CN105205484A - Synthetic aperture radar target detection method based on curvelet transformation and Wiener filtering - Google Patents
Synthetic aperture radar target detection method based on curvelet transformation and Wiener filtering Download PDFInfo
- Publication number
- CN105205484A CN105205484A CN201410691384.6A CN201410691384A CN105205484A CN 105205484 A CN105205484 A CN 105205484A CN 201410691384 A CN201410691384 A CN 201410691384A CN 105205484 A CN105205484 A CN 105205484A
- Authority
- CN
- China
- Prior art keywords
- image
- target
- filtering
- decomposition
- scale
- 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
Links
Landscapes
- Radar Systems Or Details Thereof (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to the signal and information processing technology field and relates to a target detection method of combining a curvelet transformation theory, statistical parameter estimation, Wiener filtering and a synthetic aperture radar image characteristic. The method is characterized by through curvelet transformation decomposition, acquiring a plurality of decomposition scales; on each decomposition scale, providing different direction information and setting a direction of each scale; making an influence of noise reach a minimum through sampling filtering twice; taking a low false alarm probability to detect abundant target information. Compared to the prior art, by using the method, there are the following advantages that a defect that a small target, a weak scattering target and a hidden target is not easy to detect is effectively overcome through double filtering; an influence of speckle noise is reduced; a signal to noise ratio of a SAR image is increased; the curvelet transformation is selected to carry out processing on the SAR image so that a processing result is accurate; de-noise processing is performed on the image and enhancement processing is performed on a characteristic so that a detection rate of a target is increased; a false alarm probability of the target is low and a high anti-false-alarm capability is possessed.
Description
Technical field
The invention belongs to Signal and Information Processing technical field, relate to the object detection method that warp wavelet theory, statistical parameter estimation, Wiener filtering and diameter radar image feature combine.
Background technology
Synthetic-aperture radar (SyntheticApertureRadar, hereinafter referred to as SAR) is a kind of very important space observation technological means over the ground.Due to the advantage of himself, since nineteen fifty-three obtains the first width SAR image, in ocean, geology, agricultural, environment, the various fields such as city and earth monitor be widely used.SAR imaging, except typical round-the-clock, round-the-clock, has outside certain penetration capacity, and the spatial resolution of the SAR image of acquisition and image-forming range have nothing to do.Therefore, in actual applications, it is better than remote optical sensing imaging and infrared remote sensing imaging, and only the resolution of SAR image does not also far reach its physics limit at present.SAR is a kind of coherence's microwave imaging remote sensing radar, therefore the typical feature of SAR image contains a large amount of " spot " noise (speckle), but they are not again real noises.These noises have had a strong impact on decipher and the application of SAR image, especially SAR target detection, classification with identification.
Because speckle noise is that SAR imaging mechanism produces, so it impossible, by complete filtering, can only suppress as far as possible, it is down to minimum on the impact of the detection and Identification of target as far as possible.Therefore, be the pretreated very important substance of SAR image to the filtering of speckle noise.About the method for SAR image speckle noise filtering and Theory comparison many, substantially can be divided into the multiple look processing before imaging and the filtering process after imaging.Multiple look processing mainly by segmentation doppler bandwidth, then imaging respectively, then carries out incoherent averaging process.Although this method effectively can suppress speckle noise, to lose SAR image spatial resolution for cost, if process to as if the SAR image of Area Objects, so can adopt this aspect.SAR image speckle noise is suppressed to be main treatment technology after imaging, it comprises again two aspects: one is the filtering based on spatial domain, representational filtering method has average and medium filtering, Lee filtering, Kuan filtering, Frost filtering, MAP filtering etc., these methods can the speckle noise of filtering SAR image preferably, is classical picture filtering method.They normally by regulating filter window size to carry out de-noising, although inhibit speckle noise, also reduce the resolution of image in filter window simultaneously, cause the fuzzy of image border and linear target; Some marginal information keeps better, but speckle noise is not suppressed fully.These methods use speckle noise statistical model usually, although and do not use speckle noise statistical model based on the filtering method of partial differential equation, but be also a kind of extraordinary airspace filter method, as non-linear multiple diffusing filter, P-M diffusing filter, from snake diffusion, tensor diffusion etc., these class methods all can obtain good effect in squelch and edge maintenance, but when edge noise suppresses, effect is not ideal.The another kind of important noise restraint method of Image retro-reconstruction is the filtering based on transform domain, as typical wavelet basis series of transformations is theoretical, multi-scale geometric analysis is theoretical.This kind of usually by the process of transform domain, adopt the means of shrinking, namely in different resolution-scale, slacken or remove the coefficient relevant to speckle noise.The suppression of SAR image speckle noise and the protection of geometric detail information are the problems of conflict always, and nearly all method is all both compromise, does not have a kind of method perfectly to meet this requirement.Only in actual applications, Choose filtering method is carried out according to the object used and emphasis.At present, widely used object detection method remains constant false alarm rate ((Coherence-basedConstantFalseAlarmRatio, hereinafter referred to as CCFAR) method, i.e. given false-alarm probability, makes target detection rate reach maximum.Detecting the shortcoming of target with CFAR, to be that requirement target and background clutter is distributed with obvious reflection strength poor, that is is directly difficult to faint, little and hiding target be detected by CFAR method.
The suppression of SAR image speckle noise is the basis of SAR image application, instead of final goal.How accurately, fast, effectively to obtain ground object target information in SAR image, be only the object of SAR image application.Therefore, in the application of SAR, the acquisition of the detection of SAR target, segmentation and identification and change information is the important content of target intelligence reconnaissance and surveillance and battle space awareness all the time.
Summary of the invention
For above-mentioned prior art situation, technology of the present invention is dealt with problems and is: propose a kind of statistics Wiener Filtering based on warp wavelet and SAR target detection method, effectively can not only reduce the intrinsic speckle noise of SAR image, and be very beneficial for the detection of SAR image target, the verification and measurement ratio of target can be significantly improved.
Now design of the present invention and technical solution are described below:
SAR is a kind of active microwave remote sensing imaging radar, and speckle noise is its intrinsic characteristic, but produces impact greatly to SAR target detection, especially point target, weak scattering target etc.And SAR imaging is the mapping of ground object target feature space to image space, the interaction between electromagnetic wave and ground object target is very complicated, and the signal of returning from ground object target scattering is non-stationary, nonlinear properties.Warp wavelet is that a kind of new nonstationary random response is theoretical, is again that a kind of multi-scale geometric analysis is theoretical.Meanwhile, bent Wave Decomposition can provide the information of multiple directions, makes the information of acquisition more accurate.According to the feature of SAR imaging mechanism and warp wavelet, the present invention proposes a kind of statistics Wiener Filtering based on warp wavelet and object detection method.This method has many advantages, first, is decomposed by warp wavelet, can obtain multiple decomposition scale, and each yardstick can provide different information, is conducive to the extraction of feature and the filtering of noise; Secondly, after bent Wave Decomposition, on each decomposition scale, can provide different directional informations, and wavelet transformation can only provide the information in three directions on each decomposition scale, therefore warp wavelet overcomes the shortcoming of wavelet transformation directivity limitation, simultaneously, after warp wavelet decomposes, the direction of each yardstick can be arranged arbitrarily, meets the azimuthal sensitivity of SAR imaging, can obtain more accurately abundanter target detail and geological information; 3rd, twice filtering of sampling, obtains better filter effect, and namely first according to the situation of given SAR image, utilize statistical property to carry out filtering, next utilizes Wiener filtering to carry out auto adapted filtering again, and the impact of noise can be made like this to reach minimum; 4th, propose a kind of new SAR target detection method, the method increase the anti-false-alarm ability of target detection, when detecting target, only needing lower false-alarm probability, but abundant target information can be detected.Experiment shows that carried filtering method and object detection method are all unusual effective methods, has great application potential.
According to foregoing invention design and experimental result, the present invention proposes a kind of statistics based on warp wavelet dimension and receives SAR image filtering and object detection method, is described in detail (see Figure of description 1) respectively below to the main performing step of the method.
Step 1: input synthetic-aperture radar (SAR) image;
Step 2: the statistical property calculating input SAR image;
Step 3: the gray-scale relation determining interesting target region and background area;
Step 3.1: determine target area, i.e. interested region, then split area-of-interest from whole image, obtains the subimage of target area;
Step 3.2: the gray average calculating target area;
Step 3.3: the gray average of the gray average of target area and whole image is compared, determines whether target area is strong reflection region or extremely weak reflector space; The average of target area and the average of whole image are close to then thinking that the signal to noise ratio (S/N ratio) of whole image is low; The average of target area and the average of whole image differ greatly, and think that the noise of whole image is large, determine corresponding threshold value T for when carrying out filtering in step 5;
Step 4: the multiple dimensioned geometry decomposition of march ripple;
Step 5: determine filtering threshold T;
According to the relation determining target area and background in step 3, the SAR image namely inputted is that signal to noise ratio (S/N ratio) is high or the image that signal to noise ratio (S/N ratio) is low; If the gray-scale value of target area and background clutter gray-scale value are more or less the same, be the image that signal to noise ratio (S/N ratio) is lower, so the determination of threshold value T just calculates by formula (1).
In above formula, std represents standard deviation, and E represents negentropy, and μ represents average.If the gray-scale value of target area differs larger with background clutter gray-scale value, also namely the signal to noise ratio (S/N ratio) of image is comparatively large, and so the available formula of the determination of threshold value T (2) calculates.
Step 6: first time filtering process is carried out to SAR image:
In this step, the result in step 4 and step 5 is utilized to process; In step 4, SAR image obtains the subimage in directions different on different decomposition yardstick after decomposing, in each subimage, the threshold value utilizing step 5 to determine processes; If the value of certain pixel is greater than threshold value T in subimage, so just it is filtered out; If the value of certain pixel is less than threshold value T, so just retain it; Last result obtains filtered whole subimage;
Step 7: to SAR image march ripple inverse transformation after filtering;
Step 8: filtering is again carried out to SAR image with adaptive S filter:
Although the speckle noise that the noise in SAR image is mainly caused by its image-forming principle, be also subject to the impact of other factors simultaneously, can produce corresponding noise, these noises can hinder decipher to SAR image, interpretation and application equally; Here directly utilize Wei Na (Wiener) filtering to process SAR image, obtain final filtering SAR image;
Step 9: obtain characteristic image by bent Wave Decomposition coefficient: concrete steps are:
In step 4, decompose SAR image march wave conversion and obtain the little coefficient of dissociation of different scale condition, if the coefficient of certain layer is got initial value or processes, and the coefficient of other decomposition layer is zero, then carry out contrary flexure wave conversion, just can obtain the coefficient characteristics figure of certain layer; As required, these features selected and merges, then obtaining final high frequency coefficient characteristic pattern; Characteristic pattern mainly comprises edge and geometric detail information;
Step 10: obtain the SAR image after target signature enhancing:
Because SAR image is after twice filtering, although decrease the impact of noise, but simultaneously also some high-frequency informations of filtering, the geometry in what these high-frequency informations comprised is ground object target region, edge or grain details information, for target detection with to identify extremely important.The characteristic pattern of filtered SAR image and enhancing carries out fusion treatment, can obtain not only filtering but also the SAR image of Enhanced feature, is conducive to inspection and the identification of target;
Step 11: select the coherence's constant false alarm rate object detection method detection method put forward from SAR image mechanism to detect target, obtains SAR target detection threshold value T1.
The present invention further provides a kind of based on warp wavelet and the SAR target detection method of adding up Wiener filtering, it is characterized in that: " calculating the statistical property of input SAR image " described in step 2 also can carry out in the steps below:
Step 2.1: the average calculating SAR image.Average reflection image comprise the average case of quantity of information;
Step 2.2: calculate the standard deviation of SAR image, its reflection image information depart from situation;
Step 2.3: the entropy calculating SAR image, the overall condition of its reflection amount of image information.
The present invention further provides a kind of based on warp wavelet and the SAR target detection method of adding up Wiener filtering, it is characterized in that: " the multiple dimensioned geometry decomposition of march ripple " described in step 4 also can carry out in the steps below:
Step 4.1: the determination of decomposition scale, along with the increase of decomposition scale, the high-frequency information comprised in corresponding coefficient of dissociation is less, and target information is mainly included in inside HFS, and decomposition scale is that 3-5 is just passable under normal circumstances;
Step 4.2: decompose the determination in direction: one of warp wavelet and wavelet transformation is not significantly both: after bent Wave Decomposition, every layer can obtain multiple different directional information, and the number in direction can change; And after wavelet decomposition, the direction on each decomposition scale is fixing, only have three directions, i.e. level, vertical and direction, three, diagonal angle; Decomposition scale is 5, and the direction number on each decomposition scale is respectively 1,8,8,16,16; After warp wavelet, obtain the subgraph of different directions on each decomposition scale.
The present invention further provides a kind of based on warp wavelet and the SAR target detection method of adding up Wiener filtering, it is characterized in that: " to SAR image march ripple inverse transformation after filtering " described in step 7 also can carry out in the steps below:
To different directions filtered subimage march ripple inverse transformation on each decomposition scale, the optimum configurations of inverse transformation is with step 4; After march ripple inverse transformation, the SAR image of acquisition is filtered image.The filtering mainly filtering speckle noise of this step.
The present invention's advantage is compared with prior art:
(1) advantage of method design
Design twice filtering, both considered the speckle noise that SAR coherent imaging mechanism produces, considered again the noise produced by other factors, reduced the impact of various noise as far as possible.When detecting target area, enhancing process being carried out to some features of target area, effectively can overcome Small object, weak scattering target and hide target and not easily detect defect.By this new mentality of designing, greatly can reduce the impact of speckle noise, improve the signal to noise ratio (S/N ratio) of SAR image, strengthen the detailed information of target area simultaneously, be very beneficial for detection and the classification of target.
(2) advantage of data processing
Select warp wavelet to process SAR image data, this is because warp wavelet can provide the feature of multi-resolution decomposition, simultaneously on each decomposition scale, multiple different azimuth information can be provided again.This data processing method meets SAR imaging mechanism, and SAR echo data embodies anisotropy and the orientation feature to sensitivity, from different directions or multiple directions information extraction, make the result of process more accurately, more accurate.
(3) advantage of target detection
To the noise reduction process of image and the enhancing process of feature, object improves the signal to noise ratio (S/N ratio) of SAR image, therefore carries out target detection again by the SAR image after this method process, can not only improve the verification and measurement ratio of target, and the false-alarm probability of target is very low, there is strong anti-false-alarm ability.
Accompanying drawing explanation
Fig. 1: based on statistics Wiener filtering and the SAR target detection method schematic diagram of warp wavelet
Fig. 2: coherence's constant false alarm rate detection method process flow diagram
Fig. 3: original image and filter result figure
Fig. 4: SAR object detection results figure
Embodiment
Now the present invention be described further, to survey SAR image, the present invention is further elaborated in conjunction with the embodiments:
Step 1: input original SAR image.
Step 2: the statistical property calculating whole SAR image.Here the statistical property index adopted is average, accurately deviation and entropy, in fact first-order statistics characteristic, and they mainly reflect amount of image information and departure degree.
Step 3: determine the gray value differences distance between objective area in image and background clutter region, object determines inputted SAR image high s/n ratio or low SNR images.First target area is determined, then by mask or image Segmentation Technology, extract target area, calculate the average of target area subimage, i.e. localized target regional average value, the overall average of it and whole image is compared, if differ nearer, being then low SNR images, if difference is comparatively far away, is then high signal-to-noise ratio image.
Step 4: to SAR image march Wave Decomposition, decomposition scale is set to 5, by the direction difference 16,16,8,8,1 of low yardstick to high yardstick.The subimage of each decomposition scale different directions can be obtained.
Step 5: the threshold value determining first time filtering according to the relation of overall average and local mean value, in this experiment, signal to noise ratio (S/N ratio) is higher, employing formula T=μ/(stdE) calculated threshold.
Step 6: the threshold value utilizing step 5 to obtain carries out filtering process to the subimage of each decomposition scale different directions.If the gray-scale value of certain pixel is greater than threshold value T in subimage, the value of this pixel is set to 0, if the value of this pixel is less than threshold value T, then remains unchanged.
Step 7: to each subimage obtained in step 6, carry out inverse transformation by the reverse direction of warp wavelet, the SAR image of acquisition is exactly first time filtered image.
Step 8: continue to carry out filtering process to SAR image by Wiener filtering, obtain last filtering image.
Step 9: obtain each coefficient characteristics figure with warp wavelet in step 4.First SAR image is carried out multi-resolution decomposition, obtains each decomposition scale coefficient, then with certain layer coefficients for benchmark, other coefficient is zero, carries out the process of contrary flexure wave conversion, so just can obtain coefficient characteristics figure.Certainly, in processing procedure, can to certain coefficient processing.Part or all of characteristic pattern can be selected as required to carry out fusion treatment, as high-frequency characteristic figure.
Step 10: filtering figure (characteristics of low-frequency figure) and coefficient characteristics image (high-frequency characteristic figure) are simply added process, obtains and had not only inhibit noise but also the SAR image of Enhanced feature.
Step 11: select CCFAR detection method to detect target.
Constant false alarm rate (CFAR) detection is basic theory and the method for target detection, is widely used at radar and the communications field.U.S.'s Lincoln laboratory have accumulated a large amount of achievements in research in this respect, and representative is the DP-CFAR rate method based on Gaussian distribution that Novak proposes.At present, widely used object detection method is still based on CFAR detection method, it is developed so far existing many branches, as cell-average CA-CFAR (Cell-AverageCFAR), minimum selection SO-CFAR (SmallestofCFAR), Ordered Statistic OS-CFAR (OrderStatisticCFAR) and MAXIMUM SELECTION GO-CFAR (GreatestofCFAR) etc., their key distinction is different clutter average estimation.Detecting the shortcoming of target with CFAR, to be that requirement target and background clutter is distributed with obvious reflection strength poor, that is, is directly difficult to faint, little and hiding target be detected by CFAR method.First the present invention processes SAR image, strengthens the signal to noise ratio (S/N ratio) of image, proposes coherence's CFAR (CCFAR) SAR target detection method simultaneously, overcome the weakness of CFAR method from SAR imaging mechanism.
In target detection, the determination of threshold value T is the key issue of CFAR object detection method, and it and background clutter distributed model have relation closely.Suppose the probability density function (PDF) that p (x) is radar clutter distributed model, then
Visible F (x) [0 ,+∞) on be increasing function, pass through solving equation
Threshold value T can be obtained.Wherein, P
fafor false-alarm probability, different P
fadifferent T can be obtained.Background clutter distribution mainly contains: lognormality (Log-normal) distribution, Rayleigh (Rayleigh) distribution, Wei Buer (Weibull) distribution, K distribution, Gamma distribution, Pearson came (Pearson) distribution etc., wherein rayleigh distributed is the special case of Wei Buer distribution.Land clutter is generally Wei Buer distribution, and sea clutter is generally K distribution, sometimes in order to simplify calculating, adopts Gauss (Gaussian) distribution, as typical two-parameter CFAR detection method.
Step 12: obtain object detection results.By CCFAR detection method, target detection is carried out to the SAR image improving signal to noise ratio (S/N ratio), just can obtain the testing result figure of target.
Below in conjunction with accompanying drawing, test findings of the present invention is further described
Experimental data in the present invention derives from the spaceborne TerraySAR satellite of Germany, and size is 512 × 512, and resolution is 1 meter, imaging region is city and suburb, therefore, typical ground object target is urban architecture and agricultural land, belongs to the higher image of signal to noise ratio (S/N ratio) according to this experimental image of calculating.
Fig. 4 is the result of SAR target detection
Experimental data derives from the on-board SAR image of U.S.'s Lincoln laboratory, resolution 0.3 meter, and target is tank target.Wherein Fig. 4 (a) is original SAR image, the whole medium-high frequency characteristic patterns of Fig. 4 (b) for obtaining after curvilinear transformation, Fig. 4 (c) is the whole medium-high frequency characteristic patterns obtained after carrying out strengthening process third layer, the 4th layer of decomposition scale feature, and Fig. 4 (d) is by the filtered SAR image of filtering method SWCT.Fig. 4 (A), Fig. 4 (B), Fig. 4 (C) and Fig. 4 (D) are the corresponding testing result figure of Fig. 4 (a), Fig. 4 (b), Fig. 4 (c) and Fig. 4 (d) respectively.In whole testing process, be same detection method, i.e. CCFAR method, and all optimum configurations are too, but the result detected distinguishes to some extent.As can be known from Fig. 4, the Detection results of Fig. 4 (A) and Fig. 4 (D) is better, and the Detection results of Fig. 4 (B) and Fig. 4 (C) is poor.
Fig. 3 is experimental result, wherein the original SAR image of Fig. 3 (a), and Fig. 3 (b) is once filtered result, namely carries out the result of filtering in the present invention with curvilinear transformation, well inhibits speckle noise.Fig. 3 (c) receives sef-adapting filter by dimension directly SAR image to be carried out to the result of filtering, therefrom can find out, for the suppression of speckle noise, S filter is not desirable wave filter, and its filter effect will lower than result Fig. 3 (b) Suo Shi.The result obtained based on statistics Wei Na (StatisticWienerbasedTransform, the SWCT) filtering method of curvilinear transformation that Fig. 3 (d) the present invention proposes.
In order to further illustrate the validity of this method, from the angle of parameter, SWCT method is discussed below, the parameter index of selection has average, standard deviation, equivalent number, and experimental result is as shown in table 1.As can be known from Table 1, the equivalent number through the filtered SAR image of SWCT method is maximum, shows that the filter effect of the method is best.Because equivalent number is larger, the speckle noise contained by SAR image is less, otherwise the speckle noise that SAR image comprises is more, and equivalent number is less.Equally, the standard deviation of SWCT method is minimum, illustrates that singular value number relatively more outstanding in the filtered image of the method is fewer.
Table 1SWCT and other method filtering performance compare
Claims (5)
1. receive SAR imagery filtering and object detection method based on the dimension of warp wavelet, it is characterized in that: decomposed by warp wavelet and obtain multiple decomposition scale; Each decomposition scale provides different directional informations and the direction arranging each yardstick; The impact of noise is made to reach minimum by twice filtering of sampling; Abundant target information detected with lower false-alarm probability, concrete steps are as follows:
Step 1: input diameter radar image;
Step 2: the statistical property calculating input diameter radar image;
Step 3: the gray-scale relation determining interesting target region and background area;
Step 3.1: determine target area, i.e. interested region, then split area-of-interest from whole image, obtains the subimage of target area;
Step 3.2: the gray average calculating target area;
Step 3.3: the gray average of the gray average of target area and whole image is compared, determines whether target area is strong reflection region or extremely weak reflector space; The average of target area and the average of whole image are close to then thinking that the signal to noise ratio (S/N ratio) of whole image is low; The average of target area and the average of whole image differ greatly, and think that the noise of whole image is large, determine corresponding threshold value T for when carrying out filtering in step 5;
Step 4: the multiple dimensioned geometry decomposition of march ripple;
Step 5: determine filtering threshold T;
According to the relation determining target area and background in step 3, the diameter radar image namely inputted is that signal to noise ratio (S/N ratio) is high or the image that signal to noise ratio (S/N ratio) is low; If the gray-scale value of target area and background clutter gray-scale value are more or less the same, be the image that signal to noise ratio (S/N ratio) is lower, so the determination of threshold value T just calculates by formula (1).
In above formula, std represents standard deviation, and E represents negentropy, and μ represents average; If the gray-scale value of target area differs larger with background clutter gray-scale value, also namely the signal to noise ratio (S/N ratio) of image is comparatively large, and so the available formula of the determination of threshold value T (2) calculates.
Step 6: Technologies Against Synthetic Aperture Radar image carries out first time filtering process:
The result in step 4 and step 5 is utilized to process; In step
In rapid 4, obtain the subimage in directions different on different decomposition yardstick after diameter radar image decomposes, in each subimage, the threshold value utilizing step 5 to determine processes; If the value of certain pixel is greater than threshold value T in subimage, so just it is filtered out; If the value of certain pixel is less than threshold value T, so just retain it; Last result obtains filtered whole subimage;
Step 7: to diameter radar image march ripple inverse transformation after filtering;
Step 8: carry out filtering again with adaptive S filter Technologies Against Synthetic Aperture Radar image;
Step 9: obtain characteristic image by bent Wave Decomposition coefficient: be specially:
According in step 4, Technologies Against Synthetic Aperture Radar image march wave conversion decomposes the little coefficient of dissociation of acquisition different scale condition, the coefficient of certain layer is got initial value or processes, and the coefficient of other decomposition layer is zero, then carry out contrary flexure wave conversion, obtain the coefficient characteristics figure of certain layer; As required, these features selected and merge the final high frequency coefficient characteristic pattern of acquisition; Characteristic pattern comprises edge and geometric detail information;
Step 10: obtain the diameter radar image after target signature enhancing:
The characteristic pattern of filtered diameter radar image and enhancing is carried out fusion treatment, obtains not only filtering but also the diameter radar image of Enhanced feature, for inspection and the identification of target;
Step 11: select the coherence's constant false alarm rate object detection method detection method put forward from diameter radar image mechanism to detect target, obtains synthetic aperture radar target detection threshold T1.
2. according to claim 1 based on warp wavelet and the synthetic aperture radar target detection method of adding up Wiener filtering, it is characterized in that: " calculating the statistical property of input diameter radar image " described in step 2 also can carry out in the steps below:
Step 2.1: the average calculating diameter radar image.Average reflection image comprise the average case of quantity of information;
Step 2.2: calculate the standard deviation of diameter radar image, its reflection image information depart from situation;
Step 2.3: the entropy calculating diameter radar image, the overall condition of its reflection amount of image information.
3. according to claim 1 based on warp wavelet and the synthetic aperture radar target detection method of adding up Wiener filtering, it is characterized in that: " the multiple dimensioned geometry decomposition of march ripple " described in step 4 also can carry out in the steps below:
Step 4.1: the determination of decomposition scale: along with the increase of decomposition scale, the high-frequency information comprised in corresponding coefficient of dissociation is less, and target information is included in inside HFS, and decomposition scale is 3-5;
Step 4.2: decompose the determination in direction: after bent Wave Decomposition, every layer can obtain multiple different directional information and the number in direction can change, and after warp wavelet, obtains the subgraph of different directions on each decomposition scale; After wavelet decomposition, each decomposition scale is level, vertical and direction, three, diagonal angle; Decomposition scale is 5, and the direction number on each decomposition scale is respectively 1,8,8,16,16.
4. according to claim 1 based on warp wavelet and the synthetic aperture radar target detection method of adding up Wiener filtering, it is characterized in that: " to diameter radar image march ripple inverse transformation after filtering " described in step 7 also can be: to different directions filtered subimage march ripple inverse transformation on each decomposition scale, the optimum configurations of inverse transformation is with step 4; After march ripple inverse transformation, the diameter radar image of acquisition is filtered image, with filtering speckle noise.
5. according to claim 1 based on warp wavelet and the synthetic aperture radar target detection method of adding up Wiener filtering, it is characterized in that: " carrying out filtering again with adaptive S filter Technologies Against Synthetic Aperture Radar image " described in step 8 directly utilizes Wiener filtering Technologies Against Synthetic Aperture Radar image to process, and obtains final filtering diameter radar image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410691384.6A CN105205484B (en) | 2014-11-26 | 2014-11-26 | Synthetic aperture radar target detection method based on warp wavelet and Wiener filtering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410691384.6A CN105205484B (en) | 2014-11-26 | 2014-11-26 | Synthetic aperture radar target detection method based on warp wavelet and Wiener filtering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105205484A true CN105205484A (en) | 2015-12-30 |
| CN105205484B CN105205484B (en) | 2018-07-06 |
Family
ID=54953156
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410691384.6A Expired - Fee Related CN105205484B (en) | 2014-11-26 | 2014-11-26 | Synthetic aperture radar target detection method based on warp wavelet and Wiener filtering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105205484B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910178A (en) * | 2017-01-20 | 2017-06-30 | 中国人民解放军装备学院 | A kind of multi-angle SAR image fusion method based on hue statistical property sort |
| CN108604379A (en) * | 2016-02-12 | 2018-09-28 | 高通股份有限公司 | System and method for determining the region in image |
| CN108921834A (en) * | 2018-06-27 | 2018-11-30 | 肖金保 | A kind of human body foreground detection system |
| CN109658386A (en) * | 2018-11-26 | 2019-04-19 | 江苏科技大学 | A kind of sonar image pipeline inspection system and method |
| CN110111273A (en) * | 2019-04-25 | 2019-08-09 | 四川轻化工大学 | A kind of restorative procedure of image |
| CN111275250A (en) * | 2020-01-15 | 2020-06-12 | 天津大学 | Strong current region sea surface temperature forecasting method based on anisotropy |
| CN111539291A (en) * | 2020-04-16 | 2020-08-14 | 创新奇智(合肥)科技有限公司 | Target detection method and device based on radar waves, electronic equipment and storage medium |
| CN111783784A (en) * | 2020-06-30 | 2020-10-16 | 创新奇智(合肥)科技有限公司 | Method and device for detecting building cavity, electronic equipment and storage medium |
| CN112567256A (en) * | 2018-08-20 | 2021-03-26 | 谷歌有限责任公司 | Smartphone-based radar system using coherent multi-view radar processing to detect user gestures |
| CN112614065A (en) * | 2020-12-22 | 2021-04-06 | 航天信息股份有限公司 | Method and system for filtering image data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102360500A (en) * | 2011-07-08 | 2012-02-22 | 西安电子科技大学 | Treelet curvelet domain denoising- based method for change detection of remote sensing image |
| CN102722734A (en) * | 2012-05-23 | 2012-10-10 | 南京航空航天大学 | Image target identification method based on curvelet domain bilateral two-dimension principal component analysis |
| JP2013148377A (en) * | 2012-01-17 | 2013-08-01 | Mitsubishi Electric Corp | Signal processing device |
| EP2645329A1 (en) * | 2012-03-27 | 2013-10-02 | Westfälische Wilhelms-Universität Münster | Method and system for image segmentation |
-
2014
- 2014-11-26 CN CN201410691384.6A patent/CN105205484B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102360500A (en) * | 2011-07-08 | 2012-02-22 | 西安电子科技大学 | Treelet curvelet domain denoising- based method for change detection of remote sensing image |
| JP2013148377A (en) * | 2012-01-17 | 2013-08-01 | Mitsubishi Electric Corp | Signal processing device |
| EP2645329A1 (en) * | 2012-03-27 | 2013-10-02 | Westfälische Wilhelms-Universität Münster | Method and system for image segmentation |
| CN102722734A (en) * | 2012-05-23 | 2012-10-10 | 南京航空航天大学 | Image target identification method based on curvelet domain bilateral two-dimension principal component analysis |
Non-Patent Citations (5)
| Title |
|---|
| 刘代志: "一种新的SAR图像目标检测方法", 《宇航学报》 * |
| 张金良等: "曲波变换域的SAR图像相干斑去噪", 《电子测量与仪器学报》 * |
| 杨扬: "基于多尺度分析的图像融合算法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 田养军: "基于提升小波分解曲波变换的多源遥感图像融合方法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 鲁小丽等: "基于剪切波变换的SAR图像舰船检测", 《电子测量技术》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108604379A (en) * | 2016-02-12 | 2018-09-28 | 高通股份有限公司 | System and method for determining the region in image |
| CN106910178A (en) * | 2017-01-20 | 2017-06-30 | 中国人民解放军装备学院 | A kind of multi-angle SAR image fusion method based on hue statistical property sort |
| CN106910178B (en) * | 2017-01-20 | 2020-03-06 | 中国人民解放军装备学院 | Multi-angle SAR image fusion method based on tone statistical characteristic classification |
| CN108921834A (en) * | 2018-06-27 | 2018-11-30 | 肖金保 | A kind of human body foreground detection system |
| CN112567256A (en) * | 2018-08-20 | 2021-03-26 | 谷歌有限责任公司 | Smartphone-based radar system using coherent multi-view radar processing to detect user gestures |
| CN109658386A (en) * | 2018-11-26 | 2019-04-19 | 江苏科技大学 | A kind of sonar image pipeline inspection system and method |
| CN109658386B (en) * | 2018-11-26 | 2023-07-25 | 江苏科技大学 | Sonar image pipeline detection system and method |
| CN110111273B (en) * | 2019-04-25 | 2021-02-12 | 四川轻化工大学 | Image restoration method |
| CN110111273A (en) * | 2019-04-25 | 2019-08-09 | 四川轻化工大学 | A kind of restorative procedure of image |
| CN111275250A (en) * | 2020-01-15 | 2020-06-12 | 天津大学 | Strong current region sea surface temperature forecasting method based on anisotropy |
| CN111275250B (en) * | 2020-01-15 | 2022-05-17 | 天津大学 | Strong current region sea surface temperature forecasting method based on anisotropy |
| CN111539291A (en) * | 2020-04-16 | 2020-08-14 | 创新奇智(合肥)科技有限公司 | Target detection method and device based on radar waves, electronic equipment and storage medium |
| CN111539291B (en) * | 2020-04-16 | 2022-08-26 | 创新奇智(合肥)科技有限公司 | Target detection method and device based on radar waves, electronic equipment and storage medium |
| CN111783784A (en) * | 2020-06-30 | 2020-10-16 | 创新奇智(合肥)科技有限公司 | Method and device for detecting building cavity, electronic equipment and storage medium |
| CN112614065A (en) * | 2020-12-22 | 2021-04-06 | 航天信息股份有限公司 | Method and system for filtering image data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105205484B (en) | 2018-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105205484A (en) | Synthetic aperture radar target detection method based on curvelet transformation and Wiener filtering | |
| CN101727662B (en) | SAR image nonlocal mean value speckle filtering method | |
| Alonso et al. | Edge enhancement algorithm based on the wavelet transform for automatic edge detection in SAR images | |
| Huang et al. | A novel method for speckle noise reduction and ship target detection in SAR images | |
| CN103236063B (en) | Based on the SAR image oil spilling detection method of multiple dimensioned spectral clustering and decision level fusion | |
| CN107507209B (en) | Printogram extraction method of polarized SAR image | |
| CN105160648B (en) | Radar target and Shadow segmentation method based on small echo and constant false alarm rate | |
| CN101493934B (en) | Weak target detecting method based on generalized S-transform | |
| Di Martino et al. | Scattering-based SARBM3D | |
| CN106646469B (en) | SAR ship detection optimization method based on VC Method | |
| CN110889843B (en) | SAR image ship target detection method based on maximum stable extremal region | |
| CN108133468B (en) | Constant false alarm rate ship detection method based on adaptive parameter enhancement and trail auxiliary detection | |
| CN104200471A (en) | SAR image change detection method based on adaptive weight image fusion | |
| CN107784655A (en) | A kind of visual attention model SAR naval vessels detection algorithm of adaptive threshold | |
| Harant et al. | Displacement estimation by maximum-likelihood texture tracking | |
| Huang et al. | A new SAR image segmentation algorithm for the detection of target and shadow regions | |
| CN107689051A (en) | A kind of multitemporal SAR image change detecting method based on changed factor | |
| CN103413279A (en) | SAR image denoising method based on AD-NSCT algorithm | |
| Shamsoddini et al. | Edge-detection-based filter for SAR speckle noise reduction | |
| Painam et al. | A comprehensive review of SAR image filtering techniques: systematic survey and future directions | |
| CN105204010A (en) | Ground object target detection method of low signal-to-clutter ratio synthetic aperture radar image | |
| Stagliano et al. | Ship detection from SAR images based on CFAR and wavelet transform | |
| Shang et al. | Seismic random noise suppression using an adaptive nonlocal means algorithm | |
| Zhang et al. | Anisotropic-scale-space-based salient-region detection for SAR images | |
| Voccola et al. | On the relationship between the generalized likelihood ratio test and backprojection for synthetic aperture radar imaging |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180706 Termination date: 20181126 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |