CN105303548B - SAR image feature selection approach based on mixing intelligent optimizing algorithm - Google Patents
SAR image feature selection approach based on mixing intelligent optimizing algorithm Download PDFInfo
- Publication number
- CN105303548B CN105303548B CN201510296347.XA CN201510296347A CN105303548B CN 105303548 B CN105303548 B CN 105303548B CN 201510296347 A CN201510296347 A CN 201510296347A CN 105303548 B CN105303548 B CN 105303548B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- particle
- 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
Links
Classifications
-
- 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
Abstract
The invention discloses a kind of SAR image feature selection approach based on mixing intelligent optimizing algorithm.The present invention is strengthened SAR image using fractal characteristic first, and a kind of azimuth method of estimation based on image moment is proposed based on the image after segmentation.It is then based on the image after not correcting and correcting and extracts Zernike squares, Gabor wavelet coefficient and gray level co-occurrence matrixes composition candidate feature sequence respectively.Employ a kind of hybrid optimization algorithm of combination genetic algorithm and two-value population and realize SAR image feature selecting.MSTAR database authentications are finally used to propose the validity of algorithm.Test result indicates that the characteristic set after optimization has certain generalization ability, the accuracy rate of SAR target identifications is on the one hand improved, on the other hand reduces the time of SAR image target identification.
Description
Technical field
The invention belongs to target identification and mode identification technology, it is related to a kind of based on mixing intelligent optimizing algorithm
SAR image feature selection approach.
Background technology
Synthetic aperture radar (Synthetic Aperture Radar, SAR) is a kind of microwave imaging sensor, is had complete
It when, round-the-clock, multiband, multipolarization the features such as, be widely used in national economy and national defense construction, such as rebound road lead
The system of defense of bullet, marine monitoring system, mineral reserve detection etc..
Influence the key factor bag of SAR image automatic target detection (Automatic Target Recognition, ATR)
Include feature extraction and feature selecting, and the aspect of classifier design two.The feature that SAR ATR are used at present is mainly included based on number
Learn feature, computer vision feature and the electromagnetism top grade of conversion.Feature based on mathematic(al) manipulation has wavelet transformation, PCA, ICA etc..
Because this category feature generally has higher object recognition rate, it typically can be used directly or multiple features merged.Calculate
Machine visual signature main textured, attitude angle, shape, fractal dimension, primary edge etc.；Common electromagnetic signature has in scattering
The heart, HRR sections etc..The target that this two category feature can be corresponded in image scene, because single feature classifying quality is poor, typically
By using multiple combinations of features to improve discriminating power.Can exist for the multiple features of SAR image, between them
The redundancy issue of feature set, feature set cross adjustment, the On The Choice of real-time notable feature.Feature selecting is that solve multiple spies
Levy the means of simultaneous selection, it is therefore an objective to maximally effective combinations of features is filtered out in candidate feature.
The content of the invention
In view of the deficiencies of the prior art, the present invention provides a kind of SAR image feature based on mixing intelligent optimizing algorithm
System of selection.
The present invention's comprises the concrete steps that：
Step (1) is pre-processed
1.1 SAR images strengthen
Using fractal characteristic, by setting suitable out to out εmax, each pixel in original SAR image is carried out
MFFK is calculated, and MFFK images corresponding to generation, realizes the targets improvement in SAR image；
Wherein MFFK is represented in εmaxRange scale in D dimensions area K intensity of variation.
1.2 SAR images are split
Row threshold division is entered to image after enhancing, obtains the binary image of target and background separation.
1.3 azimuthal estimations, the posture correction of image
To SAR segmentation figure pictures, using the azimuth of the constant moments estimation targets of Hu, calculation formula is
Wherein M02、M20、M11For second moment, M00For zeroth order square, (xc,yc) represent image barycenter.According to the orientation of estimation
Angle carries out posture correction to original SAR image.
1.4 cut, centralization
Read 61 × 61 image-regions centered on image center respectively to original SAR image and SAR segmentation figures picture.
Step (2) feature extractions
The 2.1 extraction rectangular-shaped features of Zernike
Segmentation figure picture based on original SAR image and correction SAR image extracts the 34 rectangular-shaped features of dimension Zernike respectively.
2.2 extraction Zernike square amplitude Characteristics
Segmentation figure picture based on original SAR image and correction SAR image extracts 34 dimension Zernike square amplitude Characteristics respectively.
2.3 extraction orientation corner characteristics
To SAR segmentation figure pictures, by the use of Hu, bending moment does not calculate the azimuth of target and is used as orientation corner characteristics.
2.4 extraction gray level co-occurrence matrixes features
Segmentation figure picture based on original SAR image and correction SAR image calculates gray level co-occurrence matrixes feature respectively, by energy
Amount, entropy, the moment of inertia, the average of correlation and standard deviation are as 8 dimension textural characteristics.
2.4 extraction Gabor textural characteristics
Gabor filter has very strong space orientation and set direction, based on original SAR image and correction SAR figures
The segmentation figure picture of picture, SAR image 16 is obtained by one group of Gabor wavelet and ties up Local textural feature.
Step (3) feature selectings
3.1 particles and chromosome coding
Coded system uses binary coding, and 1 expression this feature is selected, and 0 represents not to be selected.
3.2 fitness functions design
Fitness function is from the aspect of object recognition rate and recognition time two, using the fitness of the method acquisition of weight
Function such as formula (3) represents
Fitness=a × AC+b × (1-L0/L) (3)
Wherein AC represents the discrimination of current subsequence, L0The Characteristic Number of current subsequence is represented, L is characterized total
Number, weight coefficient a, b value are respectively 0.8,0.2.
3.3 mixing intelligent optimizing algorithms
3.3.1 particle is initialized
Position and the speed of N number of particle are initialized using random device.
3.3.2 select outstanding particle
Sorted according to the fitness function value of N number of particle, using N/2 high particle of fitness function value as outstanding particle
Retain, N/2 particle is given up in addition.
3.3.3 particle updates
The position of the N/2 particle retained first using the renewal of NBPSO algorithms and speed, the speed more new formula of particle are
WhereinRepresent that the position of particle is changed into 1,0 probability respectively, work as PibstOr PgbstDuring equal to 0,Increase
Add,Reduce；Conversely, work as PibstOr PgbstDuring equal to 1,Reduce,Increase, in this way, a certain position of particle
Change 1 and table 0 direction can keep and for particle renewal.The location updating formula of particle is
Wherein v 'ij(t)=sig (vij(t)),Represent xij(t) negating under binary system, rijIt is between (0,1)
Random value.
Then carry out GA to new particle to operate to obtain N/2 particle, finally by the renewal particle combinations obtained twice under
The N number of particle of a generation, this completes the renewal of whole population.
3.3.4 iteration judges
Judge whether to reach maximum iteration, stop iteration if meeting, otherwise continue step 3.3.2 and step
3.3.3。
The present invention extracts Zernike squares, Gabor wavelet coefficient and gray scale based on the image after not correcting and correcting and is total to respectively
Raw matrix forms candidate feature sequence.The hybrid optimization algorithm for employing a kind of combination genetic algorithm and two-value population is realized
SAR image feature selecting.Test result indicates that the characteristic set after optimization has certain generalization ability, SAR is on the one hand improved
The accuracy rate of target identification, on the other hand reduce the time of SAR image target identification.
Brief description of the drawings
Fig. 1 is the flow chart of the present invention.
Fig. 2 (a) is target T72 original graph.
Fig. 2 (b) is target T72 enhancing figure.
Fig. 2 (c) is target T72 segmentation figure.
Fig. 2 (d) is target T72 correction figure.
Fig. 3 is the flow chart of mixing intelligent optimizing algorithm.
Embodiment
Below in conjunction with accompanying drawing, the invention will be further described.
As shown in figure 1, the present invention comprises the following steps：
Step (1) is pre-processed
1.1 SAR images strengthen
Fractal model can the surface with natural forms or space structure structure phase well in the range of some scale
It coincide, and certain otherness be present with the surface of man-made target or space structure in this regular expression of fractal model.
According to the difference of fractal characteristic, the man-made target in natural background is extracted to realize the enhancing to man-made target.Multi-scale Fractal
Feature (multi-scale fractal feature related with K, MFFK) is a fractal parameter measure of variation letter
Number, MFFK can be understood as in εmaxRange scale in D dimension area (K) intensity of variation, using MFFK come realize protrusion it is artificial
Target and difference of the natural background on fractal characteristic.By setting suitable out to out εmax, as shown in Fig. 2 (a), to original
Each pixel carries out MFFK calculating, and MFFK images corresponding to generation in beginning SAR image, realizes the targets improvement in SAR image,
As shown in Fig. 2 (b)；
1.2 SAR images are split
Enter row threshold division to image after enhancing, threshold value 5, obtain the binary image of target and background separation, such as Fig. 2
(c) shown in.
1.3 azimuthal estimations, the posture correction of image
To SAR segmentation figure pictures, using the azimuth of the constant moments estimation targets of Hu, calculation formula is
Wherein M02、M20、M11For second moment, M00For zeroth order square, (xc,yc) represent image barycenter.According to the orientation of estimation
Angle carries out posture correction to original SAR image, and Fig. 2 (d) shows the image after being corrected according to the azimuth of estimation.
1.4 cut, centralization
Read 61 × 61 image-regions centered on image center respectively to original SAR image and SAR segmentation figures picture.
Step (2) feature extractions
The 2.1 extraction rectangular-shaped features of Zernike
Zernike squares are projection of the piece image on one group of Zernike multinomial.Teague etc. is with complex domain Zernike
Multinomial is base, has obtained having orthogonal, invariable rotary characteristic Zernike squares [12].One width discrete picture I n ranks m weights
Zernike squares may be defined as
Wherein n=0,1,2 ..., 0≤| m |≤n, and n- | m | it is even number.(ρ, θ) is the polar coordinates performance under unit circle
Form, Rn,mIt is radial polynomial.Zernike squares have two it is critically important the characteristics of：(1) although Zernike squares are dependent on target
Centralization is translated, but amplitude has rotational invariance, i.e., and amplitude does not change piece image after rotation, therefore can use
The shape facility of the magnitude extraction SAR image of Zernike squares；(2)Rn,mIt is mutually orthogonal, can be in the case of the shape of target be ignored
Amplitude Characteristics from region of interesting extraction Zernike squares as target.
Segmentation figure picture based on original SAR image and correction SAR image extracts the 34 rectangular-shaped features of dimension Zernike respectively.
2.2 extraction Zernike square amplitude Characteristics
Segmentation figure picture based on original SAR image and correction SAR image extracts 34 dimension Zernike square amplitude Characteristics respectively.
2.3 extraction orientation corner characteristics
To SAR segmentation figure pictures, by the use of Hu, bending moment does not calculate the azimuth of target and is used as orientation corner characteristics.
2.4 extraction gray level co-occurrence matrixes features
Gray level co-occurrence matrixes are defined with the joint probability density of two position pixels, and it is bright that it not only reflects image
The distribution character of degree, it also reflects the position distribution characteristic having in image between same brightness or the pixel of similar brightness, base
Gray level co-occurrence matrixes feature is calculated respectively in the segmentation figure picture of original SAR image and correction SAR image, by energy, entropy, inertia
Square, related average and standard deviation are as 8 dimension textural characteristics.
2.4 extraction Gabor textural characteristics
Gabor filter has very strong space orientation and set direction, based on original SAR image and correction SAR figures
The segmentation figure picture of picture, SAR image 16 is obtained by one group of Gabor wavelet and ties up Local textural feature, Gabor function medium wavelengths are set to
1.5th, two parameters of Gaussian function are all set to 0.5.
Step (3) feature selectings
3.1 particles and chromosome coding
Coded system uses binary coding, and 1 expression this feature is selected, and 0 represents not to be selected.
3.2 fitness functions design
Fitness function is from the aspect of object recognition rate and recognition time two, using the fitness of the method acquisition of weight
Function such as formula (4) represents
Fitness=a × AC+b × (1-L0/L) (4)
Wherein AC represents the discrimination of current subsequence, L0The Characteristic Number of current subsequence is represented, L is characterized total
Number, weight coefficient a, b value are respectively 0.8,0.2.
3.3 carry out SAR image feature selecting based on mixing intelligent optimizing algorithm, referring to Fig. 3.
3.3.1 particle is initialized
Position and the speed of N number of particle are initialized using random device.
3.3.2 select outstanding particle
Sorted according to the fitness function value of N number of particle, using N/2 high particle of fitness function value as outstanding particle
Retain, N/2 particle is given up in addition.
3.3.3 particle updates
The position of the N/2 particle retained first using the renewal of NBPSO algorithms and speed, the speed more new formula of particle are
WhereinRepresent that the position of particle is changed into 1,0 probability respectively, work as PibstOr PgbstDuring equal to 0,Increase
Add,Reduce；Conversely, work as PibstOr PgbstDuring equal to 1,Reduce,Increase, in this way, a certain position of particle
Change 1 and table 0 direction can keep and for particle renewal.The location updating formula of particle is
Wherein v 'ij(t)=sig (vij(t)),Represent xij(t) negating under binary system, rijIt is between (0,1)
Random value.
Then carry out GA to new particle to operate to obtain N/2 particle, finally by the renewal particle combinations obtained twice under
The N number of particle of a generation, this completes the renewal of whole population.
3.3.4 iteration judges
Judge whether to reach maximum iteration, stop iteration if meeting, otherwise continue step 3.3.2 and step
3.3.3。
Claims (1)
1. the SAR image feature selection approach based on mixing intelligent optimizing algorithm, it is characterised in that this method comprises the concrete steps that：
Step (1) is pre-processed
1.1SAR image enhaucament
Using fractal characteristic, by setting suitable out to out εmax, MFFK meters are carried out to each pixel in original SAR image
Calculate, and MFFK images corresponding to generation, realize the targets improvement in SAR image；
<mrow>
<mi>M</mi>
<mi>F</mi>
<mi>F</mi>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>&epsiv;</mi>
<mo>=</mo>
<mn>2</mn>
</mrow>
<msub>
<mi>&epsiv;</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
</munderover>
<msup>
<mrow>
<mo>&lsqb;</mo>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>&epsiv;</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mrow>
<msub>
<mi>&epsiv;</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>&epsiv;</mi>
<mo>=</mo>
<mn>2</mn>
</mrow>
<msub>
<mi>&epsiv;</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
</munderover>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>&epsiv;</mi>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein MFFK is represented in εmaxRange scale in D dimensions area K intensity of variation；
1.2SAR images are split
Row threshold division is entered to image after enhancing, obtains the binary image of target and background separation；
1.3 azimuthal estimations, the posture correction of image
To SAR segmentation figure pictures, using the azimuth of the constant moments estimation targets of Hu, calculation formula is
<mrow>
<mi>&theta;</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msup>
<mi>tan</mi>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<mn>2</mn>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>M</mi>
<mn>11</mn>
</msub>
<msub>
<mi>M</mi>
<mn>00</mn>
</msub>
</mfrac>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>c</mi>
</msub>
<msub>
<mi>y</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>M</mi>
<mn>20</mn>
</msub>
<msub>
<mi>M</mi>
<mn>00</mn>
</msub>
</mfrac>
<mo>-</mo>
<msup>
<msub>
<mi>x</mi>
<mi>c</mi>
</msub>
<mn>2</mn>
</msup>
<mo>)</mo>
<mo>-</mo>
<mo>(</mo>
<mfrac>
<msub>
<mi>M</mi>
<mn>02</mn>
</msub>
<msub>
<mi>M</mi>
<mn>00</mn>
</msub>
</mfrac>
<mo>-</mo>
<msup>
<msub>
<mi>y</mi>
<mi>c</mi>
</msub>
<mn>2</mn>
</msup>
<mo>)</mo>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
</mrow>
<mn>2</mn>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein M02、M20、M11For second moment, M00For zeroth order square, (xc,yc) represent image barycenter；According to the azimuth pair of estimation
Original SAR image carries out posture correction；
1.4 cut, centralization
Read 61 × 61 image-regions centered on image center respectively to original SAR image and SAR segmentation figures picture；
Step (2) feature extractions
The 2.1 extraction rectangular-shaped features of Zernike
Segmentation figure picture based on original SAR image and correction SAR image extracts the 34 rectangular-shaped features of dimension Zernike respectively；
2.2 extraction Zernike square amplitude Characteristics
The segmentation figure picture of segmentation figure picture and correction SAR image based on original SAR image extracts 34 dimension Zernike square amplitudes respectively
Feature；
2.3 extraction orientation corner characteristics
To SAR segmentation figure pictures, by the use of Hu, bending moment does not calculate the azimuth of target and is used as orientation corner characteristics；
2.4 extraction gray level co-occurrence matrixes features
Based on original SAR image and correction SAR image segmentation figure picture calculate gray level co-occurrence matrixes feature respectively, by energy, entropy,
The moment of inertia, related average and standard deviation are as 8 dimension textural characteristics；
2.4 extraction Gabor textural characteristics
Gabor filter has very strong space orientation and set direction, based on original SAR image and corrects SAR image
Segmentation figure picture, SAR image 16 is obtained by one group of Gabor wavelet and ties up Local textural feature；
Step (3) feature selectings
3.1 particles and chromosome coding
Coded system uses binary coding, and 1 expression this feature is selected, and 0 represents not to be selected；
3.2 fitness functions design
Fitness function is from the aspect of object recognition rate and recognition time two, using the fitness function of the method acquisition of weight
As formula (3) represents
Fitness=a × AC+b × (1-L0/L) (3)
Wherein AC represents the discrimination of current subsequence, L0The Characteristic Number of current subsequence is represented, L is characterized total number, weight
Coefficient a, b value are respectively 0.8,0.2；
3.3 mixing intelligent optimizing algorithms
3.3.1 particle is initialized
Position and the speed of N number of particle are initialized using random device；
3.3.2 select outstanding particle
Sorted according to the fitness function value of N number of particle, N/2 high particle of fitness function value is protected as outstanding particle
Stay, N/2 particle is given up in addition；
3.3.3 particle updates
The position of the N/2 particle retained first using the renewal of NBPSO algorithms and speed, the speed more new formula of particle are
WhereinRepresent that the position of particle is changed into 1,0 probability respectively, work as PibstOr PgbstDuring equal to 0,Increase,
Reduce；Conversely, work as PibstOr PgbstDuring equal to 1,Reduce,Increase, in this way, the change 1 of a certain position of particle
It can keep with the direction of table 0 and for the renewal of particle；The location updating formula of particle is
Wherein v 'ij(t)=sig (vij(t)),Represent xij(t) negating under binary system, rijBe between (0,1) with
Machine value；
Then carry out GA to new particle to operate to obtain N/2 particle, finally by the renewal particle combinations obtained twice into N of future generation
Individual particle, this completes the renewal of whole population；
3.3.4 iteration judges
Judge whether to reach maximum iteration, stop iteration if meeting, otherwise continue step 3.3.2 and step
3.3.3。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510296347.XA CN105303548B (en) | 2015-06-02 | 2015-06-02 | SAR image feature selection approach based on mixing intelligent optimizing algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510296347.XA CN105303548B (en) | 2015-06-02 | 2015-06-02 | SAR image feature selection approach based on mixing intelligent optimizing algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105303548A CN105303548A (en) | 2016-02-03 |
| CN105303548B true CN105303548B (en) | 2017-12-22 |
Family
ID=55200772
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510296347.XA Active CN105303548B (en) | 2015-06-02 | 2015-06-02 | SAR image feature selection approach based on mixing intelligent optimizing algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105303548B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111323815B (en) * | 2020-02-17 | 2021-04-02 | 成都理工大学 | Method for predicting carbonate rock fracture reservoir based on azimuth gray level co-occurrence matrix |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593674A (en) * | 2013-11-19 | 2014-02-19 | 太原理工大学 | Cervical lymph node ultrasonoscopy feature selection method |
| CN103678680A (en) * | 2013-12-25 | 2014-03-26 | 吉林大学 | Image classification method based on region-of-interest multi-element spatial relation model |
| CN104537108A (en) * | 2015-01-15 | 2015-04-22 | 中国矿业大学 | High-dimensional data feature selecting method |
-
2015
- 2015-06-02 CN CN201510296347.XA patent/CN105303548B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593674A (en) * | 2013-11-19 | 2014-02-19 | 太原理工大学 | Cervical lymph node ultrasonoscopy feature selection method |
| CN103678680A (en) * | 2013-12-25 | 2014-03-26 | 吉林大学 | Image classification method based on region-of-interest multi-element spatial relation model |
| CN104537108A (en) * | 2015-01-15 | 2015-04-22 | 中国矿业大学 | High-dimensional data feature selecting method |
Non-Patent Citations (7)
| Title |
|---|
| Genetic algorithm based feature selection for target detection in SAR images;Bhanu B 等;《Image and Vision Computing》;20031231;第21卷(第7期);第591-608页 * |
| 利用SVM的极化SAR图像特征选择与分类;吴永辉 等;《电子与信息学报》;20081031;第30卷(第10期);第2347-2351页 * |
| 利用特征选择自适应决策树的层次SAR图像分类;何楚 等;《武汉大学学报·信息科学版》;20120131;第37卷(第1期);第46-49页 * |
| 基于Zernike矩的字符特征选择;李了了;《内江师范学院学报》;20080630;第23卷(第6期);第49-51页 * |
| 基于复合Zernike矩相角估计的图像配准;易盟 等;《光学精密工程》;20120531;第20卷(第5期);第1117-1125页 * |
| 基于遗传算法的SAR图像目标鉴别特征选择;高贵 等;《电子学报》;20080630;第36卷(第6期);第1041-1046页 * |
| 基于随机复杂度约束的高维特征自动选择算法;刘峤 等;《电子学报》;20110228;第39卷(第2期);第370-374页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105303548A (en) | 2016-02-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gao et al. | Automatic change detection in synthetic aperture radar images based on PCANet | |
| CN105184309B (en) | Classification of Polarimetric SAR Image based on CNN and SVM | |
| CN103218831B (en) | A kind of video frequency motion target classifying identification method based on profile constraint | |
| CN101814144B (en) | Water-free bridge target identification method in remote sensing image | |
| CN102346850B (en) | DataMatrix bar code area positioning method under complex metal background | |
| CN110097093A (en) | A kind of heterologous accurate matching of image method | |
| CN109508710A (en) | Based on the unmanned vehicle night-environment cognitive method for improving YOLOv3 network | |
| CN102902956B (en) | A kind of ground visible cloud image identifying processing method | |
| CN104598878A (en) | Multi-modal face recognition device and method based on multi-layer fusion of gray level and depth information | |
| CN104102920A (en) | Pest image classification method and pest image classification system based on morphological multi-feature fusion | |
| CN109858368A (en) | A kind of recognition of face attack defense method based on Rosenbrock-PSO | |
| CN113963222B (en) | High-resolution remote sensing image change detection method based on multi-strategy combination | |
| CN102147867B (en) | Method for identifying traditional Chinese painting images and calligraphy images based on subject | |
| CN103246894B (en) | A kind of ground cloud atlas recognition methods solving illumination-insensitive problem | |
| CN103927511A (en) | Image identification method based on difference feature description | |
| CN104392462B (en) | A kind of SAR image registration method based on significantly segmentation subregion pair | |
| CN103984963B (en) | Method for classifying high-resolution remote sensing image scenes | |
| CN104537353A (en) | Three-dimensional face age classifying device and method based on three-dimensional point cloud | |
| CN106503748A (en) | A kind of based on S SIFT features and the vehicle targets of SVM training aids | |
| CN105160686A (en) | Improved scale invariant feature transformation (SIFT) operator based low altitude multi-view remote-sensing image matching method | |
| CN102903109A (en) | Integrated partition registering method of optical image and synthetic aperture radar (SAR) image | |
| CN104143091A (en) | Single-sample face recognition method based on improved mLBP | |
| CN104463878A (en) | Novel depth image local descriptor method | |
| CN105303548B (en) | SAR image feature selection approach based on mixing intelligent optimizing algorithm | |
| CN110458019A (en) | The waterborne target detection method for excluding inverted image interference under rare cognition sample conditions |
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 |