here, Γ i ={j|ρ j <ρ i , j=1,2,...,I,j≠i} represents a set of superpixels with a lower density than the i-th superpixel.
Figure BDA0002936728400000045
represents the empty set.

5)分别计算SAR图像的目标聚类中心与杂波聚类中心。5) Calculate the target clustering center and the clutter clustering center of the SAR image respectively.

首先将各密度特征ρi和各密度距离特征ri分别进行归一化到区间[0,1]，得到归一化的密度特征

Figure BDA0002936728400000046
和归一化的密度距离特征
Figure BDA0002936728400000047
归一化的方式可采用以下方式：First, each density feature ρ i and each density distance feature ri are normalized to the interval [0, 1], and the normalized density feature is obtained.
Figure BDA0002936728400000046
and the normalized density distance feature
Figure BDA0002936728400000047
Normalization can be done in the following ways:

Figure BDA0002936728400000048
Figure BDA0002936728400000048

再利用归一化的密度特征

Figure BDA0002936728400000051
和归一化的密度距离特征
Figure BDA0002936728400000052
寻找该图像目标聚类中心i目标和杂波的聚类中心i杂波：Reuse normalized density features
Figure BDA0002936728400000051
and the normalized density distance feature
Figure BDA0002936728400000052
Find the image target cluster center i target and the cluster center i clutter of the clutter :

Figure BDA0002936728400000053
Figure BDA0002936728400000053

Figure BDA0002936728400000054
Figure BDA0002936728400000054

其中，C目标表示所有超像素中作为目标聚类中心的超像素索引，C杂波表示所有超像素中作为杂波聚类中心的超像素索引。Among them, C target represents the superpixel index of all superpixels as the target cluster center, and Cclutter represents the superpixel index of all superpixels as the clutter cluster center.

6)利用最近邻分类器删除SAR图像中的杂波超像素；6) Use the nearest neighbor classifier to remove clutter superpixels in SAR images;

对每个超像素，判定如下：For each superpixel, the decision is as follows:

Figure BDA0002936728400000055
Figure BDA0002936728400000055

其中

Figure BDA0002936728400000056
表示第i个超像素与杂波聚类中心的差异，
Figure BDA0002936728400000057
表示当前超像素与目标聚类中心的差异。
Figure BDA0002936728400000058
表示图像中作为杂波聚类中心的超像素的密度特征值，
Figure BDA0002936728400000059
表示图像中作为目标聚类中心的超像素的密度特征值，
Figure BDA00029367284000000510
表示图像中作为杂波聚类中心的超像素的密度距离特征值，
Figure BDA00029367284000000511
表示图像中作为目标聚类中心的超像素的密度距离特征值。in
Figure BDA0002936728400000056
represents the difference between the ith superpixel and the clutter cluster center,
Figure BDA0002936728400000057
Represents the difference between the current superpixel and the target cluster center.
Figure BDA0002936728400000058
represents the density eigenvalues of the superpixels in the image that are the centers of clutter clusters,
Figure BDA0002936728400000059
represents the density feature value of the superpixel in the image as the target cluster center,
Figure BDA00029367284000000510
represents the density distance eigenvalues of the superpixels in the image that are the centers of clutter clusters,
Figure BDA00029367284000000511
Represents the density distance eigenvalues of the superpixels in the image that are the centers of the target clusters.

7)输出SAR图像中经过步骤6)后保留的超像素(视为目标超像素的候选)。7) Output the superpixels retained after step 6) in the output SAR image (as candidates for the target superpixels).

本发明中，由于式(5)删除了大量的杂波超像素，后续的超像素检测方法的运行时间得到显著减少。In the present invention, since a large number of clutter superpixels are removed by Equation (5), the running time of the subsequent superpixel detection method is significantly reduced.

步骤7)中输出的保留的超像素可直接作为现有的SAR图像超像素舰船目标检测方法的输入，即，现有的SAR图像超像素舰船目标检测方法无需对SAR图像中所有的超像素进行精细检测，仅需要对少量的步骤7)中输出的保留超像素进行精细检测，大大降低了完成检测所需的时间，提升了计算运行效率。The reserved superpixels output in step 7) can be directly used as the input of the existing SAR image superpixel ship target detection method, that is, the existing SAR image superpixel ship target detection method does not need to detect all the superpixels in the SAR image. For fine detection of pixels, only a small number of reserved superpixels output in step 7) need to be finely detected, which greatly reduces the time required to complete the detection and improves the computational efficiency.

Claims (3)

1.一种基于密度审查的SAR图像舰船目标检测加速方法，其特征在于，该方法首先对SAR图像进行超像素分割；分别计算每个超像素的密度特征和密度距离特征，筛选出该SAR图像作为目标聚类中心的超像素和作为杂波聚类中心的超像素；利用最近邻分类器比较每个超像素与杂波聚类中心的差异以及该超像素与目标聚类中心的差异，删除与目标聚类中心差异更大的超像素，将最终保留的超像素作为之后舰船目标检测方法的输入。1. a SAR image ship target detection acceleration method based on density review, is characterized in that, this method first carries out superpixel segmentation to SAR image; The image is used as the superpixel of the target cluster center and the superpixel as the clutter cluster center; the nearest neighbor classifier is used to compare the difference between each superpixel and the clutter cluster center and the difference between the superpixel and the target cluster center, The superpixels that are more different from the target cluster center are deleted, and the final retained superpixels are used as the input of the ship target detection method. 2.如权利要求1所述的方法，其特征在于，该方法包括以下步骤：2. The method of claim 1, wherein the method comprises the steps of: 1)获取一张SAR图像，该图像像素数量为N；设置超像素尺寸S,则该图像中超像素个数为
Figure FDA0002936728390000011
Figure FDA0002936728390000012
表示向上取整；设置正则化参数λ＞0；1) Obtain a SAR image, the number of pixels in the image is N; set the superpixel size S, then the number of superpixels in the image is
Figure FDA0002936728390000011
Figure FDA0002936728390000012
Indicates rounding up; set the regularization parameter λ>0; 2)超像素分割；2) Superpixel segmentation; 将超像素的尺寸S、正则化参数λ、以及SAR图像作为输入，利用简单线性迭代聚类SLIC算法获得该SAR图像中的所有超像素；Taking the size S of the superpixel, the regularization parameter λ, and the SAR image as input, use the simple linear iterative clustering SLIC algorithm to obtain all the superpixels in the SAR image; 3)计算每个超像素的密度特征ρi：3) Calculate the density feature ρ i of each superpixel:
Figure FDA0002936728390000013
Figure FDA0002936728390000013
 其中，i＝1,2,…,I，i为超像素的索引；j表示图像中除了第i个像素外其他像素的索引，Di,j＝|μi-μj|表示第i个超像素的灰度均值μi和第j个超像素灰度均值μj的绝对值差异，μ表示超像素灰度均值，
Figure FDA0002936728390000014
表示软截断距离，α是输入的尺度因子，α∈(0,1)；Among them, i = 1, 2, . is the absolute value difference between the gray mean value μ i of the superpixel and the jth superpixel gray mean value μ j , where μ represents the gray mean value of the superpixel,
Figure FDA0002936728390000014
represents the soft truncation distance, α is the scale factor of the input, α∈(0,1); 4)计算每个超像素的密度距离特征ri：4) Calculate the density distance feature ri for each superpixel :
Figure FDA0002936728390000015
Figure FDA0002936728390000015
 其中，Γi＝{j|ρj＜ρi,j＝1,2,...,I,j≠i}表示比第i个超像素密度低的超像素的集合，
Figure FDA0002936728390000016
表示空集；where Γ i ={j|ρ j ＜ρ i ,j=1,2,...,I,j≠i} represents the set of superpixels with a lower density than the ith superpixel,
Figure FDA0002936728390000016
represents the empty set; 5)分别计算SAR图像的目标聚类中心与杂波聚类中心；5) Calculate the target clustering center and the clutter clustering center of the SAR image respectively; 首先将各密度特征ρi和各密度距离特征ri分别进行归一化到区间[0,1]，得到归一化的密度特征
Figure FDA0002936728390000021
和归一化的密度距离特征
Figure FDA0002936728390000022
First, each density feature ρ i and each density distance feature ri are normalized to the interval [0, 1], and the normalized density feature is obtained.
Figure FDA0002936728390000021
and the normalized density distance feature
Figure FDA0002936728390000022
 再利用归一化的密度特征
Figure FDA0002936728390000023
和归一化的密度距离特征
Figure FDA0002936728390000024
寻找该图像目标聚类中心i目标和杂波的聚类中心i杂波：Reuse normalized density features
Figure FDA0002936728390000023
and the normalized density distance feature
Figure FDA0002936728390000024
Find the image target cluster center i target and the cluster center i clutter of the clutter :
Figure FDA0002936728390000025
Figure FDA0002936728390000025
Figure FDA0002936728390000026
Figure FDA0002936728390000026
 其中，C目标表示所有超像素中作为目标聚类中心的超像素索引，C杂波表示所有超像素中作为杂波聚类中心的超像素索引；Among them, C target represents the superpixel index as the target cluster center in all superpixels, and Cclutter represents the superpixel index as the clutter cluster center in all superpixels; 6)利用最近邻分类器删除SAR图像中的杂波超像素；6) Use the nearest neighbor classifier to remove clutter superpixels in SAR images; 对每个超像素，判定如下：For each superpixel, the decision is as follows:
Figure FDA0002936728390000027
Figure FDA0002936728390000027
 其中，
Figure FDA0002936728390000028
表示第i个超像素与杂波聚类中心的差异；
Figure FDA0002936728390000029
表示当前超像素与目标聚类中心的差异；
Figure FDA00029367283900000210
表示图像中作为杂波聚类中心的超像素的密度特征值，
Figure FDA00029367283900000211
表示图像中作为目标聚类中心的超像素的密度特征值，
Figure FDA00029367283900000212
表示图像中作为杂波聚类中心的超像素的密度距离特征值，
Figure FDA00029367283900000213
表示图像中作为目标聚类中心的超像素的密度距离特征值；in,
Figure FDA0002936728390000028
represents the difference between the i-th superpixel and the clutter cluster center;
Figure FDA0002936728390000029
Represents the difference between the current superpixel and the target cluster center;
Figure FDA00029367283900000210
represents the density eigenvalues of the superpixels in the image that are the centers of clutter clusters,
Figure FDA00029367283900000211
represents the density feature value of the superpixel in the image as the target cluster center,
Figure FDA00029367283900000212
represents the density distance eigenvalues of the superpixels in the image that are the centers of clutter clusters,
Figure FDA00029367283900000213
Represents the density distance feature value of the superpixel in the image as the target cluster center; 7)输出SAR图像中经过步骤6)后保留的超像素。7) Output the superpixels retained after step 6) in the output SAR image. 3.如权利要求2所述的方法，其特征在于，所述步骤5)中归一化方法如下：3. method as claimed in claim 2, is characterized in that, in described step 5), normalization method is as follows:
Figure FDA00029367283900000214
Figure FDA00029367283900000214
CN202110161176.5A 2021-02-05 2021-02-05 SAR image ship target detection acceleration method based on density examination Active CN112766287B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110161176.5A CN112766287B (en) 2021-02-05 2021-02-05 SAR image ship target detection acceleration method based on density examination

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110161176.5A CN112766287B (en) 2021-02-05 2021-02-05 SAR image ship target detection acceleration method based on density examination

Publications (2)

Publication Number Publication Date
CN112766287A true CN112766287A (en) 2021-05-07
CN112766287B CN112766287B (en) 2021-09-17

Family

ID=75705159

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110161176.5A Active CN112766287B (en) 2021-02-05 2021-02-05 SAR image ship target detection acceleration method based on density examination

Country Status (1)

Country Link
CN (1) CN112766287B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9389311B1 (en) * 2015-02-19 2016-07-12 Sandia Corporation Superpixel edges for boundary detection
CN107067039A (en) * 2017-04-25 2017-08-18 西安电子科技大学 SAR image Ship Target quick determination method based on super-pixel
CN108629783A (en) * 2018-05-02 2018-10-09 山东师范大学 Image partition method, system and medium based on the search of characteristics of image density peaks
CN108765491A (en) * 2018-05-31 2018-11-06 成都信息工程大学 A kind of SAR image Ship Target Detection method
CN109886218A (en) * 2019-02-26 2019-06-14 西安电子科技大学 SAR image ship target detection method based on statistical dissimilarity of superpixels

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9389311B1 (en) * 2015-02-19 2016-07-12 Sandia Corporation Superpixel edges for boundary detection
CN107067039A (en) * 2017-04-25 2017-08-18 西安电子科技大学 SAR image Ship Target quick determination method based on super-pixel
CN108629783A (en) * 2018-05-02 2018-10-09 山东师范大学 Image partition method, system and medium based on the search of characteristics of image density peaks
CN108765491A (en) * 2018-05-31 2018-11-06 成都信息工程大学 A kind of SAR image Ship Target Detection method
CN109886218A (en) * 2019-02-26 2019-06-14 西安电子科技大学 SAR image ship target detection method based on statistical dissimilarity of superpixels

Also Published As

Publication number Publication date
CN112766287B (en) 2021-09-17

Similar Documents

Publication Publication Date Title
CN110210463B (en) Radar target image detection method based on Precise ROI-Faster R-CNN
CN110378308B (en) Improved method for nearshore ship detection in port SAR images based on Faster R-CNN
Zhang et al. Pedestrian detection method based on Faster R-CNN
CN108898065B (en) A deep network ship target detection method based on rapid screening of candidate regions and scale adaptation
CN108171193B (en) Polarized SAR (synthetic aperture radar) ship target detection method based on super-pixel local information measurement
CN105930790B (en) Human behavior recognition method based on kernel sparse coding
CN114067118B (en) Processing method of aerial photogrammetry data
Huang et al. Survey of target detection algorithms in SAR images
Li et al. Evaluation the performance of fully convolutional networks for building extraction compared with shallow models
CN116363535A (en) Ship detection method in unmanned aerial vehicle aerial image based on convolutional neural network
CN116645608A (en) Remote sensing target detection based on Yolox-Tiny biased feature fusion network
CN112766286A (en) Ship target detection method based on global-local features of SAR image
CN110837769B (en) Image processing and deep learning embedded far infrared pedestrian detection method
Zhang et al. Infrared detection of small moving target using spatial–temporal local vector difference measure
Xili et al. Feature attention pyramid-based remote sensing image object detection method
CN113014923B (en) Behavior identification method based on compressed domain representation motion vector
Sun et al. Pig detection algorithm based on sliding windows and PCA convolution
CN112766287B (en) SAR image ship target detection acceleration method based on density examination
Yang et al. Sar images target detection based on yolov5
CN102156879A (en) Human target matching method based on weighted terrestrial motion distance
CN111950476A (en) Automatic identification method of river vessels in complex environment based on deep learning
Sun et al. A fast multi-target detection method based on improved YOLO
CN114758135B (en) An unsupervised image semantic segmentation method based on attention mechanism
Qiwei et al. SAR image ship detection based on improved YOLOv4
Chen et al. SAR ship detection based on improved YOLOv3

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