CN104966295B - A kind of naval vessel extracting method based on wire-frame model - Google Patents
A kind of naval vessel extracting method based on wire-frame model Download PDFInfo
- Publication number
- CN104966295B CN104966295B CN201510332404.5A CN201510332404A CN104966295B CN 104966295 B CN104966295 B CN 104966295B CN 201510332404 A CN201510332404 A CN 201510332404A CN 104966295 B CN104966295 B CN 104966295B
- Authority
- CN
- China
- Prior art keywords
- ship
- candidate
- line
- fore
- feature
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
Abstract
The invention provides a kind of naval vessel extracting method based on wire-frame model, including step：S1, the pretreatment of remote sensing image, Threshold segmentation and line drawing；S2, utilize model ship detection candidate's ship；S3, using detect mask verify candidate's ship；S4, using SVM non-candidate ship region is classified again, then repeatedly S2 ~ S3.The present invention can reduce loss, improve the recall ratio and accuracy rate of ship detection；In addition, the present invention is not only restricted to ship size and texture, polytype ship can be detected, it is applied widely.
Description
Technical field
It is more particularly to a kind of to be based on wire-frame model the present invention relates to high-resolution remote sensing image Objective extraction technical field
Naval vessel extracting method.
Background technology
In order to preferably monitor and manage the ship of offshore and offshore, information more effectively is provided for military combat, is hit
Illegal fishing, the relevant issues that ship is extracted are current urgent problems to be solved.High-resolution remote sensing image is more and more to obtain quilt
For doing ship extraction, but ship extraction method is primarily present problem at present：It can only can typically be extracted for volume is larger
The ship of texture information is effectively extracted, and for offshore, ship in complex background, then Detection results are not good.Pin
To these problems, the method based on line feature can be efficiently solved.
Current high-resolution remote sensing image ship extraction method mainly has based on texture, based on conspicuousness and based on line feature
Three class methods.
(1) method based on texture
Because ship and its background there are different textures, by analyzing texture information, ship can be extracted.
But the precondition of this method is that ship is sufficiently large, its texture information can be extracted.For small volume, can not be very
The ship of texture is extracted well, and this method can not obtain good effect.
(2) method based on conspicuousness
Because ship and its background pixel value difference are big, it is possible to use this characteristic carries out the extraction of ship.This method
Effective for the ship of the distribution of being scattered of offshore, for the ship of the dense distribution of offshore, this method can not obtain making us full
The result of meaning.
(3) method based on line feature
Propose on the basis of " V " fore model, On-line testing, find the line segment pair of satisfaction " V " fore model, then pass through ship
Shape facility only finally determines whether candidate's ship is that ship is true.But line drawing sometimes can not will constitute the two of fore
Bar line segments extraction comes out, and causes missing inspection.
The content of the invention
In view of the deficienciess of the prior art, ship detection accuracy rate, the scope of application can be improved the invention provides one kind
A kind of wide naval vessel extracting method based on wire-frame model.
In order to solve the above technical problems, the present invention is adopted the following technical scheme that：
A kind of naval vessel extracting method based on wire-frame model, including step：
S1, remote sensing image pretreatment, i.e., carry out denoising, enhancing successively to remote sensing image；
S2, row threshold division is entered to remote sensing image after pretreatment obtain connected region, the line for extracting original remote sensing image is special
Levy；
S3, based on connected region and line feature, using the model ship based on fore part structure in the line feature that S2 is extracted
Ship is extracted, the connected region where ship is obtained, the connected region that camber is more than threshold value s is first candidate's ship, otherwise
For the first non-candidate ship, s is empirical value；
S3 also includes the line feature to first candidate's ship fore line segment to beyond, using the ship mould based on hull configuration
Type extracts ship, obtains the connected region where ship, the connected region of size within a preset range is second candidate's ship, no
It is then the second non-candidate ship, preset range is rule of thumb set；Wherein, the model ship based on hull configuration is by (1) distance
Less than d1And angle is less than 10 degree of line feature pair and (2) are less than d with line feature centering any line characteristic distance3And angle is
[θ1,θ2] line feature constitute, d1And d3For empirical value, [θ1,θ2] represent the angular range of fore and hull；
S4, first candidate's ship and the second candidate are really detected using the detection mask being made up of fan-shaped mask and semicircle mask
Ship, be specially：
S4-1 covers fan-shaped mask and semicircle mask in the bianry image of first candidate's ship or second candidate's ship,
The summit of fan-shaped mask is located at the nearly fore end points line of the line feature of fore two of first candidate's ship or second candidate's ship
Midpoint, its radius for the line feature of fore two nearly hull end points to the distance on summit higher value, its both sides respectively with fore two
Line feature is parallel；Semicircle mask circle is the semicircle for eliminating fan-shaped mask, and its center of circle is located at apex, and its radius is covered with sector
Film radius, its symmetry axis is overlapped with the angular bisector of fan-shaped mask central angle；
The overlapping region area of the fan-shaped mask of S4-2 orders and first candidate's ship or second candidate's ship and fan-shaped mask face
The ratio between product is ratio1, make the overlapping region area and semicircle mask of semicircle mask and first candidate's ship or second candidate's ship
Area ratio is ratio2, ratio1More than η1And ratio2Less than η2Candidate's ship be true ship, η1And η2Respectively
The empirical value of value in 50%~90% and 20%~50%；
S5 sets up the minimum area-encasing rectangle of true ship, with true ship and true ship in minimum area-encasing rectangle with outskirt
The pixel value training grader in domain, is classified using grader to the first non-candidate ship or second candidate's ship, acquisition the
One ship connected region or the second ship connected region；
S6 judges whether the camber of the first ship connected region is more than threshold value s, if being more than, the first ship connected region
That is first candidate's ship, is otherwise the first non-candidate ship；S4 is performed to first candidate's ship that this step is obtained, it is non-to first
Candidate's ship performs S5, until can't detect the first non-candidate ship；
Judge the second ship connected region size whether in preset range [c1,c2], if in the second ship connected region
That is second candidate's ship, is otherwise the second non-candidate ship；S4 is performed to second candidate's ship of this acquisition, it is non-to second to wait
Ship is selected to perform S5, until can't detect the second non-candidate ship.
In S3, ship is extracted in the line feature that S2 is extracted using the model ship based on fore part structure, is specially：
The line feature that S2 is extracted is traveled through, the line feature to the connected region where the feature of front is operated：Find with working as
The distance of front feature is less than d and angle is [α1,α2] line feature, the line feature with when front feature constitute fore line segment pair；
If being less than d and angle is [θ in the presence of with the distance of fore line segment centering any line feature1,θ2] line feature, the line feature and ship
Head line segment is to constituting ship；Wherein, [α1,α2] represent fore angular range, [θ1,θ2] angular range of fore and hull is represented,
D is empirical value, and value is carried out based on experimental verification.
In S3, the connected region where ship is obtained, is specially：
By the fore line segment of ship to expanding, folding the maximum connected region of area with fore line segment counterweight after expansion is
Connected region where ship.
Above-mentioned used grader is SVM classifier.
Compared to the prior art, the invention has the advantages that and beneficial effect：
1st, loss can be reduced, the recall ratio and accuracy rate of ship detection is improved.
2nd, ship size and texture are not only restricted to, polytype ship can be detected, it is applied widely.
Brief description of the drawings
Fig. 1 is the flow chart of step 1~3 of the embodiment of the present invention；
Fig. 2 is the model ship schematic diagram that the embodiment of the present invention is used, wherein, figure (a) is the ship based on fore part structure
Model, figure (b) is the model ship based on hull configuration；
Fig. 3 is detection mask schematic diagram, wherein, the candidate that figure (a)~(b) obtains for the model ship based on fore part structure
The detection mask of ship, the detection mask for candidate's ship that figure (c)~(d) obtains for the model ship based on hull configuration；
Fig. 4 is the flow chart of step 4 of the embodiment of the present invention；
The ship that Fig. 5 is the present invention extracts result figure, wherein, figure (a)~(c) is that three kinds of specific ships extract result
Figure.
Embodiment
In order that the purpose of the present invention, technical scheme and beneficial effect are more clearly understood, below with reference to accompanying drawing and reality
Example is applied, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only used for explaining this hair
It is bright, it is not intended to limit the present invention.
Wen Zhong, " high-resolution remote sensing image " is referred to as " remote sensing image ".
The flow chart of step 1~3 of the present invention is shown in Fig. 1, and the flow chart of step 4 is shown in Fig. 4, and this is illustrated below in conjunction with the accompanying drawings
The implementation steps of invention.
Step 1, the pretreatment of remote sensing image, Threshold segmentation and line drawing.
The pretreatment of step 1.1 remote sensing image.
The pretreatment of remote sensing image is routine techniques, and its purpose is to carry out denoising and enhancing to remote sensing image.
In specific implementation, denoising and increasing are carried out using median filtering method and histogram equalization method respectively to remote sensing image
By force.In median filtering method, remote sensing image is handled for 5 × 5 template using size, to remove noise in remote sensing image.Then it is sharp
Enhancing processing is carried out to the remote sensing image after medium filtering with histogram equalization method, formula (1) is seen.
In formula (1), L represents number of greyscale levels in remote sensing image, njRepresent in number of pixels of the gray level as j, n representative images
Pixel summation, max and min are respectively original image maximum gradation value and minimum value, and s is the remote sensing figure after histogram equalization
Picture.
The Threshold segmentation and line drawing of step 1.2 remote sensing image.
Enter row threshold division to pretreated remote sensing image using classical maximum variance between clusters, obtain connected region
Domain.Line drawing is carried out to original remote sensing image using EDlines or LSD line drawings method, line feature is obtained, described line feature is
Line segment.It is UNUSED by the wired signature of institute of extraction.
Step 2, candidate's ship detection.
Step 2.1 is using model ship detection candidate's ship based on fore part structure.
This sub-step is prior art, and for ease of understanding, this sub-step will be described in detail below.
Model ship based on fore part structure is shown in Fig. 2 (a).The line feature that traversal step 1 is extracted, where the feature of front
In the line feature of connected region, find and when the distance of front feature is less than d and angle is in [α1,α2] in line feature, the line is special
Levy and constitute fore line segment pair with working as front feature.If any with fore line segment centering when existing in connected region where the feature of front
The distance of line feature is less than d and angle is [θ1,θ2] line feature, then the line feature and fore line segment extract ship to constituting ship
Connected region where only.If the camber of the connected region is more than threshold value s, the connected region is candidate's ship, is labeled as
CANDIDATE, the fore line segment of candidate's ship is to mark；, will if the camber of the connected region is not more than threshold value s
The connected component labeling is REMAIN1.
There is distance between four distances, this specific implementation center line feature to use Euclidean distance between line segment, Euclidean distance is this
Minimum range in four distances.D is empirical value, carries out value based on experimental verification, d preferred scope is 15~30.[α1,
α2] fore angular range is represented, because fore angle is general between 15~100 degree, so, in this specific implementation, by α1And α2Point
15 degree and 100 degree are not set to.Certain α1、α2Value can also rule of thumb be adjusted.[θ1,θ2] represent fore and ship
The angular range of body, because the angle of fore and hull is general between 135~170 degree, so, in this specific implementation, θ1And θ2Point
135 degree and 170 degree are not set to.S is empirical value, typically can be in [0.6,1] interior value based on experimental verification value, and this is specific
In implementation, s=0.3 is made.
The definition of connected region is：Assuming that pixel A and pixel B are marked as 1 and the Euclidean distance of pixel A and pixel B existsWithin, then pixel A is connected with pixel B.The specific method of connected region is where obtaining ship：By fore line segment to entering
Row expansion, the fore line segment pair and which connected region overlapping area after expansion is maximum, and it to be considered as by ship the connected region
Connected region.
Camber is calculated as prior art, and circular is：Connected region area divided by its minimum surround convex polygon
Shape area is the camber of connected region.
Step 2.2 is using model ship detection candidate's ship based on hull configuration.
The present invention proposes a kind of model ship based on hull configuration, sees Fig. 2 (b), special for unmarked USED line
Extraction ship is levied, unmarked USED line feature is line feature of candidate's ship fore line segment to beyond.Travel through unmarked USED
Line feature, in the line feature as the unmarked USED of connected region where the feature of front, find with when front feature away from
From less than d1And angle is less than 10 degree of line feature, the line feature constitutes hull line segment pair, hull line segment pair with working as front feature
Angle is less than 10 degree, that is, shows hull line segment to for approximately parallel line segment pair.If being deposited when in connected region where the feature of front
It is being less than d with any line characteristic distance in hull line segment pair3And angle is [θ1,θ2] unmarked USED line feature, the line is special
Levy with hull line segment to constituting ship, extract the connected region where the ship.If the connected region size is in [c1,c2] model
Enclose, the connected region is candidate's ship, labeled as CANDIDATE, be to mark by the hull line segment of candidate's ship；Size
Not in [c1,c2] connected region of scope is the second non-candidate ship, labeled as REMAIN2.
Distance also uses Euclidean distance between this sub-step center line feature.d1And d3For empirical value, obtained according to verification experimental verification；
[c1,c2] rule of thumb set with verification experimental verification.
Step 3, detection mask is set up, for detecting candidate's ship, so as to confirm true ship.
Fig. 3 is detection mask, for confirming whether candidate's ship is true ship.
The detection mask for candidate's ship that step 3.1 establishment step 2.1 is obtained.
Detect that mask includes fan-shaped mask and semicircle mask.Fig. 3 (a) is that the sector for candidate's ship that step 2.1 is obtained is covered
Film builds schematic diagram, and fan-shaped mask summit is located at the midpoint of the nearly fore end points line of the line feature (1) of fore two, and its radius is ship
The nearly hull end points of first two lines feature (1) is to the higher value of the distance on fan-shaped mask summit, fan-shaped mask both sides and the line of fore two
Feature (1) is parallel.Fan-shaped summit is the center of circle of circle where sector, and the described line feature of fore two is fore line segment pair.
Fig. 3 (b) is that the semicircle mask for candidate's ship that step 2.1 is obtained builds schematic diagram, and semicircle mask is to remove sector
The semicircle mask of mask, the semicircle mask center of circle is located at fan-shaped mask apex, and its radius is with fan-shaped mask radius, its symmetry axis
The angular bisector of fan-shaped mask central angle.
Step 3.2, the detection mask for candidate's ship that establishment step 2.2 is obtained.
Fig. 3 (c) is that the fan-shaped mask for candidate's ship that step 2.2 is obtained builds schematic diagram, and fan-shaped mask summit is located at
The nearly fore end points of heading line feature (1), radius is heading line feature (1) length, fan-shaped mask both sides and (1) heading line feature
And (2) fan-shaped mask summit and the line coincident of the nearly fore end points of heading line feature offside hull line feature (2) (1).
Fig. 3 (d) is that the semicircle mask for candidate's ship that step 2.2 is obtained builds schematic diagram, and semicircle mask is to remove sector
The semicircle mask of mask, the semicircle mask center of circle is located at fan-shaped mask apex, and its radius is with fan-shaped mask radius, its symmetry axis
The angular bisector of fan-shaped mask central angle.
Step 3.3 is confirmed using fan-shaped mask and semicircle mask to candidate's ship, obtains true ship.
In the bianry image that fan-shaped mask is covered to candidate's ship, the overlapping region face of the fan-shaped mask of order and candidate's ship
Product is ratio with fan-shaped mask area ratio1.Semicircle mask is covered in the bianry image where candidate's ship, make semicircle
The overlapping region area and semicircle mask area ratio of mask and candidate's ship are ratio2If, ratio1More than η1And ratio2
Less than η2, it is determined that candidate's ship is true ship, otherwise it is assumed that candidate's ship is false retrieval.
η1And η2For empirical value, η1The value in the range of 50%~90%, η2Value, this tool in the range of 20%~50%
By η in body implementation1And η2It is set to 60% and 30%.
Step 4, the classification of non-candidate ship.
In the minimum area-encasing rectangle for the true ship that extraction step 3 is determined, minimum area-encasing rectangle, outside true ship region
Region is non-ship region.True ship region randomly selects m segment, regard segment pixel value as positive sample.Segment size
Do not require, rule of thumb can be chosen with demand with quantity, for example, can randomly select the segment that 50 sizes are 5 × 5, or
Randomly select the segment that 60 sizes are 3 × 3.Non- ship region randomly selects m segment, regard its pixel value as negative sample.
It will be trained in positive sample and negative sample input SVM (SVMs).
In segmentation figure, ship class pixel value is 1, and non-ship class pixel value is 0, using the characteristic, utilizes the SVM of training
Grader is classified to the first non-candidate ship or the second non-candidate ship, you can ship connected region.To the first non-candidate
The ship connected region that ship classification is obtained, judges whether its camber is more than threshold value s, if being more than, the ship connected region is
Candidate's ship, is otherwise the first non-candidate ship.The ship connected region obtained to the classification of the second non-candidate ship, judges that its is big
It is small whether in preset range [c1,c2] in, if the ship connected region is candidate's ship, is otherwise the second non-candidate ship.
Step 3 is performed to candidate's ship, this step is continued executing with to the first non-candidate ship or the second non-candidate ship, until being not present
Non-candidate ship.
The present invention can overcome the disadvantages that by the missing inspection of inaccurate the brought ship of line drawing；Meanwhile, the present invention utilizes mask
Confirm candidate's ship, reduce fallout ratio.
It should be appreciated that for those of ordinary skills, can according to the above description be improved or converted,
And all these modifications and variations should all belong to the protection domain of appended claims of the present invention.
Claims (5)
1. a kind of naval vessel extracting method based on wire-frame model, it is characterised in that including step：
S1, remote sensing image pretreatment, i.e., carry out denoising, enhancing successively to remote sensing image；
S2, remote sensing image after pretreatment is entered row threshold division obtain connected region, extract original remote sensing image line feature；
S3, based on connected region and line feature, extracted using the model ship based on fore part structure in the line feature that S2 is extracted
Ship, obtains the connected region where ship, the connected region that camber is more than threshold value s is first candidate's ship, is otherwise the
One non-candidate ship, s is empirical value；
S3 also includes the line feature to first candidate's ship fore line segment to beyond, is carried using the model ship based on hull configuration
Take ship, obtain the connected region where ship, the connected region of size within a preset range is second candidate's ship, otherwise for
Second non-candidate ship, preset range is rule of thumb set；Wherein, the model ship based on hull configuration is less than by (1) distance
d1And angle is less than 10 degree of line feature pair and (2) are less than d with line feature centering any line characteristic distance3And angle is [θ1,
θ2] line feature constitute, d1And d3For empirical value, [θ1,θ2] represent the angular range of fore and hull；
S4, first candidate's ship and second candidate's ship are really detected using the detection mask being made up of fan-shaped mask and semicircle mask
Only, it is specially：
S4-1 covers fan-shaped mask and semicircle mask in the bianry image of first candidate's ship or second candidate's ship, fan-shaped
The summit of mask is located at the midpoint of the nearly fore end points line of the line feature of fore two of first candidate's ship or second candidate's ship,
Its radius for the line feature of fore two nearly hull end points to the distance on summit higher value, its both sides respectively with the line feature of fore two
It is parallel；Semicircle mask circle is the semicircle for eliminating fan-shaped mask, and its center of circle is located at apex, and its radius is with fan-shaped mask half
Footpath, its symmetry axis is overlapped with the angular bisector of fan-shaped mask central angle；
The overlapping region area and sector mask area of the fan-shaped mask of S4-2 orders and first candidate's ship or second candidate's ship it
Than for ratio1, make the overlapping region area and semicircle mask area of semicircle mask and first candidate's ship or second candidate's ship
The ratio between be ratio2, ratio1More than η1And ratio2Less than η2Candidate's ship be true ship, η1And η2Respectively 50%~
The empirical value of value in 90% and 20%~50%；
S5 sets up the minimum area-encasing rectangle of true ship, with true ship and true ship in minimum area-encasing rectangle with exterior domain
Pixel value trains grader, and the first non-candidate ship or the second non-candidate ship are classified using grader, obtains first
Ship connected region or the second ship connected region；
S6 judges whether the camber of the first ship connected region is more than threshold value s, if being more than, the first ship connected region is the
One candidate's ship, is otherwise the first non-candidate ship；S4 is performed to first candidate's ship that this step is obtained, to the first non-candidate
Ship performs S5, until can't detect the first non-candidate ship；
Judge the second ship connected region size whether in preset range [c1,c2], if the second ship connected region is
Two candidate's ships, are otherwise the second non-candidate ship；S4 is performed to second candidate's ship of this acquisition, to the second non-candidate ship
S5 is only carried out, until can't detect the second non-candidate ship.
2. the naval vessel extracting method as claimed in claim 1 based on wire-frame model, it is characterised in that：
In S3, ship is extracted in the line feature that S2 is extracted using the model ship based on fore part structure, is specially：
The line feature that S2 is extracted is traveled through, the line feature to the connected region where the feature of front is operated：Find and work as front
The distance of feature is less than d and angle is [α1,α2] line feature, the line feature with when front feature constitute fore line segment pair；If depositing
It is being less than d and angle is [θ with the distance of any line feature in fore line segment pair1,θ2] line feature, the line feature and heading line
Section is to constituting ship；Wherein, [α1,α2] represent fore angular range, [θ1,θ2] angular range of fore and hull is represented, d is
Empirical value, value is carried out based on experimental verification.
3. the naval vessel extracting method as claimed in claim 1 based on wire-frame model, it is characterised in that：
Model ship of the described use based on hull configuration extracts ship：
Carried out for line feature of the first candidate's ship fore line segment to beyond, travel through line feature, to connecting where the feature of front
The line feature in logical region is operated：Find and be less than d with the distance when front feature1, should and angle is less than 10 degree of line feature
Line feature constitutes hull line segment pair with working as front feature；If being less than d in the presence of with hull line segment centering any line characteristic distance3And folder
Angle is [θ1,θ2] line feature, the line feature and hull line segment are to constituting ship；Wherein, d1And d3For empirical value, [θ1,θ2] table
Show the angular range of fore and hull.
4. the naval vessel extracting method as claimed in claim 1 based on wire-frame model, it is characterised in that：
In S3, the connected region where ship is obtained, is specially：
By the fore line segment of ship to expanding, it is ship to fold the maximum connected region of area with fore line segment counterweight after expansion
The connected region at place.
5. the naval vessel extracting method as claimed in claim 1 based on wire-frame model, it is characterised in that：
Described grader is SVM classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510332404.5A CN104966295B (en) | 2015-06-16 | 2015-06-16 | A kind of naval vessel extracting method based on wire-frame model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510332404.5A CN104966295B (en) | 2015-06-16 | 2015-06-16 | A kind of naval vessel extracting method based on wire-frame model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104966295A CN104966295A (en) | 2015-10-07 |
| CN104966295B true CN104966295B (en) | 2017-10-24 |
Family
ID=54220326
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510332404.5A Active CN104966295B (en) | 2015-06-16 | 2015-06-16 | A kind of naval vessel extracting method based on wire-frame model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104966295B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295503A (en) * | 2016-07-25 | 2017-01-04 | 武汉大学 | The high-resolution remote sensing image Ship Target extracting method of region convolutional neural networks |
| CN107610114B (en) * | 2017-09-15 | 2019-12-10 | 武汉大学 | optical satellite remote sensing image cloud and snow fog detection method based on support vector machine |
| CN109447082B (en) * | 2018-08-31 | 2020-09-15 | 武汉尺子科技有限公司 | Scene moving object segmentation method, system, storage medium and equipment |
| CN111784606B (en) * | 2020-06-30 | 2023-11-24 | 长江大学 | Remote sensing image classification post-processing method, storage medium and system |
| CN114612799B (en) * | 2022-03-11 | 2022-09-16 | 应急管理部国家自然灾害防治研究院 | Space self-adaptive positive and negative sample generation method and system based on landslide/non-landslide area ratio |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5305985B2 (en) * | 2009-02-27 | 2013-10-02 | 三菱スペース・ソフトウエア株式会社 | Artifact detection device, artifact detection method, and artifact detection program |
| CN103544502A (en) * | 2013-10-29 | 2014-01-29 | 上海市城市建设设计研究总院 | High-resolution remote-sensing image ship extraction method based on SVM |
| CN104077777A (en) * | 2014-07-04 | 2014-10-01 | 中国科学院大学 | Sea surface vessel target detection method |
| CN104166853A (en) * | 2014-04-24 | 2014-11-26 | 中国人民解放军海军航空工程学院 | Method for quickly extracting regularized ship section from high resolution remote sensing image |
-
2015
- 2015-06-16 CN CN201510332404.5A patent/CN104966295B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5305985B2 (en) * | 2009-02-27 | 2013-10-02 | 三菱スペース・ソフトウエア株式会社 | Artifact detection device, artifact detection method, and artifact detection program |
| CN103544502A (en) * | 2013-10-29 | 2014-01-29 | 上海市城市建设设计研究总院 | High-resolution remote-sensing image ship extraction method based on SVM |
| CN104166853A (en) * | 2014-04-24 | 2014-11-26 | 中国人民解放军海军航空工程学院 | Method for quickly extracting regularized ship section from high resolution remote sensing image |
| CN104077777A (en) * | 2014-07-04 | 2014-10-01 | 中国科学院大学 | Sea surface vessel target detection method |
Non-Patent Citations (2)
| Title |
|---|
| Ship Detection and Classification in High-Resolution Remote Sensing Imagery Using Shape-Driven Segmentation Method;Chao Tao et al.;《Proceedings of SPIE》;20091231;第7495卷(第4期);全文 * |
| 高分辨率遥感影像上基于形状特征的船舶提取;汪闽 等;《武汉大学学报.信息科学版》;20050831;第30卷(第8期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104966295A (en) | 2015-10-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104966295B (en) | A kind of naval vessel extracting method based on wire-frame model | |
| CN103971102B (en) | Static gesture recognition method based on finger contour and decision-making trees | |
| CN105701483B (en) | A kind of urban border extracting method of fusion Multi-spectral Remote Sensing Data and night lights remotely-sensed data | |
| CN108009472B (en) | Finger back joint print recognition method based on convolutional neural network and Bayes classifier | |
| CN101551863B (en) | Method for extracting roads from remote sensing image based on non-sub-sampled contourlet transform | |
| CN105654097B (en) | The detection method of quadrangle marker in image | |
| CN107203990A (en) | A kind of labeling damage testing method based on template matches and image quality measure | |
| CN107301638B (en) | A kind of ellipse detection method for supporting line segment based on arc | |
| CN106446952A (en) | Method and apparatus for recognizing score image | |
| CN103020975A (en) | Wharf and ship segmentation method combining multi-source remote sensing image characteristics | |
| CN104077577A (en) | Trademark detection method based on convolutional neural network | |
| CN109993099A (en) | A kind of lane line drawing recognition methods based on machine vision | |
| CN103413119A (en) | Single sample face recognition method based on face sparse descriptors | |
| CN104680130A (en) | Chinese character recognition method for identification cards | |
| CN103020965A (en) | Foreground segmentation method based on significance detection | |
| CN105160686B (en) | A kind of low latitude various visual angles Remote Sensing Images Matching Method based on improvement SIFT operators | |
| CN104766344B (en) | Vehicle checking method based on movement edge extractor | |
| CN110008900B (en) | Method for extracting candidate target from visible light remote sensing image from region to target | |
| CN105718552A (en) | Clothing freehand sketch based clothing image retrieval method | |
| CN104408711A (en) | Multi-scale region fusion-based salient region detection method | |
| CN106709500A (en) | Image feature matching method | |
| CN107704865A (en) | Fleet Targets Detection based on the extraction of structure forest edge candidate region | |
| CN109409356B (en) | Multi-direction Chinese print font character detection method based on SWT | |
| CN108509950A (en) | Railway contact line pillar number plate based on probability characteristics Weighted Fusion detects method of identification | |
| Zhao et al. | The edge detection of river model based on self-adaptive Canny Algorithm and connected domain segmentation |
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 |