CN108038856A - Based on the infrared small target detection method for improving Multi-scale Fractal enhancing - Google Patents
Based on the infrared small target detection method for improving Multi-scale Fractal enhancing Download PDFInfo
- Publication number
- CN108038856A CN108038856A CN201711403183.1A CN201711403183A CN108038856A CN 108038856 A CN108038856 A CN 108038856A CN 201711403183 A CN201711403183 A CN 201711403183A CN 108038856 A CN108038856 A CN 108038856A
- Authority
- CN
- China
- Prior art keywords
- scale
- formula
- image
- pixel
- target
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw 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
-
- 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
- G06T2207/10048—Infrared image
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to based on the infrared small target detection method for improving Multi-scale Fractal enhancing.Method of Target Detection in Infrared of the tradition based on fractal characteristic enhancing has the problem of computation complexity is high.The difference of the present invention by the use of the maxima and minima of pixel in multiple dimensioned region is measured as one kind, then the Multi-scale Fractal vector of every pixel is obtained using its region averages, the conspicuousness of every pixel is evaluated by the new significance measure criterion of definition, enhanced image is finally based on and target detection is carried out by auto-thresholding algorithm.Method proposed by the present invention reduces the calculation amount of algorithm, improves detection speed.
Description
Technical field
The invention belongs to infrared small target detection field, is related to a kind of based on the infrared small of improvement Multi-scale Fractal enhancing
Object detection method.
Background technology
Small IR targets detection is one of key technology of Infra-Red Search ＆ Track System, can effectively improve monitoring model
Enclose, play an important role in fields such as navigation, air defense and security monitorings.There are two difficult points for small IR targets detection：(1)
Small object does not have obvious texture and style characteristic；(2) influenced by target context radiation and imaging sensor, infrared image
In there are random noise and a large amount of clutters, signal noise ratio (snr) of image are low.Although existing big quantity algorithm proposes have under complex environment
Detection Small object in effect ground is still not yet be fully solved the problem of.
Small IR target detection is broadly divided into two classes：Detection (DBT) and the preceding tracking (TBD) of detection before tracking.
TBD technologies are accumulated using multiple image to detect weak signal target, and DBT is to utilize mesh in the first two field picture that target occurs
Detection algorithm positioning target is marked, then utilizes tracking technique estimation target using space-time consistency of the target in consecutive image
Position.Priori of the Small object without master goal is detected in single image, because its computational efficiency is higher, is usually used in
Some detection devices.Past 20 years, although much research focused on DBT technologies, it is proposed that some effective infrared small targets
Detection algorithm, such as TopHat Filtering, MaxMean and MaxMedian etc..
At present, the algorithm that researcher proposes is that the grey-scale contrast based on target and its surrounding environment is more than the back of the body mostly
The contrast of scene area this it is assumed that strengthen first input picture, then split using threshold value, obtained to be checked
Survey the candidate region of target.According to Fractal Geometry Theory, because natural target and man-made target have different immanent structures, divide
Shape model is more suitable for natural target of the scales such as mountain, cloud, water, plant within certain particular range, but is not suitable for artificial
Target, therefore fractal theory can be based on and design effective Multi-scale Fractal, first image is strengthened, is then based on adaptive
Threshold segmentation method is answered to realize the detection of infrared small target.For example, average gray difference maximum absolute value mapping algorithm
(AGADMM) using every pixel coordinate as central point, the target area of different scale is defined, while define a scale bigger
Background area, commented by calculating the maximum of each target area and the poor absolute value of background field pixel grey scale average
Estimate the conspicuousness of the pixel, image enhancement is carried out in this, as measurement, based on enhanced image using adaptive threshold point
Cut algorithm and realize target detection.
The content of the invention
The present invention considers that Method of Target Detection in Infrared computation complexity of the tradition based on fractal characteristic enhancing is high and asks
Topic, by analyzing existing algorithm, it is proposed that a kind of new significance measure criterion based on Multi-scale Fractal,
Then algorithm is simplified, devised a kind of based on the infrared small target detection method for improving Multi-scale Fractal enhancing, reduction
The calculation amount of algorithm, on the premise of target detection rate is ensured real-time gets a promotion.The present invention is using in multiple dimensioned region
The difference of the maxima and minima of pixel is as a kind of more rulers measured, every pixel is then obtained using its region averages
Fractal characteristic vector is spent, the conspicuousness of every pixel is evaluated by the new significance measure criterion of definition, is finally based on increasing
Image after strong carries out target detection by auto-thresholding algorithm.The algorithm of target detection of design is only needed with simple
Arithmetic can obtain enhanced image, algorithm is realized simple, is very suitable for Embedded Application, is ensureing to detect
The real-time of algorithm is improved on the premise of rate.
The technical solution adopted by the present invention comprises the following steps：
Step (1) obtains the Multi-scale Fractal vector of every pixel in original infrared image I '.
Step (2) calculates every picture based on Multi-scale Fractal vector by the new significance measure criterion of definition
The conspicuousness of element, image after being strengthened.
Step (3) is based on image after enhancing, Target Segmentation is carried out using auto-thresholding algorithm, after obtaining detection
Target.
Compared with prior art, the present invention its remarkable advantage is：(1), it is necessary to the parameter of setting during progress image enhancement
It is few, it is only necessary to which that the out to out for calculating Multi-scale Fractal is set.With the increase of out to out, target in image after enhancing
Broadening effect it is more obvious, thus be more advantageous to the visual and Threshold segmentation of Small object；(2) target partial zones are effectively utilized
The contrast information in domain, can detect most of target in the case of there are less false-alarm, and the method for design can be examined at the same time
Measure the bright target in image and dark target.(3) the method is applicable not only to small target deteection, by designing rational scale
And detection threshold value, it is equally applicable to big target detection, including the SAR image with stronger coherent speckle noise.(4) traditional algorithm
Complexity is calculated, and inventive algorithm has pertained only to arithmetic, it is convenient low in Implementation of Embedded System, algorithm calculation amount, in real time
Property is good.
Brief description of the drawings
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is influence of the out to out to infrared image enhancement effect；
Fig. 3 is based on the infrared image enhancement and testing result (Small object) for improving Multi-scale Fractal enhancing；Wherein (a)-
(f) it is the infrared image under different scenes；
Fig. 4 is based on the infrared image enhancement and testing result (big target) for improving Multi-scale Fractal enhancing；Wherein (a)-
(f) it is the infrared image under different scenes；
Fig. 5 is based on the SAR image enhancing and testing result (big target) for improving Multi-scale Fractal enhancing；Wherein (a)-(d)
For the infrared image under different scenes；
The enhancing effect of Fig. 6 and other five kinds of typical IRs small target deteection algorithms contrast；Wherein (1)-(6) are difference
Infrared image under scene；(a)-(d) is the targets improvement effect of original image and different detection algorithms.
Embodiment
Below in conjunction with attached drawing, the invention will be further described.
As shown in Figure 1, specific implementation step of the present invention is as follows：
Step (1) obtains the Multi-scale Fractal vector of every pixel in original infrared image I '.It is specific as follows：
According to Fractal Geometry Theory, the relation of fractal measure and scale can be described as by Richardson laws：
M (ε)=K εd-FDFormula (1)
Wherein, ε represents scale, and ε=1,2 ..., M (ε) represent estimating under scale ε, and FD and d represent fractal dimension respectively
And topological dimension, K are fractal parameter.It can be described as 2-D gray image：
A (x, y, ε)=K (x, y, ε) ε2-FD(x,y,ε)Formula (2)
Wherein A (x, y, ε) represents that the surface area on the gradation of image surface under scale ε is estimated.
When the scale of measurement is respectively ε1,ε2When, it can be obtained by formula (2),
logA(x,y,ε1)=(2-FD (x, y, ε1))log(ε1)+logK(x,y,ε1) formula (3)
logA(x,y,ε2)=(2-FD (x, y, ε2))log(ε2)+logK(x,y,ε2) formula (4)
Understood according to Fractal Geometry Theory, preferably divide shape for one, its fractal dimension FD is unrelated with all scales
Amount, be a constant all the time.Thus, it is supposed that under the different scale of measurement, the FD in formula (3) and (4) is constant, and
If ε1=ε, ε2=ε -1, then can be obtained by formula (3) and (4), when the scale of measurement is ε, its corresponding D dimension area K (x, y, ε)
It is represented by
When calculating A (x, y, ε) using covering blanket method, have
Wherein, V (x, y, ε) is the volume at coordinate (x, y) place under scale ε, shown in calculation formula such as formula (7), U (x, y, ε)
It is respectively the maximum and minimum value of pixel in coordinate (x, y) place contiguous range under the scale with B (x, y, ε).
Formula (6) is substituted into formula (5), is obtained
Carry out analysis to formula (7) and formula (8) to can be seen that under different scale, the body of homogeneous every pixel in background area
Product V is small, close to 0；And in nonhomogeneous background area, although the value of V is larger, due to the fluctuating of background, therefore under different scale
Its change rate is also relatively large, i.e., Part II absolute value is big in formula (8), therefore K values reduce；In target area, either bright mesh
Mark or dark target, the maximum or minimum value of its regional area are fixed, and V is maximum, and its change rate is small, therefore in target area K
Value is maximum.
Analyzed based on more than, if ignoring rear two parts in formula (8), have K (x, y, ε)=V (x, y, ε -1), at this time may be used
The difference of the maximum of pixel and minimum value adds up and is used as the region centered on coordinate points (x, y) in by the use of a region
Significance measure.
Analyzed based on more than, if scale is ε=2 ... εmax, it is original infrared image as ε=1, εmaxFor out to out.
The pixel coordinate of any is (x, y) in former infrared image I ', passes through picture in the coordinate points contiguous range under formula (9) calculating scale ε
The difference of element value maxima and minima, as contrast of the pixel under scale ε, obtains contrast image I；To disappear
Except the influence for the cumulative effect that different scale is brought, average contrast of the every pixel under scale ε is calculated by formula (10),
Obtain the fractal characteristic vector under scale ε；And then using scale ε as variable, acquisition Multi-scale Fractal vector t (x, y,:).
Step (2) be based on Multi-scale Fractal vector t (x, y,:) calculate every pixel conspicuousness, strengthened
Image afterwards.It is specific as follows：
Based on the analysis to formula (8), for target area, the Multi-scale Fractal vector average that formula (10) defines is big,
Change at the same time small, and for background area, especially non-homogeneous background area, although its average is also larger, it changes
Greatly, therefore using formula (11) come the significance measure value at coordinates computed point (x, y) place, image after being strengthened, is denoted as E (x, y).
E (x, y)=mean (t (x, y,:))2-std(t(x,y,:))2Formula (11)
Wherein mean () and std () is respectively the function averaged with standard deviation.
Step (3) is based on image after enhancing, Target Segmentation is carried out using auto-thresholding algorithm, after obtaining detection
Target.It is specific as follows：
Target detection is carried out using auto-thresholding algorithm, to eliminate the influence of boundary effect, splits threshold calculating
During value, 2 times of out to out ε are removedmaxBorderline region, specific formula for calculation is as follows：
μ=mean (E (2 × εmax+1:rows-2×εmax,2×εmax+1:cols-2×εmax)) formula (12)
δ=std (E (2 × εmax+1:rows-2×εmax,2×εmax+1:cols-2×εmax)) formula (13)
PSR=(255- μ)/δ formulas (14)
T=c × PSR × δ+μ formulas (15)
D (x, y)=E (x, y) >=T formulas (16)
Wherein, μ is that image removes the pixel average after fringe region after strengthening, and pixel is divided after δ removes border for image after enhancing
The standard deviation of cloth；PSR is peak sidelobe ratio；C is division coefficient, and for small target deteection, value range may be set to [0.5
0.65], for big target detection, value range may be set to [0.15 0.45].Rows and cols is original infrared image I '
Height and width；T is the threshold value for target detection, and D is to detect the pixel for belonging to target.
To verify effectiveness of the invention, the out to out ε first by experimental analysismaxTo inputting infrared image enhancement
The influence of effect, from figure 2 it can be seen that with the increase of scale, target is imitated with obvious broadening in image after enhancing
Should, and out to out is bigger, broadening effect is more obvious, and target contrast gets a promotion in infrared image after enhancing, improvement of visual effect
Enhancing is obvious.The performance of this method is tested using 6 groups of infrared images with Small object, strengthens image and testing result such as Fig. 3 institutes
Show.From figure 3, it can be seen that in addition to bright target is detected, suspicious dark target can be equally detected.Work as scene
In there are during big target, by selecting suitable division coefficient c, be set as c=0.35, for infrared image and SAR images
Object detection results as shown in Figure 4 and Figure 5.Even if SAR image has stronger coherent speckle noise, the algorithm of proposition is effective
Ground utilizes the contrast information of target regional area, and most of target can be detected in the case of there are less false-alarm.
Fig. 6 is proposition method of the present invention and TopHat Filtering, MaxMean, MaxMedian, AGADMM and NWIE
Deng the targets improvement Contrast on effect of five kinds of typical IR small target deteection algorithms.Wherein NWIE algorithms are to be weighted based on local entropy
AGADMM algorithms.From fig. 6 it can be seen that method proposed by the present invention is obvious for Small object enhancing effect, simultaneously for
Larger target conditions present in image, as in the fifth line in Fig. 6 target for the submarine that just emerged bridge, using it
The target area is divided into several small point targets after the enhancing of its method, and this method can be intactly for submarine
Bridge part is strengthened, and this also illustrates this method is applicable not only to small target deteection, is also applied for including the figure of big target
As detection.
To verify the real-time of the present invention, using following PC machine hardware configuration：CPU is Intel (R) Core (TM) i5-
3230M 2.6GHz, inside save as 12GB, and video card is NVIDIA NVS5400M, 2G independence video memorys；Realized by Matlab and C++
Algorithm.Target detection experiment is carried out when out to out is 4, the time required to the infrared image detection for different resolution such as
Shown in table 1.Experiment proves that the present invention disclosure satisfy that requirement of real-time, and is very suitable for Implementation of Embedded System.
Under the different images resolution ratio of the present invention of table 1 the time required to detection (ms)
| Sequence number | Image type | Resolution ratio (pixel) | Invention algorithm | Multi-scale Fractal algorithm |
| 1 | IR | 200×150 | 14.42 | 20.70 |
| 2 | IR | 280×228 | 28.40 | 39.37 |
| 3 | IR | 250×200 | 21.71 | 31.93 |
| 4 | IR | 281×240 | 29.41 | 40.49 |
| 5 | IR | 220×140 | 14.60 | 18.55 |
| 6 | IR | 320×240 | 33.51 | 48.16 |
Claims (5)
1. based on the infrared small target detection method for improving Multi-scale Fractal enhancing, it is characterised in that the specific steps of this method
It is：
Step (1), the Multi-scale Fractal vector for obtaining every pixel in original infrared image I '；
Step (2), the conspicuousness based on Multi-scale Fractal vector every pixel of calculating, after obtaining enhancing according to formula (1)
Image E (x, y)；
E (x, y)=mean (t (x, y,:))2-std(t(x,y,:))2Formula (1)
Wherein mean () and std () is respectively the function averaged with standard deviation, and (x, y) is the pixel in former infrared image I '
Coordinate；t(x,y,:) represent that the Multi-scale Fractal of pixel coordinate (x, y) is vectorial；
Step (3), based on image after enhancing, Target Segmentation is carried out using auto-thresholding algorithm, according to formula (2)-(6)
Target after being detected；
μ=mean (E (2 × εmax+1:rows-2×εmax,2×εmax+1:cols-2×εmax)) formula (2)
δ=std (E (2 × εmax+1:rows-2×εmax,2×εmax+1:cols-2×εmax)) formula (3)
PSR=(255- μ)/δ formulas (4)
T=c × PSR × δ+μ formulas (5)
D (x, y)=E (x, y) >=T formulas (6)
Wherein, μ is that image removes the pixel average after fringe region after strengthening, and pixel is divided after δ removes border for image after enhancing
The standard deviation of cloth；PSR is peak sidelobe ratio；C represents Small object division coefficient；Rows and cols is respectively original infrared image I '
Height and width；T is the threshold value for target detection；D is to detect the pixel for belonging to target.
2. according to claim 1 existed based on the infrared small target detection method for improving Multi-scale Fractal enhancing, its feature
It is specifically in step (1)：
According to Fractal Geometry Theory, derive：When the scale of measurement is ε, its corresponding D dimension area K (x, y, ε) is represented by
Wherein A (x, y, ε) represents that the surface area on the gradation of image surface under scale ε is estimated；
When calculating A (x, y, ε) using covering blanket method, have
Wherein, V (x, y, ε) is the volume at coordinate (x, y) place under scale ε, shown in calculation formula such as formula (9), U (x, y, ε) and B
(x, y, ε) is respectively the maximum and minimum value of pixel in coordinate (x, y) place contiguous range under the scale；
Formula (6) is substituted into formula (5), is obtained
Formula (9) and formula (10) analysis understand that under different scale, the volume V of homogeneous every pixel in background area is small, close to 0；And
In nonhomogeneous background area, although V values are larger, due to the fluctuating of background, therefore under different scale, its change rate is also relatively
Greatly, i.e., Part II absolute value is big in formula (10), therefore K values reduce；In target area, either bright target or dark target, its
The maximum or minimum value of regional area are fixed, and V is maximum, and its change rate is small, therefore maximum in target area K values；
Analyzed based on more than, if ignoring rear two parts in formula (10), have K (x, y, ε)=V (x, y, ε -1), one can be used at this time
The difference of the maximum of pixel and minimum value adds up and as the notable of the region centered on coordinate points (x, y) in a region
Property measurement；
Analyzed based on more than, if scale is ε=2 ... εmax, it is original infrared image as ε=1, εmaxFor out to out；It is former red
The pixel coordinate of any is (x, y) in outer image I ', passes through pixel value in the coordinate points contiguous range under formula (11) calculating scale ε
The difference of maxima and minima, as contrast of the pixel under scale ε, obtains contrast image I；It is different to eliminate
The influence for the cumulative effect that scale is brought, calculates average contrast of the every pixel under scale ε by formula (12), obtains scale
Fractal characteristic vector under ε；And then using scale ε as variable, acquisition Multi-scale Fractal vector t (x, y,:)；
3. based on the infrared big object detection method for improving Multi-scale Fractal enhancing, using the method described in claim 1 or 2,
Division coefficient c in formula (5) is adjusted to big Target Segmentation coefficient.
4., will using the method described in claim 1 or 2 based on the SAR small target detecting methods for improving Multi-scale Fractal enhancing
Original image is substituted for SAR image by infrared image, and division coefficient c is adjusted to SAR image Small object segmentation system in formula (5)
Number.
5., will using the method described in claim 1 or 2 based on the big object detection methods of SAR for improving Multi-scale Fractal enhancing
Original image is substituted for SAR image by infrared image, and division coefficient c is adjusted to the big Target Segmentation system of SAR image in formula (5)
Number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711403183.1A CN108038856B (en) | 2017-12-22 | 2017-12-22 | Infrared small target detection method based on improved multi-scale fractal enhancement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711403183.1A CN108038856B (en) | 2017-12-22 | 2017-12-22 | Infrared small target detection method based on improved multi-scale fractal enhancement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108038856A true CN108038856A (en) | 2018-05-15 |
| CN108038856B CN108038856B (en) | 2020-08-04 |
Family
ID=62100448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711403183.1A Active CN108038856B (en) | 2017-12-22 | 2017-12-22 | Infrared small target detection method based on improved multi-scale fractal enhancement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108038856B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109461164A (en) * | 2018-09-21 | 2019-03-12 | 武汉大学 | A kind of infrared small target detection method based on direction nuclear reconstitution |
| CN110084778A (en) * | 2019-01-31 | 2019-08-02 | 电子科技大学 | It is a kind of based on the infrared imaging cirrus detection method for dividing shape dictionary learning |
| CN110310264A (en) * | 2019-06-25 | 2019-10-08 | 北京邮电大学 | A kind of large scale object detection method, device based on DCNN |
| CN113343758A (en) * | 2021-04-26 | 2021-09-03 | 西安卓越视讯科技有限公司 | Long-distance unmanned aerial vehicle small target detection method based on infrared image |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5671294A (en) * | 1994-09-15 | 1997-09-23 | The United States Of America As Represented By The Secretary Of The Navy | System and method for incorporating segmentation boundaries into the calculation of fractal dimension features for texture discrimination |
| EP1708140A2 (en) * | 2005-03-31 | 2006-10-04 | Lockheed Martin Corporation | Unresolved target detection improvement by use of multiple matched filters approach at different spatial phases |
| US20100104191A1 (en) * | 2007-03-26 | 2010-04-29 | Mcgwire Kenneth C | Data analysis process |
| CN102521831A (en) * | 2011-12-02 | 2012-06-27 | 南京信息工程大学 | Robot vision image segmentation method based on multi-scale fractal dimension and neural network |
| CN103077534A (en) * | 2012-12-31 | 2013-05-01 | 南京华图信息技术有限公司 | Space-time multi-scale moving target detection method |
| CN103218782A (en) * | 2013-04-11 | 2013-07-24 | 杭州电子科技大学 | Infrared image strengthening method based on multiscale fractal characteristics |
| CN105741253A (en) * | 2016-01-27 | 2016-07-06 | 北京理工大学 | Enhancement estimation method of image fractal feature on the basis of merge replication |
| US20160371851A1 (en) * | 2014-12-30 | 2016-12-22 | Huazhong University Of Science And Technology | Infrared image-spectrum associated intelligent detection method and apparatus |
-
2017
- 2017-12-22 CN CN201711403183.1A patent/CN108038856B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5671294A (en) * | 1994-09-15 | 1997-09-23 | The United States Of America As Represented By The Secretary Of The Navy | System and method for incorporating segmentation boundaries into the calculation of fractal dimension features for texture discrimination |
| EP1708140A2 (en) * | 2005-03-31 | 2006-10-04 | Lockheed Martin Corporation | Unresolved target detection improvement by use of multiple matched filters approach at different spatial phases |
| US20100104191A1 (en) * | 2007-03-26 | 2010-04-29 | Mcgwire Kenneth C | Data analysis process |
| CN102521831A (en) * | 2011-12-02 | 2012-06-27 | 南京信息工程大学 | Robot vision image segmentation method based on multi-scale fractal dimension and neural network |
| CN103077534A (en) * | 2012-12-31 | 2013-05-01 | 南京华图信息技术有限公司 | Space-time multi-scale moving target detection method |
| CN103218782A (en) * | 2013-04-11 | 2013-07-24 | 杭州电子科技大学 | Infrared image strengthening method based on multiscale fractal characteristics |
| US20160371851A1 (en) * | 2014-12-30 | 2016-12-22 | Huazhong University Of Science And Technology | Infrared image-spectrum associated intelligent detection method and apparatus |
| CN105741253A (en) * | 2016-01-27 | 2016-07-06 | 北京理工大学 | Enhancement estimation method of image fractal feature on the basis of merge replication |
Non-Patent Citations (1)
| Title |
|---|
| 刘俊: "基于红外图像的内河运动船舶目标检测和跟踪技术研究", 《中国博士学位论文全文数据库（信息科技辑）》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109461164A (en) * | 2018-09-21 | 2019-03-12 | 武汉大学 | A kind of infrared small target detection method based on direction nuclear reconstitution |
| CN110084778A (en) * | 2019-01-31 | 2019-08-02 | 电子科技大学 | It is a kind of based on the infrared imaging cirrus detection method for dividing shape dictionary learning |
| CN110084778B (en) * | 2019-01-31 | 2021-04-13 | 电子科技大学 | Infrared imaging cirrus cloud detection method based on fractal dictionary learning |
| CN110310264A (en) * | 2019-06-25 | 2019-10-08 | 北京邮电大学 | A kind of large scale object detection method, device based on DCNN |
| CN110310264B (en) * | 2019-06-25 | 2021-07-20 | 北京邮电大学 | DCNN-based large-scale target detection method and device |
| CN113343758A (en) * | 2021-04-26 | 2021-09-03 | 西安卓越视讯科技有限公司 | Long-distance unmanned aerial vehicle small target detection method based on infrared image |
| CN113343758B (en) * | 2021-04-26 | 2022-03-15 | 西安卓越视讯科技有限公司 | Long-distance unmanned aerial vehicle small target detection method based on infrared image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108038856B (en) | 2020-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hou et al. | Multilayer CFAR detection of ship targets in very high resolution SAR images | |
| CN108038856A (en) | Based on the infrared small target detection method for improving Multi-scale Fractal enhancing | |
| CN104361582B (en) | Method of detecting flood disaster changes through object-level high-resolution SAR (synthetic aperture radar) images | |
| CN110837768B (en) | Online detection and identification method for rare animal protection | |
| CN102494675B (en) | High-speed visual capturing method of moving target features | |
| CN109583293A (en) | Aircraft Targets detection and discrimination method in satellite-borne SAR image | |
| CN109712112B (en) | Aerial photography insulator image positioning method based on local features | |
| CN105427350B (en) | Color reproduction image altering detecting method based on local quaternary number Uniformly bounded | |
| CN104834915B (en) | A kind of small infrared target detection method under complicated skies background | |
| CN104680538A (en) | SAR image CFAR target detection method on basis of super pixels | |
| Hou et al. | SAR image ship detection based on visual attention model | |
| CN110110675A (en) | A kind of wavelet field of combination of edge information divides shape infrared cirrus detection method | |
| CN110889843A (en) | SAR image ship target detection method based on maximum stable extremal region | |
| CN105005983A (en) | SAR image background clutter modeling and target detection method | |
| Jin et al. | Improved direction estimation for Di Zenzo's multichannel image gradient operator | |
| CN101482969A (en) | SAR image speckle filtering method based on identical particle computation | |
| Fu et al. | Infrared sea-sky line detection utilizing self-adaptive Laplacian of Gaussian filter and visual-saliency-based probabilistic Hough transform | |
| CN103279954A (en) | Remote-sensing image change detecting method based on land utilization database | |
| CN111161308A (en) | Dual-band fusion target extraction method based on key point matching | |
| CN108647693A (en) | A kind of sea infrared target detection method based on binary significant characteristics | |
| CN107369163A (en) | A kind of quick SAR image object detection method based on best entropy Double Thresholding Segmentation | |
| CN107729903A (en) | SAR image object detection method based on area probability statistics and significance analysis | |
| Yuan et al. | New Watershed segmentation algorithm based on hybrid gradient and self-adaptive marker extraction | |
| ZHU et al. | Dim small targets detection based on horizontal-vertical multi-scale grayscale difference weighted bilateral filtering | |
| CN108573236A (en) | Method for detecting infrared puniness target under skies background based on discrete fraction Brownian random field |
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 |