CN107808383B - Rapid detection method for SAR image target under strong sea clutter - Google Patents

Rapid detection method for SAR image target under strong sea clutter Download PDF

Info

Publication number
CN107808383B
CN107808383B CN201710952653.3A CN201710952653A CN107808383B CN 107808383 B CN107808383 B CN 107808383B CN 201710952653 A CN201710952653 A CN 201710952653A CN 107808383 B CN107808383 B CN 107808383B
Authority
CN
China
Prior art keywords
target
sub
block
image
sar image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201710952653.3A
Other languages
Chinese (zh)
Other versions
CN107808383A (en
Inventor
顾丹丹
张元�
梁子长
鲁炳坦
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.)
Shanghai Radio Equipment Research Institute
Original Assignee
Shanghai Radio Equipment Research Institute
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Radio Equipment Research Institute filed Critical Shanghai Radio Equipment Research Institute
Priority to CN201710952653.3A priority Critical patent/CN107808383B/en
Publication of CN107808383A publication Critical patent/CN107808383A/en
Application granted granted Critical
Publication of CN107808383B publication Critical patent/CN107808383B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • 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
    • G01S13/9027—Pattern recognition for feature extraction
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G01S13/9004—SAR image acquisition techniques
    • G01S13/9011—SAR image acquisition techniques with frequency domain processing of the SAR signals in azimuth
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/414—Discriminating targets with respect to background clutter
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10032—Satellite or aerial image; Remote sensing
    • G06T2207/10044—Radar image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20092—Interactive image processing based on input by user
    • G06T2207/20104—Interactive definition of region of interest [ROI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30181—Earth observation

Landscapes

  • Engineering & Computer Science (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Theoretical Computer Science (AREA)
  • Electromagnetism (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

A method for quickly detecting SAR image targets under strong sea clutter comprises the following steps: s1, dividing the large scene SAR image into a plurality of sub-block SAR images by adopting a region blocking strategy; s2, acquiring a sub-block coherent image with enhanced target-clutter contrast corresponding to the sub-block SAR image based on sub-aperture coherent processing; s3, screening out a target subblock coherent image containing a target and a corresponding target subblock SAR image as candidate target subblock images based on the significance of the subblock coherent image; s4, carrying out fine target detection on the candidate target sub-block images, wherein the fine target detection comprises the following steps: detecting low-scattering target pixels of the target sub-block coherent image, detecting a target detail contour of the target sub-block SAR image, and fusing the two detection results; and S5, removing false alarms from the fine target detection result, and carrying out region combination to obtain a final target detection result. The method can detect the targets with complex structures and weak scattering intensity in the sea surface large-scene SAR image, and has the advantages of high detection rate, low false alarm rate, strong adaptability and high calculation speed.

Description

Rapid detection method for SAR image target under strong sea clutter
Technical Field
The invention relates to a target rapid detection method, in particular to a target rapid detection method of an SAR image under a strong sea clutter, and belongs to the technical field of radar target identification.
Background
Synthetic Aperture Radar (SAR) imaging is an important means for sea surface target detection, monitoring and identification, and has important application value in the fields of marine traffic supervision, fishery management, even marine military target reconnaissance and the like. Target detection is a key processing technology in the SAR image target recognition system, and can extract an interested target Region (ROI) from a complex large-scene marine environment, so that necessary information input is provided for further target feature extraction and recognition, and therefore the performance Of the target detection has important influence on the precision and the efficiency Of the whole recognition system.
At present, the most deep and widely applied target detection research of the SAR image is the Constant False-Alarm Rate (CFAR) detection algorithm. The algorithm is based on the assumption that relatively obvious scattering brightness difference exists between a target and a background clutter, combines a given constant false alarm rate Parameter (PFA), and adaptively determines a decision threshold according to local statistical characteristics of the clutter to realize target detection. However, when the sea clutter is strong (such as high sea conditions and small incident angle), the signal-to-clutter ratio in the SAR image is low, and the traditional CFAR detection is easy to generate a large amount of false alarms or target missing detection phenomena. In addition, with the improvement of the resolution of the SAR image and the increase of the scene width, the details and the content of the interference clutter in the SAR imaging scene are extremely complex, and the data volume is increased greatly. Therefore, the capability of the CFAR algorithm for carrying out efficient target detection on SAR images of complex ocean large scenes is to be improved.
In a patent of layered high-resolution SAR image ship detection method based on multilayer CFAR (synthetic aperture radar) (publication number: CN104166838A) applied by the inventor of Houbiao and the like in 2014 at the university of Western Ann electronic technology, it is recorded that the influence of a strong scattering target pixel on the estimation of clutter statistical parameters is removed by using log-normal distribution as a clutter statistical distribution model and through multilayer iteration indexes based on the CFAR detection, the detection precision and the applicability of the algorithm are improved, and false alarms are removed by combining priori knowledge. However, the method has limited applicability to the sea surface target detection of the complex large-scene SAR image based on the iterative index strategy of the global threshold, and as the complex and variable ocean current and climate conditions easily cause the difference of clutter statistical characteristics of different sea areas, the corresponding optimal detection thresholds are different, so that the target is difficult to be detected with high precision by using the global threshold. In addition, clutter suppression preprocessing is not considered in the algorithm design, so that the phenomena of missing detection or high false alarm rate are easy to occur when weak and small targets under strong clutter are detected. Moreover, the logarithm positive-phase distribution adopted by the algorithm has 'over' fitting to a low-value part in the SAR image histogram, and false alarms are easily introduced.
In a 2016 patent of SAR image target detection method based on visual attention mechanism model and constant false alarm rate (publication number: CN105354541A) applied by Liu Feng et al, Sian electronic science and technology university, it is described that a potential target area is preliminarily selected through saliency map estimation and threshold detection based on Fourier spectrum residual error information; therefore, the potential target area is subjected to more fine target detection by adopting the K distribution-based adaptive sliding window CFAR algorithm. However, the algorithm is more suitable for target positioning because the detailed contour of the target in the saliency map is fuzzy and the shape is easily distorted, thereby causing the target to be detected in an 'over' mode. In addition, the adopted K distribution model parameter is complex in calculation and is not suitable for modeling the extremely inhomogeneous sea clutter, so that the calculation burden is large, and the target detection false alarm rate under the strong clutter is high.
In a paper "An adaptive and fast CFAR adaptive base on automatic centering for target detection in high-resolution SAR images", published in 2009 by the inventor of national defense science and technology university, An automatic index-based CFAR detection algorithm is provided, the interference of a target pixel on clutter statistical parameter estimation is removed through single index, and a G-based detection algorithm is constructed0And the distributed parameter estimation acceleration strategy is used for improving the algorithm efficiency. However, the generation of the target index map in the algorithm requires a priori index depth (i.e. target pixel number), and the parameter is unknown in an actual environment, which results in that the optimal threshold is difficult to determine, so that the adaptivity and detection accuracy of the algorithm are limited. In addition, although the algorithm adopts an acceleration strategy, the processing time of the algorithm for the SAR image with the size of 1000 × 1000 pixels is still in the order of "minutes", and the requirement of large scene rapid target detection is difficult to meet.
In a paper "On the Iterative indexing for Target Detection in SAR Images" published by the inventor of Yi Cui of the university of Qinghua, an Iterative indexing strategy (ICS) is provided to remove the influence of a Target pixel On clutter distribution, no prior information such as index depth is needed, and the Target can be detected in a self-adaptive manner.
Further, in a paper "An Improved estimated effective sensing Scheme for CFAR shield Detection With SAR image", published by the inventor of Wentao An, etc. of the national satellite and marine application service unit, the initial target index map generation algorithm and the target pixel index manner in the method proposed by the Yi Cui, etc. are Improved to more thoroughly filter the influence of the target and its neighborhood strong scattering pixels on clutter distribution, improve Detection accuracy and reduce iteration times. However, the two algorithms are directly applied to iterative index processing of a large scene image and complex K-distribution parameter estimation, which brings a large computational burden and makes the target detection speed slow. In addition, clutter suppression preprocessing is not considered in the index CFAR algorithms, and target omission or high false alarm rate is easy to occur under strong clutter.
Based on the above, the invention provides a method for rapidly detecting an SAR image target under a strong sea clutter, so as to solve the defects and limitations in the prior art.
Disclosure of Invention
The invention aims to provide a method for rapidly detecting an SAR image target under a strong sea clutter, which can effectively detect a target with a complex structure and weak scattering intensity in a sea surface large scene SAR image; the method has the advantages of high detection rate, low false alarm rate, strong adaptability and high calculation speed.
In order to achieve the above object, the present invention provides a method for rapidly detecting an SAR image target under a strong sea clutter, comprising the following steps:
s1, dividing the large scene SAR image into a plurality of sub-block SAR images by adopting a region blocking strategy;
s2, acquiring a target-clutter contrast enhanced sub-block coherent image corresponding to each sub-block SAR image based on a sub-aperture coherent processing method;
s3, pre-screening a target sub-block coherent image containing a target and a target sub-block SAR image corresponding to the target sub-block coherent image as candidate target sub-block images based on the significance of each sub-block coherent image;
s4, carrying out fine target detection on the candidate target sub-block images; the method comprises the following steps: detecting low-scattering target pixels of the coherent image of each target sub-block; detecting a target detail contour of each target sub-block SAR image, and fusing two detection results;
s5, post-processing: and removing false alarms from the fine target detection result, and carrying out region combination to obtain a final detection result.
In S2, the specific steps are: and performing sub-aperture decomposition on each sub-block SAR image along the azimuth direction of the sub-block SAR image to obtain a plurality of sub-view images, and calculating the coherence coefficient of each sub-view image to obtain a sub-block coherent image after the target-clutter contrast enhancement corresponding to the sub-block SAR image.
In S2, a partially overlapping sub-aperture division method is used in the sub-aperture decomposition process.
In S3, the specific steps are: calculating a significance parameter of each sub-block coherent image; screening out subblock coherent images with significance parameters larger than a detection threshold, wherein the subblock coherent images are target subblock coherent images containing targets; screening out sub-block SAR images corresponding to the target sub-block coherent images, wherein the sub-block SAR images are target sub-block SAR images containing targets, and therefore candidate target sub-block images are obtained; the rest non-screened sea clutter sub-block images comprise: the sea clutter sub-block coherent image and the sea clutter sub-block SAR image.
In S3, the detection threshold is calculated according to the histogram of the saliency parameter of each sub-block coherent image.
In S4, the specific steps are: detecting low-scattering target pixels of the coherent images of each target sub-block by adopting a Pareto distribution-based global threshold detection algorithm; detecting a target detail contour of each target sub-block SAR image by adopting a fast iterative index CFAR detection algorithm based on an integral image; and fusing the two detection results to obtain a fine target detection result.
In S4, the detection of the target sub-block coherent image and the detection of the target sub-block SAR image are performed in parallel.
In S4, the method for detecting the low-scattering target pixel in the target sub-block coherent image based on the Pareto distribution global threshold detection algorithm specifically includes: accurately modeling the statistical characteristics of the sea clutter based on Pareto distribution to obtain an amplitude probability density function of the Pareto distribution; calculating Pareto distribution parameters by using the sea clutter sub-block coherent image obtained by pre-screening in the S3; calculating to obtain a global threshold according to a CFAR detection criterion based on Pareto distribution, Pareto distribution parameters and a constant false alarm rate set by a priori; and comparing the pixels in the coherent image of each target sub-block with a global threshold value respectively, judging the pixels larger than the global threshold value as targets, and judging the pixels as clutter if the pixels are not larger than the global threshold value, so as to obtain the detection result of the low-scattering target pixels.
In S4, the method for detecting the target detail profile of each target sub-block SAR image based on the fast iterative index CFAR detection algorithm of the integral image specifically includes: for each target sub-block SAR image, an initial target index map is generated by adopting a Pareto distribution-based global threshold detection algorithm
Figure BDA0001433197310000041
Filtering out target pixels and 4-neighborhood pixels in the target sub-block SAR image based on a target index map of the current iteration to obtain a corresponding sea clutter image of the current iteration; calculating a two-dimensional integral image of the current iteration sea clutter image, and quickly calculating to obtain a corresponding local threshold value based on a CFAR detection criterion; generating a target index map of the current iteration through sliding window scanning and threshold judgment; and comparing the target index map of the current iteration with the target index map of the previous iteration, if the target index maps are different, continuing the iteration, and if the target index maps are the same, obtaining the detection result of the target detail outline.
In S5, false alarms are removed from the fine target detection result by counting filtering and morphological processing, and the fine target detection result is subjected to region merging to obtain a final detection result.
In summary, the method for rapidly detecting the target of the SAR image under the strong sea clutter provided by the invention can effectively detect the target with complex structure and weak scattering intensity in the SAR image under the sea surface large scene through the parallel processing based on the region blocking, the preprocessing of the target-clutter contrast enhancement, the target detection strategy from coarse to fine and the improved adaptive target detection and fusion technology. The method has the advantages of high detection rate, low false alarm rate, strong adaptability and high calculation speed; therefore, the defects of low target detection speed, low weak and small target detection rate and high false alarm rate of the strong sea clutter large scene SAR image in the prior art can be overcome.
Drawings
FIG. 1 is a flow chart of a fast detection method of SAR image targets under strong sea clutter according to the present invention;
FIG. 2 is a flow chart of enhancing the target-clutter contrast of sub-block SAR images based on sub-aperture coherent processing method in the present invention;
FIG. 3 is a flow chart of an integral image based fast iterative index CFAR detection algorithm in the present invention;
FIG. 4 is a schematic diagram of a sea surface large scene SAR image in the invention;
FIG. 5 is an enlarged schematic view of the portion of FIG. 4 encircled with a box;
FIG. 6 is a histogram of the significance parameter of each sub-block coherent image in the present invention;
FIG. 7 is a diagram illustrating the pre-screening result of the target sub-block coherent image in the present invention;
fig. 8 is a schematic diagram of the final target detection result in the present invention.
Detailed Description
A preferred embodiment of the present invention will be described in detail below with reference to fig. 1 to 8.
As shown in fig. 1, a flowchart of the method for rapidly detecting an SAR image target under a strong sea clutter according to the present invention includes the following steps:
s1, dividing the large scene SAR image into a plurality of sub-block SAR images by adopting a region blocking strategy;
s2, acquiring a target-clutter contrast enhanced sub-block coherent image corresponding to each sub-block SAR image based on a sub-aperture coherent processing method;
s3, pre-screening a target sub-block coherent image containing a target and a target sub-block SAR image corresponding to the target sub-block coherent image as candidate target sub-block images based on the significance of each sub-block coherent image;
s4, carrying out fine target detection on the candidate target sub-block images; the method comprises the following steps: detecting low-scattering target pixels of the coherent image of each target sub-block; detecting a target detail contour of each target sub-block SAR image, and fusing two detection results;
s5, post-processing: and removing false alarms from the fine target detection result, and carrying out region combination to obtain a final detection result.
In S1, the purpose of dividing the large-scale scene SAR image into a plurality of sub-block SAR images is to facilitate the subsequent parallel computation and pre-screening of each sub-block SAR image, thereby achieving coarse-to-fine target detection and improving the target detection efficiency of the large-scale scene SAR image. The size of the sub-block SAR image obtained after division can be specifically set according to the pixel resolution unit, the target maximum size, the sea situation and the like.
In a preferred embodiment of the present invention, it is assumed that the sea surface large scene SAR image is as shown in fig. 4, and has a size of 3000 × 3000, and a pixel resolution unit of 3 × 3 m. The large scene SAR image not only contains dozens of ships, but also contains some interferences such as strong scattering clutter and target orientation blur, which are circled by a box in fig. 4 and can be clearly shown by an enlarged view of fig. 5. The size of each sub-block SAR image is set to be 300 x 300, and the large scene SAR image can be divided into 100 sub-block SAR images by adopting a region blocking strategy. The obtained sub-block SAR images are mutually independent during subsequent steps, so that parallel calculation is facilitated; meanwhile, each sub-block SAR image can be preliminarily screened.
As shown in fig. 2, in S2, specifically, the following steps are performed: for each sub-block SAR image, sub-aperture decomposition is carried out along the azimuth direction of the sub-block SAR image to obtain a plurality of sub-view images (n sub-view images are displayed in total in fig. 2), the coherence coefficient of each sub-view image is calculated, and a target-clutter contrast enhanced sub-block coherent image corresponding to the sub-block SAR image is obtained:
Figure BDA0001433197310000061
wherein the content of the first and second substances,
Figure BDA0001433197310000062
a k sub-block coherent image corresponding to the k sub-block SAR image; n is a radical ofBThe total number of sub-block SAR images (and the total number of sub-block coherent images); spAnd sp+1Respectively representing the result of sub-aperture decomposition of the k sub-block SAR imageUp to the ith and (i + 1) th sub-view images; superscript denotes complex conjugate operation;<·>representing a spatial averaging operation.
Because the ship targets in different sub-visual images have coherence and the sea clutter has the characteristic of non-coherence, the sea clutter in the obtained sub-block coherent image can be inhibited and the targets can be enhanced, thereby improving the target-clutter contrast.
As shown in fig. 2, the method for performing sub-aperture decomposition on each sub-block SAR image in S2 specifically includes the following steps:
firstly, converting a sub-block SAR image to a range-Doppler domain along the azimuth direction to obtain SAR data;
secondly, estimating an azimuth window function and removing the influence of the azimuth window function on SAR data;
thirdly, dividing the SAR data into a plurality of partial overlapped sub-apertures along the azimuth direction, and keeping the spatial resolution as much as possible when subsequently acquiring coherent images by adopting the mode;
and finally, windowing and up-sampling the SAR data of each sub-aperture, and transforming the SAR data into an image domain to obtain each sub-view image.
The step S3 specifically includes the following steps:
s31, calculating the Significance parameter (S, Significance) of each subblock coherent image:
Figure BDA0001433197310000071
wherein S iskRepresenting the kth sub-block coherent image
Figure BDA0001433197310000072
A significance parameter of (d); n is a radical ofBIs the total number of sub-block coherent images; mu.skAnd σkAre respectively as
Figure BDA0001433197310000073
The amplitude mean and variance of (d); | · | represents a modulo operation; max (. cndot.) denotes maximizingPerforming value operation;
s32, because the scattering intensity of the target is usually larger than that of the sea clutter, the corresponding peak intensity is also larger; and because the number of pixels of the target is less, the influence on the mean value and the variance is less; the saliency parameters of the sub-block coherent image containing the object are therefore generally large;
screening out subblock coherent images with significance parameters larger than a detection threshold, wherein the subblock coherent images are target subblock coherent images containing targets; screening out sub-block SAR images corresponding to the target sub-block coherent images, wherein the sub-block SAR images are target sub-block SAR images containing targets;
the detection threshold is obtained by estimating according to a histogram of significance parameters of each sub-block coherent image;
s33, taking the screened target sub-block coherent image and the target sub-block SAR image as candidate target sub-block images, so as to realize more precise target detection based on the candidate target sub-block coherent image and the target sub-block SAR image; the non-screened images are sea clutter sub-block images, and the method comprises the following steps: the sea clutter sub-block coherent image and the sea clutter sub-block SAR image can be applied to modeling of sea clutter statistical characteristics in a subsequent fine detection stage, as shown in FIG. 1.
In a preferred embodiment of the present invention, as shown in fig. 6, the detection threshold is determined to be 23 according to the histogram of the significance parameter of each sub-block coherent image, so as to screen out the sub-block coherent image with the significance parameter greater than 23 and the corresponding sub-block SAR image to obtain a candidate target sub-block image, and the rest of the sub-block coherent images that are not screened out are sub-block clutter images, as shown in fig. 7, that is, results obtained after pre-screening. As is apparent from fig. 7, after the pre-screening, the large-area sea clutter sub-block image, especially the sub-block image containing the clutter with strong scattering shown in fig. 5, can be effectively filtered, so as to reduce the calculation burden and the false alarm rate of the subsequent fine detection processing.
As shown in fig. 1, in S4, the step of performing more detailed target detection processing on the candidate target sub-block image includes the following steps:
s4a, detecting low-scattering target pixels of the coherent image of each target sub-block by adopting a Pareto (Pareto) distribution-based global threshold detection algorithm;
s4b, detecting a target detail contour of each target sub-block SAR image by adopting a fast iterative index CFAR detection algorithm based on an integral image;
s4c, fusing the detection results in the S4a and the S4b to obtain a fine target detection result;
the detection of the target sub-block coherent image by the S4a and the detection of the target sub-block SAR image by the S4b can be performed in a parallel mode, so that the calculation efficiency of the algorithm is effectively improved.
In S4a, in order to detect the low-scattering target pixel under the strong sea clutter, detection needs to be performed based on the target sub-block coherent image obtained by target-clutter contrast enhancement. Particularly, the method adopts Pareto distribution which has the modeling capability of uniform, non-average and extremely-non-uniform clutter areas, has a simple expression and is easy for parameter estimation, and accurately models the statistical characteristics of the sea clutter; and further determining a global threshold according to a CFAR detection criterion, and judging a detection target through the global threshold. The global threshold detection algorithm based on Pareto distribution specifically comprises the following steps:
s4a1, modeling the statistical characteristics of the sea clutter based on Pareto distribution, wherein the expression of the amplitude Probability Density Function (PDF) of the Pareto distribution is as follows:
Figure BDA0001433197310000081
wherein, vpAnd ηpRespectively representing the shape and scale parameters of Pareto distribution, both being greater than 0;
when v ispOn → infinity, the Pareto distribution degenerates into a rayleigh distribution, which is suitable for modeling uniform sea clutter;
Figure BDA0001433197310000082
at the moment, the Pareto distribution has an extremely long trailing characteristic and is suitable for modeling extremely non-uniform sea clutter;
s4a2, sea clutter obtained by pre-screening in S33Block coherent image estimation Pareto distribution parameter vpAnd ηpThe influence of the target pixel on the statistical distribution of the sea clutter can be reduced;
the Pareto distribution parameter vpAnd ηpThe method can be obtained by fast calculating the 2-order and 4-order sample moments of a sample set consisting of all pixel points in the sea clutter sub-block coherent image:
Figure BDA0001433197310000091
wherein min (-) represents the minimum calculation; mu.s2And mu4Respectively representing the sample moments of 2-order and 4-order, and calculated by the following method:
Figure BDA0001433197310000092
wherein, muκRepresenting a k-order sample moment; z (t) is each reference unit; n is a radical ofsIs the number of reference cells;
s4a3, setting the CFAR detection criterion based on Pareto distribution as follows:
Figure BDA0001433197310000093
wherein p isfaRepresents a constant false alarm rate of 0<pfa<<1；fp(z) is the amplitude probability density function of the Pareto distribution; dzRepresents the derivation with respect to variable z; t represents a global threshold to be determined and can be defined by a Pareto distribution parameter vpAnd ηpQuickly solving the following steps:
Figure BDA0001433197310000094
in this embodiment, a constant false alarm rate p is determinedfa＝10-4；
S4a4, respectively connecting the pixels in the coherent image of each target sub-block with the full imageComparing local threshold values; judging the pixels larger than the global threshold value as targets, and marking 1; otherwise, judging the pixel as clutter, marking 0 and obtaining a detection result B of the low-scattering target pixel1。
As shown in fig. 3, in S4b, for the problem that the spatial resolution of the target sub-block coherent image is lower than that of the target sub-block SAR image, so that part of the target detail contour is lost and the target pixel is missed, the fast iterative index CFAR detection algorithm based on the integral image shown in fig. 3 is further adopted for each target sub-block SAR image, so as to more finely and completely extract the target detail contour features therein. The algorithm mainly utilizes an iterative index strategy (ICS) to more thoroughly remove the influence of a target pixel on the sea clutter statistical distribution, improves the applicability and the detection precision of the algorithm on a complex scene, and realizes algorithm acceleration based on an integral image operator. The method specifically comprises the following steps:
s4b1, for each target sub-block SAR image, generating an initial target index map by adopting the Pareto distribution-based global threshold detection algorithm in S4a
Figure BDA0001433197310000095
The Pareto distribution parameters and the global threshold are obtained by utilizing the sea clutter sub-block SAR image obtained by pre-screening in S33;
s4b2, filtering out target pixels and 4-neighborhood pixels in the target sub-block SAR image based on the target index map of the current iteration to obtain the corresponding clutter image of the current iteration
Figure BDA00014331973100001011
Figure BDA0001433197310000101
Figure BDA0001433197310000102
Wherein i is the number of iterations(ii) a I represents any target sub-block SAR image; represents a matrix dot product operation;
Figure BDA0001433197310000103
the method comprises the steps that a sea clutter index map of current iteration is used for filtering target pixels and 4-neighborhood pixels in a target sub-block SAR image, wherein the sea clutter pixel value is 1, the pixel values of a target and the 4-neighborhood pixels are 0, and NOT (DEG) represents logical negation operation;
Figure BDA0001433197310000104
representing a morphological dilation operator, and H represents a corresponding morphological probe element;
s4b3, calculating a two-dimensional integral image of the current iteration sea clutter image, and rapidly calculating a sample moment of a reference unit in any sliding window through simple addition and subtraction operations based on the two-dimensional integral image, so that a corresponding local threshold value is rapidly estimated through the CFAR detection criteria recorded in S4a2 and S4a 3;
s4b4, generating a target index map of the current iteration through sliding window scanning and threshold value judgment
Figure BDA0001433197310000105
S4b5, comparing the target index map of the current iteration
Figure BDA0001433197310000106
And the target index map of the previous iteration
Figure BDA0001433197310000107
If not, returning to S4b2, continuing the next iteration, and utilizing the target index map of the current iteration
Figure BDA0001433197310000108
Regenerating a sea clutter image and continuing to detect for the next time; if they are the same, the difference pixel number NiIf the detected detail contour is 0, the detection is terminated, and the detection result of the target detail contour is obtained and output
Figure BDA0001433197310000109
Wherein the content of the first and second substances,
Figure BDA00014331973100001010
namely, the pixels of the target index map of the two iterations are compared in sequence according to rows and columns.
In S4c, the specific steps are: and fusing the detection results of S4a and S4B to obtain a more refined target detection result B:
B＝B2||B3＝B2||Sign+(B1-B2)；
wherein, | | represents a logical or operation; b is3＝Sign+(B1-B2) Representing weak target pixels, Sign, detectable only by coherent images of target sub-blocks+Indicating positive sign operation, i.e. when B1-B2When the pixel value of (B) is greater than 0, B3The pixel value of the corresponding position is 1, otherwise, 0.
In S5, false alarms are removed from the fine target detection result by counting filtering and morphological processing, and the fine target detection result is subjected to region merging to obtain a final detection result. As shown in fig. 8, which is a final detection result obtained after the target is rapidly detected by using the method of the present invention, compared with the original image before detection shown in fig. 4, the ship information therein is effectively extracted, and the sea clutter existing in the original image is effectively filtered.
In summary, according to the method for rapidly detecting the target of the SAR image under the strong sea clutter, firstly, the large-scene SAR image is subjected to region blocking, so that each sub-block SAR image can be subjected to parallel calculation and pre-screening processing in the following process; then, the target-sea clutter contrast of each sub-block SAR image is improved through sub-aperture coherent processing, and a pre-screening algorithm based on significance is adopted to quickly obtain a candidate target sub-block image; furthermore, for each candidate target sub-block image, an improved self-adaptive target detection and fusion strategy is adopted, so that a small target and a target detail outline in the candidate target sub-block image are effectively extracted, and a more precise target detection result is obtained; and finally, removing false alarms through automatic post-processing to obtain a final target detection result.
Therefore, the method for rapidly detecting the SAR image target under the strong sea clutter provided by the invention can effectively detect the target with complex structure and weak scattering intensity in the sea surface large scene SAR image through the parallel processing based on the region blocks, the preprocessing of the target-clutter contrast enhancement, the target detection strategy from coarse to fine and the improved self-adaptive target detection and fusion technology, and has the advantages of high detection rate, low false alarm rate, strong adaptability and high calculation speed.
While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.

Claims (7)

1. A rapid detection method for SAR image targets under strong sea clutter is characterized by comprising the following steps:
s1, dividing the large scene SAR image into a plurality of sub-block SAR images by adopting a region blocking strategy;
s2, acquiring a target-clutter contrast enhanced sub-block coherent image corresponding to each sub-block SAR image based on a sub-aperture coherent processing method;
s3, pre-screening a target sub-block coherent image containing a target and a target sub-block SAR image corresponding to the target sub-block coherent image as candidate target sub-block images based on the significance of each sub-block coherent image;
s4, carrying out fine target detection on the candidate target sub-block images, wherein the fine target detection comprises the following steps:
s4a, detecting low-scattering target pixels of the coherent image of each target sub-block by adopting a Pareto distribution-based global threshold detection algorithm;
s4b, detecting a target detail contour of each target sub-block SAR image by adopting a fast iterative index CFAR detection algorithm based on an integral image;
s4c, fusing the detection results in the S4a and the S4b to obtain a fine target detection result;
s5, removing false alarms from the fine target detection result, and carrying out region merging to obtain a final detection result;
wherein, in S4b, the method specifically includes the following steps:
s4b1, for each target sub-block SAR image, generating an initial target index map by adopting a Pareto distribution-based global threshold detection algorithm
Figure FDA0002871749250000011
The Pareto distribution parameters and the global threshold are obtained by calculation by using the sea clutter sub-block SAR image obtained in S33;
s4b2, filtering out target pixels and 4-neighborhood pixels in the target sub-block SAR image based on the target index map of the current iteration to obtain the corresponding clutter image of the current iteration
Figure FDA0002871749250000012
Figure FDA0002871749250000013
Figure FDA0002871749250000014
Wherein i is the number of iterations; i represents any target sub-block SAR image; represents a matrix dot product operation;
Figure FDA0002871749250000021
the method comprises the steps that a sea clutter index map of current iteration is used for filtering target pixels and 4-neighborhood pixels in a target sub-block SAR image, wherein the sea clutter pixel value is 1, the pixel values of a target and the 4-neighborhood pixels are 0, and NOT (DEG) represents logical negation operation;
Figure FDA0002871749250000022
representing a morphological dilation operator, and H represents a corresponding morphological probe element;
s4b3, calculating a two-dimensional integral image of the current iteration sea clutter image, and quickly calculating to obtain a corresponding local threshold value based on a CFAR detection criterion;
s4b4, generating a target index map of the current iteration through sliding window scanning and threshold value judgment
Figure FDA0002871749250000023
S4b5, comparing the target index map of the current iteration
Figure FDA0002871749250000024
And the target index map of the previous iteration
Figure FDA0002871749250000025
If not, returning to S4b2 and continuing the next iteration; if the target detail contour is the same as the target detail contour, obtaining and outputting a detection result of the target detail contour
Figure FDA0002871749250000026
In S4c, the specific steps are: and fusing the detection results of S4a and S4B to obtain a fine target detection result B:
B＝B2||B3＝B2||Sign+(B1-B2)；
wherein, | | represents a logical or operation; b is3＝Sign+(B1-B2) Representing weak target pixels, Sign, detectable only by coherent images of target sub-blocks+Indicating a positive sign operation.
2. The method for rapidly detecting the target of the SAR image under the strong sea clutter according to claim 1, wherein in S2, the method specifically comprises: for each sub-block SAR image, sub-aperture decomposition is carried out along the azimuth direction of the sub-block SAR image to obtain a plurality of sub-view images, the coherence coefficient of each sub-view image is calculated to obtain a target-clutter contrast enhanced sub-block coherent image corresponding to the sub-block SAR image:
Figure FDA0002871749250000027
wherein the content of the first and second substances,
Figure FDA0002871749250000028
a k sub-block coherent image corresponding to the k sub-block SAR image; NB is the total number of the sub-block SAR images; spAnd sp+1Respectively representing the ith and (i + 1) th sub-view images obtained by sub-aperture decomposition of the kth sub-block SAR image; superscript denotes complex conjugate operation;<·>representing a spatial averaging operation.
3. The method for rapidly detecting the SAR image target under the strong sea clutter according to claim 2, wherein a partial overlapping sub-aperture division mode is adopted in the sub-aperture decomposition process.
4. The method for rapidly detecting the target of the SAR image under the strong sea clutter according to claim 2, wherein the step S3 specifically comprises the following steps:
s31, calculating the significance parameter of each sub-block coherent image:
Figure FDA0002871749250000031
wherein S iskRepresenting the kth sub-block coherent image
Figure FDA0002871749250000032
A significance parameter of (d); n is a radical ofBIs the total number of sub-block coherent images; mu.skAnd σkAre respectively as
Figure FDA0002871749250000033
The amplitude mean and variance of (d); | · | represents a modulo operation; max (·) represents the maximum operation;
s32, screening out subblock coherent images with significance parameters larger than a detection threshold, wherein the subblock coherent images are target subblock coherent images containing targets; screening out sub-block SAR images corresponding to the target sub-block coherent images, wherein the sub-block SAR images are target sub-block SAR images containing targets;
s33, taking the screened target sub-block coherent image and the target sub-block SAR image as candidate target sub-block images; the non-screened images are sea clutter sub-block images, and the method comprises the following steps: the sea clutter sub-block coherent image and the sea clutter sub-block SAR image.
5. The method for rapidly detecting the SAR image target under the strong sea clutter according to claim 4, wherein the detection threshold is obtained according to histogram estimation of the significance parameter of each sub-block coherent image.
6. The method for rapidly detecting the SAR image target under the strong sea clutter as claimed in claim 5, wherein the S4a and S4b are performed in a parallel manner.
7. The method for rapidly detecting the target of the SAR image under the strong sea clutter according to claim 5, wherein the step S4a specifically comprises the following steps:
s4a1, modeling the statistical characteristics of the sea clutter based on Pareto distribution, wherein the expression of the amplitude probability density function of the Pareto distribution is as follows:
Figure FDA0002871749250000034
wherein, vpAnd ηpRespectively are Pareto distribution parameters;
s4a2, sea clutter sub-block coherent image meter obtained by S33Calculating Pareto distribution parameter vpAnd ηp；
S4a3, setting the CFAR detection criterion based on Pareto distribution as follows:
Figure FDA0002871749250000035
wherein p isfaRepresents constant false alarm rate and 0 < pfa＜＜1；fp(z) is the amplitude probability density function of the Pareto distribution; dzRepresents the derivation with respect to variable z; t represents a global threshold value which can be defined by a Pareto distribution parameter vpAnd ηpObtaining:
Figure FDA0002871749250000041
s4a4, comparing the pixels in the target sub-block coherent images with a global threshold respectively; judging the pixels larger than the global threshold value as targets, and marking 1; otherwise, judging the pixel as clutter, marking 0 and obtaining a detection result B of the low-scattering target pixel1。
CN201710952653.3A 2017-10-13 2017-10-13 Rapid detection method for SAR image target under strong sea clutter Active CN107808383B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710952653.3A CN107808383B (en) 2017-10-13 2017-10-13 Rapid detection method for SAR image target under strong sea clutter

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710952653.3A CN107808383B (en) 2017-10-13 2017-10-13 Rapid detection method for SAR image target under strong sea clutter

Publications (2)

Publication Number Publication Date
CN107808383A CN107808383A (en) 2018-03-16
CN107808383B true CN107808383B (en) 2021-04-02

Family

ID=61584967

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710952653.3A Active CN107808383B (en) 2017-10-13 2017-10-13 Rapid detection method for SAR image target under strong sea clutter

Country Status (1)

Country Link
CN (1) CN107808383B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113555737A (en) * 2021-09-17 2021-10-26 长沙泰科阳微电子有限公司 SAR image detection method and system
US20220005180A1 (en) * 2020-09-24 2022-01-06 Beijing Baidu Netcom Science And Technology Co., Ltd. Method assisting quality inspection of map data and electronic device

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108960085B (en) * 2018-06-20 2021-10-01 中国电子科技集团公司第五十四研究所 Method for rapidly extracting synthetic aperture radar image ship target candidate area
CN109191483A (en) * 2018-09-19 2019-01-11 南昌航空大学 A kind of quick watershed detection method of helicopter blade Circle in Digital Images shape mark
CN109856604B (en) * 2018-11-13 2020-12-22 上海无线电设备研究所 Two-dimensional scaling rapid constant false alarm detection method
CN112598572B (en) * 2019-10-01 2022-04-15 浙江大学 Method and device for screening subblock images and processing units
CN111381216B (en) * 2020-03-10 2022-03-11 中国科学院微电子研究所 Hybrid distribution radar sea clutter analysis method and device
CN111931593B (en) * 2020-07-16 2024-04-26 上海无线电设备研究所 Weak target detection method based on deep neural network and time-frequency image sequence
CN112505701B (en) * 2020-12-14 2023-09-22 大连海事大学 Method for inhibiting speckle of non-local mean value of approximate half-declining integral graph number
CN113034366A (en) * 2021-03-25 2021-06-25 中国科学院空天信息创新研究院 Splicing seam-free SAR image segmentation parallel acceleration processing method

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101464956A (en) * 2009-01-21 2009-06-24 电子科技大学 Method for classifying polarization synthetic aperture radar image based on sub-aperture analysis
CN101609151B (en) * 2009-07-17 2011-11-02 重庆大学 Moving target detection method based on eigenvalue decomposition of single-channel synthetic aperture radar (SAR) image sequence
CN102288963B (en) * 2011-07-21 2013-06-12 西安电子科技大学 Bistatic inverse synthetic aperture radar (ISAR) image fusion method based on sub aperture parameter estimation
CN102609709B (en) * 2012-02-03 2014-04-16 清华大学 Sea surface oil spilling segmentation method based on polarized SAR (synthetic aperture radar) data fusion
CN103217677A (en) * 2013-05-10 2013-07-24 重庆大学 Single-channel SAR (synthetic aperture radar) moving target detecting method based on joint detection amount
CN105093188B (en) * 2015-08-24 2017-05-24 中国科学院电子学研究所 SAR ocean image sea wave clutter suppression method based on sub-aperture image
CN106910177B (en) * 2017-01-20 2019-10-29 中国人民解放军装备学院 A kind of multi-angle SAR image fusion method that local image index optimizes

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220005180A1 (en) * 2020-09-24 2022-01-06 Beijing Baidu Netcom Science And Technology Co., Ltd. Method assisting quality inspection of map data and electronic device
CN113555737A (en) * 2021-09-17 2021-10-26 长沙泰科阳微电子有限公司 SAR image detection method and system

Also Published As

Publication number Publication date
CN107808383A (en) 2018-03-16

Similar Documents

Publication Publication Date Title
CN107808383B (en) Rapid detection method for SAR image target under strong sea clutter
CN107274401B (en) High-resolution SAR image ship detection method based on visual attention mechanism
CN107145874B (en) Ship target detection and identification method in complex background SAR image
CN101727662B (en) SAR image nonlocal mean value speckle filtering method
CN103942803B (en) SAR (Synthetic Aperture Radar) image based automatic water area detection method
EP2871610B1 (en) Infrared image based early detection of oil spills in water
CN106170819A (en) A kind of diameter radar image Ship Target method for quick
CN102968799B (en) Integral image-based quick ACCA-CFAR SAR (Automatic Censored Cell Averaging-Constant False Alarm Rate Synthetic Aperture Radar) image target detection method
US20120328161A1 (en) Method and multi-scale attention system for spatiotemporal change determination and object detection
CN110889843B (en) SAR image ship target detection method based on maximum stable extremal region
CN104166838B (en) Layering High Resolution SAR image Ship Detection based on multi-layer C FAR
CN113536963B (en) SAR image airplane target detection method based on lightweight YOLO network
CN111311644B (en) Moving target detection method based on video SAR
WO2018000252A1 (en) Oceanic background modelling and restraining method and system for high-resolution remote sensing oceanic image
CN113674308B (en) SAR image ship target rapid detection method based on image enhancement and multiple detection
Chen et al. A novel infrared small target detection method based on BEMD and local inverse entropy
CN108305265B (en) Real-time processing method and system for weak and small target image
AlAzawee et al. Using morphological operations—Erosion based algorithm for edge detection
Li et al. Automatic infrared ship target segmentation based on structure tensor and maximum histogram entropy
CN107369163B (en) Rapid SAR image target detection method based on optimal entropy dual-threshold segmentation
CN107729903A (en) SAR image object detection method based on area probability statistics and significance analysis
CN105204010A (en) Ground object target detection method of low signal-to-clutter ratio synthetic aperture radar image
CN116188944A (en) Infrared dim target detection method based on Swin-transducer and multi-scale feature fusion
CN108564054B (en) Accurate oil spill detection method based on CFAR
CN113963017A (en) Real-time infrared small and weak target detection method and device and computer equipment

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