CN103279758B - There is the remote sensing image Ship Detection of cloud noise - Google Patents
There is the remote sensing image Ship Detection of cloud noise Download PDFInfo
- Publication number
- CN103279758B CN103279758B CN201310256096.3A CN201310256096A CN103279758B CN 103279758 B CN103279758 B CN 103279758B CN 201310256096 A CN201310256096 A CN 201310256096A CN 103279758 B CN103279758 B CN 103279758B
- Authority
- CN
- China
- Prior art keywords
- target
- length
- feature
- remote sensing
- area
- 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
Abstract
The invention discloses a kind of Ship Target Detection method based on the judgement of connection in series-parallel mixing multiple features fusion, this technology belongs to field of remote sensing image processing.In existing Oceanic remote sensing image higher, to the Ship Target Detection poor performance under complicated cloud noise background to false alarm rate under cloud noise of the detection algorithm on naval vessel.Geometric shape feature and the edge energy feature of target are combined by the present invention, comprehensive utilization connection in series-parallel mixing multiple features fusion decision method can take into account geometric shape feature and the feature of edge energy feature of target, extracts clarification of objective and describes son formation detection statistic.This detection method has been combined with the advantage of feature fusion, and the Ship Target in complicated cloud noise is had good power of test, and the method is also applied for the Ship Target Detection in smooth remote sensing image simultaneously, has application value.
Description
One, technical field
The present invention relates to the geometric shape feature utilizing target in remote sensing image process field and edge energy Feature Fusion warship
Ship detection method, i.e. connection in series-parallel mixing multiple features fusion decision method, it is adaptable to containing in the case of cloud noise in remote sensing image
To the detection of naval vessel, sea.
Two, background technology
Marine vessel is monitored significant by satellite remote sensing technology, and in the world, there is important application in each coastal strip country.
Owing to SAR image has round-the-clock advantage, and not by sky cloud effect, it is used widely in naval vessel, ocean monitors.Therewith
Comparing, optical satellite remote sensing is easily affected by weather, but the resolution of optical satellite is higher and has predominance, at certain bar
Can be as the means of supplementing out economy of satellite-borne SAR under part.
In recent years, utilize remote sensing image to carry out the supervision of naval vessel, ocean and cause the concern of many research worker.At present, use
Method in the detection of Oceanic remote sensing image naval vessel can be largely classified into method based on gray-scale statistical characteristics, side based on marginal information
Method, method based on fractal model and fuzzy theory, the method for view-based access control model mechanism of perception.The background of marine remote sensing image
The most single, but still there is many difficulties, such as the separation of Hai Lu, the interference on island and the interference of cloud layer.Traditional naval vessel inspection
Survey method is difficult to the interference overcoming cloud layer to detect naval vessel, there is substantial amounts of false-alarm in testing result.Single utilization is a kind of or several
Plant target characteristic, it is difficult to differentiate the false-alarm that Ship Target produces with cloud layer.Traditional method does not the most design one the most quickly to be had
The method of effect detects the Ship Target in the remote sensing image containing cloud noise.
Three, summary of the invention
1. to solve the technical problem that
It is an object of the invention to provide and a kind of can detect the most special based on connection in series-parallel mixing of Ship Target under complicated cloud noise
Levy the detection method of amalgamation judging.Geometric shape and edge energy two category feature of target are combined by this detection method, first solve
Determine the rapid extraction problem of Feature Descriptor of target area, established the judgement of a kind of connection in series-parallel mixing the most on this basis
Structure.If individually using the judgement of parallel organization, get rid of the false-alarm produced by cloud layer etc. relatively difficult, to each decision rule
Require higher；Individually using the judgement of serial structure, the ability that detection is suppressed target is more weak, easily causes false dismissal.Use
Connection in series-parallel mixed structure as above, can effectively detect repressed target and repel the generation of false-alarm.Meanwhile, should
Area-of-interest is expressed by detection method, can take into account geometric shape feature and the edge energy feature of target, accomplish two
The integrated use of category feature.
2. technical scheme
The remote sensing image Ship Detection having cloud noise of the present invention, including techniques below measure: be primarily based on
Diversity between Ship Target and the geometric shape of cloud layer target and edge energy two category feature in remote sensing image, is extracted
Four kinds describe sub: drift rate, length-width ratio, area change ratio, compactedness, and set a kind of for distinguishing warship on this basis
Description of the edge geometric shape feature of ship and cloud layer, then, is comprehensively utilized these five kinds and describes son, mixed by connection in series-parallel
Target is detected by judgement structure, finally, differentiates target according to given testing conditions in judgement structure, determines warship
Ship target also gets rid of cloud layer false-alarm.
3. beneficial effect
The present invention compares background technology and has the advantage that
(1) this detection method reduces the demand during remote sensing image naval vessel detects to Sea background；
(2) this detection method sets a kind of description based on object edge geometrical property, under the conditions of improve cloud noise
The effect of naval vessel detection；
(3) this detection method devises the judgement structure of a kind of connection in series-parallel mixing, by geometric shape and edge energy two category feature
Connected applications, improves Detection results；
(4) this detection method has the ability detecting Ship Target in strong cloud noise background.
Four, accompanying drawing explanation
Figure of description is the enforcement principle flow chart of the present invention.
Five, detailed description of the invention
Below in conjunction with Figure of description, the present invention is described in further detail.With reference to Figure of description, the present invention is embodied as
Mode divides following step:
(1) the remote sensing image data input computing device 1 obtained in advance is carried out region segmentation, carry out target and carry on the back with sea
Scape separates, and the remote sensing image data in this step are to obtain from the SPOT4 satellite remote sensing images of 10m resolution.
The method used is gray level threshold segmentation, and wherein gray threshold is estimated by general image average and the variance of remote sensing images.
(2) calculate device 2 to accept to calculate the partition data that device 1 obtains, each cut zone is carried out morphological dilations and fills out
Fill, and then obtain optimal segmentation effect, make target information to retain as far as possible.
(3) judgment means 3 accept calculate device 2 obtain based on each district of the partition data after Morphological scale-space, first labelling
The position in territory and area, and calculate minimum enclosed rectangle (by area minimum of computation), then judge whether its area and length-width ratio accord with
Close Ship Target size, i.e. Preliminary screening and meet the target area of naval vessel feature, reduce amount of calculation below, improve and calculate speed；
Area span is [20,600], and length-width ratio threshold value is 1.4.
(4) if the judged result of judgment means 3 is " target has the feature similar to Ship Target ", then flow process to device 4,
Device 5, device 6 and device 7 perform, and calculate clarification of objective and describe son；Device 4 utilizes the average in target window region
And variance, use different weight coefficients to estimate two gray threshold (K1, K2, wherein K1＜ K2) window is carried out twice segmentation,
Obtaining two centres of form of target area, based on object edge energy feature, the distance calculated between two centres of form obtains off-centring degree
l；Device 5, based on object edge energy feature, according to the ratio of the target area area that twice segmentation obtains, obtains the face of target
S is compared in long-pending change；Device 6, based on target geometric shape feature, uses threshold k1To the segmentation result of target in window, calculate it
Minimum enclosed rectangle, obtains the area filling degree f of target compared with rectangular area with target area；Device 7 is based on target geometric form
State feature, uses threshold k1To the segmentation result of target in window, target area, length-width ratio and edge pixels length is utilized to calculate
Obtaining tortuosity c of target, this describes sub-principle and is defined as follows:
What generally definition one was oval is c like circularity1=(4 π SOval/L2), wherein L is oval girth；First calculate in section
The length-width ratio of target area, the length in pixels of edges of regions and the major and minor axis of target, then obtain one actual like circularity c2,
From morphological characteristic, target can be differentiated by comparing the value obtained twice；
Wherein, L1For target area length of side number of pixels.Ship Target profile is more smooth and similar to elliptical shape, cloud layer target wheel
Wide complex.Objective contour complexity causes contour pixel length to become big, and its tortuosity is the biggest.
(5) judgment means 8 accepts to calculate the result of device 4, it is judged that whether the off-centring degree of target meets Ship Target feature.
(6) if the judged result of judgment means 8 is " target's center's drift rate has the feature similar to Ship Target ", then flow
Journey is to judgment means 9；Ship Target is thin-and-long, and length-width ratio is relatively big, and sea cloud layer is almost without the longest and the narrowest shape,
The Ship Target that cloud layer is bigger with length-width ratio is easily distinguished, and therefore judgment means 9 describes son to the length-width ratio of target and judges,
Length-width ratio threshold value M=5 herein.
(7) if the judged result of judgment means 9 is " target length-width ratio is more than threshold value ", then judge that target is as Ship Target；
If the judged result of judgment means 9 is " target length-width ratio is less than threshold value ", then flow process performs to device 10, and device 5,
Result of calculation is delivered separately to device 10 and calculates by device 6 and device 7；Three description that transmission is come by device 10 enter
Row weighted calculation, obtains three-dimensional feature detection statistic, and weighted formula is:
Wherein k1, k2, k3For weight coefficient.
(8) comparator 11 accepts to calculate the result of device 10, compares with judgement thresholding T set in advance, carries out there is driftlessness
Judgement.
(9) device 12 accepts the output result of comparator, if (s, c, f) higher than detection door for detection statistic V in comparator 11
Limit T, then device 12 shows that court verdict is Ship Target；If (s, c, f) less than detection door for detection statistic V in comparator 11
Limit T, then device 12 court verdict is cloud layer target.
Claims (1)
1. there is the remote sensing image Ship Detection of cloud noise, it is characterised in that include techniques below measure:
(1) set the Feature Descriptor tortuosity for differentiating Ship Target and cloud noise, specifically comprise the following steps that generally definition one
Individual ellipse is c like circularity1=(4 π SOval/L2), wherein L is oval girth；First calculate section in target area length-width ratio,
The length in pixels of edges of regions and the major and minor axis of target, then obtain one actual like circularity c2=(4πSTarget/L1 2), pass through
Target can be differentiated from morphological characteristic by the value relatively obtained for twice, has the advantages that effectively differentiate naval vessel and cloud noise,
Tortuosity expression formula is as follows:
Wherein, L1For target area length of side number of pixels, r is target area length-width ratio, and it is external that a and b is respectively target area minimum
The length of rectangle and width；
(2) the judgement structure of connection in series-parallel mixing is set, and to drift rate, length-width ratio, area change ratio, compactedness, tortuosity
Five kinds of Feature Descriptors carry out merging formation detection statistic, compare with setting thresholding, if less than thresholding, adjudicating as naval vessel
Target, otherwise judgement is cloud noise, specifically comprises the following steps that what the size by screening target area and length-width ratio obtained
After area-of-interest, can judge that target, as naval vessel, first ensures that the off-centring degree of target meets bar by two kinds of parallel organizations
Part；This method is divided into support, indeterminacy section and degree of refuting aspect ratio features, and we define a length-width ratio support threshold
Value M, when target length-width ratio belongs to based on the evidence supported more than M i.e. aspect ratio features, it is believed that this target is Ship Target；
When target aspect ratio features belongs to indeterminacy section, the area change of statistics target than s, compactedness f and tortuosity c, by this three
Kind feature one objective characteristic vector V of composition (s, c, f), express area-of-interest:
V (s, c, f)=k1·s+k2·c+k3F, wherein k1,k2,k3For weight coefficient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310256096.3A CN103279758B (en) | 2013-06-26 | 2013-06-26 | There is the remote sensing image Ship Detection of cloud noise |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310256096.3A CN103279758B (en) | 2013-06-26 | 2013-06-26 | There is the remote sensing image Ship Detection of cloud noise |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103279758A CN103279758A (en) | 2013-09-04 |
| CN103279758B true CN103279758B (en) | 2016-09-28 |
Family
ID=49062273
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310256096.3A Active CN103279758B (en) | 2013-06-26 | 2013-06-26 | There is the remote sensing image Ship Detection of cloud noise |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103279758B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101520896A (en) * | 2009-03-30 | 2009-09-02 | 中国电子科技集团公司第十研究所 | Method for automatically detecting cloud interfering naval vessel target by optical remote sensing image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5503141B2 (en) * | 2008-12-22 | 2014-05-28 | 日本無線株式会社 | Target detection device |
-
2013
- 2013-06-26 CN CN201310256096.3A patent/CN103279758B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101520896A (en) * | 2009-03-30 | 2009-09-02 | 中国电子科技集团公司第十研究所 | Method for automatically detecting cloud interfering naval vessel target by optical remote sensing image |
Non-Patent Citations (5)
| Title |
|---|
| A novel method for speckle noise reduction and ship target detection in SAR images;Shi-qi Huang等;《Pattern Recognition》;20090731;第42卷(第7期);第1533-1542页 * |
| 一种光学遥感图像海面舰船检测算法;高立宁等;《清华大学学报（自然科学版）》;20110131;第51卷(第1期);第105-110页 * |
| 可见光遥感图像中舰船目标检测的多阶阈值分割方法;王保云等;《中国科学技术大学学报》;20110430;第41卷(第4期);第293-298页 * |
| 基于海陆分割的舰船目标变化检测研究;蔡姝等;《电视技术》;20100531;第34卷(第5期);第109-112页 * |
| 基于视觉显著性的雷达视频舰船检测;周伟等;《雷达科学与技术》;20120229;第10卷(第1期);第54-58页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103279758A (en) | 2013-09-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104036239B (en) | Fast high-resolution SAR (synthetic aperture radar) image ship detection method based on feature fusion and clustering | |
| CN101975940B (en) | Segmentation combination-based adaptive constant false alarm rate target detection method for SAR image | |
| CN103198332B (en) | A kind of far infrared vehicle-mounted pedestrian detection method of real-time robust | |
| CN102855622B (en) | A kind of infrared remote sensing image sea ship detection method based on significance analysis | |
| CN106022288A (en) | Marine oil spill information identification and extraction method based on SAR image | |
| CN107818326A (en) | A kind of ship detection method and system based on scene multidimensional characteristic | |
| CN102542277B (en) | Method for detecting ship trail of ocean synthetic aperture radar image | |
| CN107527023B (en) | Polarized SAR image classification method based on superpixels and topic models | |
| CN104732215A (en) | Remote-sensing image coastline extracting method based on information vector machine | |
| CN106682586A (en) | Method for real-time lane line detection based on vision under complex lighting conditions | |
| CN109359661B (en) | Sentinel-1 radar image classification method based on convolutional neural network | |
| CN107992818B (en) | Method for detecting sea surface ship target by optical remote sensing image | |
| CN103020975A (en) | Wharf and ship segmentation method combining multi-source remote sensing image characteristics | |
| CN105138982A (en) | Crowd abnormity detection and evaluation method based on multi-characteristic cluster and classification | |
| CN103400156A (en) | CFAR (Constant False Alarm Rate) and sparse representation-based high-resolution SAR (Synthetic Aperture Radar) image ship detection method | |
| CN104331886A (en) | Port region ship and warship detection method based on high resolution SAR image | |
| CN102830404B (en) | Method for identifying laser imaging radar ground target based on range profile | |
| CN105427301B (en) | Based on DC component than the extra large land clutter Scene Segmentation estimated | |
| CN102831600B (en) | A kind of image level dividing method merged based on Weighted Cut | |
| CN103996017A (en) | Ship detection method based on Hu invariant moment and support vector machine | |
| CN104134004A (en) | Marine environment safety assessment method based on D-S evidence theory | |
| CN104680183A (en) | SAR target identification method based on scattering point and K-center one-class classifier | |
| CN106803078A (en) | A kind of SAR image Ship Target dividing method | |
| CN112068085A (en) | Radar sea clutter original data rapid preprocessing method based on deep learning | |
| CN105139034A (en) | Spectrum filtering based ship detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Zhou Wei Inventor after: Xu Chengbin Inventor after: Jiang Baichen Inventor after: Hu Wenchao Inventor after: Sun Lu Inventor after: He Dongliang Inventor after: Cong Yu Inventor after: Guan Jian Inventor before: Zhou Wei Inventor before: Xu Chengbin Inventor before: Yao Libo Inventor before: Cong Yu Inventor before: He Dongliang Inventor before: Guan Jian |
|
| COR | Change of bibliographic data | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200724 Address after: 264001 Research and Academic Department, 188 Erma Road, Zhifu District, Yantai City, Shandong Province Patentee after: NAVAL AERONAUTICAL University Address before: 264001 No. two, 188 Road, Zhifu District, Shandong, Yantai Patentee before: NAVAL AERONAUTICAL AND ASTRONAUTICAL University PLA |