CN108376247A - Strategic coarse-fine combined sea-land separation method applied to optical remote sensing ship detection - Google Patents
Strategic coarse-fine combined sea-land separation method applied to optical remote sensing ship detection Download PDFInfo
- Publication number
- CN108376247A CN108376247A CN201810113117.9A CN201810113117A CN108376247A CN 108376247 A CN108376247 A CN 108376247A CN 201810113117 A CN201810113117 A CN 201810113117A CN 108376247 A CN108376247 A CN 108376247A
- Authority
- CN
- China
- Prior art keywords
- land
- sea
- extra large
- image
- segmentation
- 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.)
- Pending
Links
- 238000000926 separation method Methods 0.000 title claims abstract description 47
- 238000001514 detection method Methods 0.000 title claims abstract description 24
- 230000003287 optical effect Effects 0.000 title claims abstract description 12
- 230000011218 segmentation Effects 0.000 claims abstract description 57
- 238000000605 extraction Methods 0.000 claims description 16
- 230000000007 visual effect Effects 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 12
- 230000002902 bimodal effect Effects 0.000 claims description 4
- 238000000151 deposition Methods 0.000 claims description 4
- 230000000877 morphologic effect Effects 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 239000000686 essence Substances 0.000 description 18
- 238000000034 method Methods 0.000 description 16
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 5
- 238000000638 solvent extraction Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 3
- 238000003709 image segmentation Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
Aiming at the limitation of the existing remote sensing image sea-land separation method, the invention provides a strategy rough and fine combined sea-land separation method applied to optical remote sensing ship detection. In the first step, a sea-land rough separation strategy based on a sea-land segmentation lookup table is adopted, and rapid rough-precision sea-land type classification is realized. The large-field remote sensing image can be quickly and accurately divided into three sea and land type areas with the coexistence of the remote-shore whole sea, the remote-shore whole land and the sea and land, and the subsequent sea and land division operation is only needed to be carried out on the condition of the coexistence of the sea and the land, so that the algorithm efficiency is greatly improved. And secondly, adopting a sea-land boundary region fine segmentation strategy based on a gradient integral graph, and performing segmentation based on the gradient integral graph on the region judged as the sea-land boundary type in the previous step. And thirdly, sea and land information generated in the coarse separation stage and sea and land information acquired in the fine segmentation stage are integrated, so that the large-field remote sensing image can be quickly and accurately divided into an ocean area and a land area, and guidance is provided for a subsequent ship detection task.
Description
Technical field
The present invention relates to a kind of point tactful coarse-fine combination sea land separation methods applied in optical remote sensing ship detection.
Background technology
In recent years, the ship detection in complicated big visual field is carried out using remote sensing images, it has also become ocean remote sensing application field
Research hotspot.Compared with the remote sensing application of typical local small field of view, most significant difference is for it：(1) real ship detects
In system operation, the big view field image of input is not only pure extra large image-region, it is also possible to including pure land and extra large land and the field deposited
Scape situation, and ship detection task is only concerned water area；And since the characteristics of image of land area is complex, at system
A large amount of false-alarms are will produce when reason, to seriously affect the timeliness and detection performance of detection system operation.(2) due to empty balance table
How in systems in practice the input data of the sensor of carrying, acquisition has the characteristics that breadth is big, data transfer rate is high, efficient land productivity
With limited computing resource, it is equally a critical issue to improve efficiency of algorithm.Therefore, how remote sensing images ship target is directed to
Detection task, it is the key technology for being badly in need of solving to design efficiently accurate extra large land separation method.
The method that remote sensing images sea land isolation technics is all based on greatly geography information (such as coastline database) at present, it is such
Method only needs the matching of geography information, therefore has higher computational efficiency, but there is also intrinsic defects.Such as：Seashore line number
It is often relatively low (general to arrive km magnitude roughly) according to library precision, and also there are buffer strips near coastline when operation.Therefore,
Although the extra large land separation method based on coastline database has higher computational efficiency, the precision of segmentation is not high, is easy to lead
The leak detection of littoral false-alarm or offshore ship is generated when cause ship detection.Another kind of method, be using image segmentation algorithm into
The extra large land of row remote sensing images detaches, and one of the technological approaches of mainstream at present.The precision of partitioning algorithm can reach Pixel-level, than
Coastline database separation method is accurate.But sea land partitioning algorithm is all based on greatly gray feature realization at present, and what is divided is accurate
Property be easy influenced by land complexity atural object grey-scale is more, difference is big.Further, since partitioning algorithm is based on Pixel-level
Operation, needed in the case of no guidance information traversal full figure all pixels, cause operand higher, influence practical application when
Effect property.Therefore, it is necessary to the application characteristic for big visual field remote sensing images ship detection, accurately and efficiently extra large land separation side is designed
Method.
Invention content
For the limitation of above-mentioned existing remote sensing images sea land separation method, the present invention proposes a kind of applied to optical remote sensing
Point tactful coarse-fine combination sea land separation method in ship detection, including：
The first step：Divide the extra large land crude separation of look-up table based on extra large land
(1.1) step divides the extra large land classification of type of look-up table based on extra large land.
The present invention uses the extra large land established by digital complex demodulation (Digital Elevation Model) segmentation to look into
Table is looked for, the extra large land classification of type for inputting big visual field remote sensing images is carried out.
(1.2) step judges the area type of present image according to matching result.
Current big visual field remote sensing images can be divided into three categories by the matching result based on previous step：The first kind is that remote bank is complete
Sea chart picture, the second class are the full land image of remote bank, and third class is land and sea junction area image, later according to area type point strategy
Carry out follow-up respective operations.
Second step：Land and sea junction region essence segmentation based on gradient integrogram
To solve to utilize intensity slicing method, accuracy to be easy by land complexity atural object grey-scale more, the big shadow of difference
Loud problem.The present invention proposes the land and sea junction region essence dividing method based on gradient integrogram.Crude separation on last stage is sentenced
Break and be used as input for region (simultaneously depositing region in extra large land) image of land and sea junction, carries out the Accurate Segmentation based on gradient integrogram.
The gradient integrogram in (2.1) step land and sea junction area calculates
First, difference processing is made to the land and sea junction types of image after mean value down-sampling, finds out Gradient Features image；It connects
It, integrogram is obtained on the basis of Gradient Features image, series steps are redrawn including generation look-up table, gradient integrogram.
Extra large land segmentation template of (2.2) step based on gradient integrogram obtains
Using bimodal mean value adaptivenon-uniform sampling algorithm, the integrogram after redrawing is split, obtains land mask two-value
Figure, the extra large land essence that land and sea junction region is then obtained by sequence of operations divide two-value template result.
Land and sea junction of (2.3) step based on extra large land template distinguishes strategy processing
According to the two-value template of the Hai Lu of acquisition the extraction of ship target is no longer carried out to being identified as the region on land；Mark
Know the region for ocean, then can enter the extraction process of subsequent ship target.Third walks：The sea of comprehensive crude separation and essence segmentation
Land information
The extra large land information that the extra large land information that the crude separation stage generates is obtained with the smart segmentation stage is integrated, will can greatly be regarded
Field remote sensing images are fast and accurately divided into ocean and land area, and guiding is provided to follow-up ship detection task.
A kind of point tactful coarse-fine combination sea land separation method applied in optical remote sensing ship detection, for quick and precisely
The land segmentation of carry out sea, including following series steps：
A) divide the extra large land crude separation of look-up table based on extra large land, including：
A1) first, the extra large land classification of type that look-up table is divided in extra large land is established based on digital elevation model；
A2) secondly, according to matching result, judge the area type of present image, quick and precisely by big visual field remote sensing images
Three kinds of extra large land type areas for being divided into the pure sea of large stretch of remote bank, large stretch of remote pure land of bank and extra large land and deposit,
Extra large, the large stretch of remote pure land type area of bank wherein pure to large stretch of remote bank directly carries out subsequent operation, it is not necessary to carry out sea again
The fine differentiation of land information, and fine extra large land cutting operation need to be only carried out to extra large land and the case where deposit, significant increase algorithm
Efficiency；
B) the land and sea junction region essence segmentation based on gradient integrogram, by being judged as land and sea junction type to previous step
Area image carries out the segmentation based on gradient integrogram：The gradient integrogram for finding out land and sea junction area first, secondly based on to ladder
Integrogram processing is spent, extra large land is obtained and divides two-value template, the land and sea junction for being finally based on extra large land template distinguishes strategy processing；
C) the extra large land information of comprehensive crude separation strategy and smart segmentation strategy, final merging obtain efficiently and accurately sea land segmentation knot
Fruit integrates the extra large land information that the extra large land information that the crude separation stage generates is obtained with the smart segmentation stage, can be distant by big visual field
Sense image is fast and accurately divided into ocean and land area,
Wherein：
Step A1) include：
Divide the extra large land classification of type of look-up table based on extra large land；Including using by digital complex demodulation (Digital
Elevation Model) extra large land segmentation look-up table is established, the extra large land classification of type for inputting big visual field remote sensing images is carried out, including
Using the elevation information of each latitude coordinates point in digital elevation model (DEM), height height above sea level is less than equal to 0
Extensive area is identified as ocean, and extensive area of the height height above sea level more than 0 is identified as land, and carries out expansive working to land
Small " cavity " is filled caused by low-lying in region, thus generates extra large land look-up table, wherein the coordinate and longitude and latitude of look-up table
It is a linear transformation relationship between degree coordinate, classification of type processing is by pending image-region geography latitude coordinates range
As input, and determine the corresponding region of four vertex longitudes and latitudes of Region Of Interest in a lookup table,
Based on the corresponding data in extra large land look-up table, the result of extra large land classification of type is obtained.Output the result is that
Description for this region Land-sea Distributions,
Step A2) include：
According to matching result, the area type of present image is judged, extra large land segmentation is established based on digital elevation model and is searched
The remote sensing images of current matching, can be divided into three categories by the matching result of the extra large land classification of type of table according to geographical location：First
Class is the full sea chart picture of remote bank, and the second class is the full land image of remote bank, and third class is land and sea junction area image, according to region class
Type, point strategy carry out subsequent operation to image, including：
When image is judged as the image in the full sea of remote bank, it will be directly entered the follow-up ship Objective extraction stage；
When image is judged as the image on the remote full land of bank, it will do not enter back into the ship Objective extraction stage；
When image is judged as land and sea junction area image, it will the stage for being directly entered subsequent fine segmentation is based on
Integrate the accurate extra large land segmentation of nomography.
Description of the drawings
Fig. 1 is point coarse-fine combination of strategy according to an embodiment of the invention applied in optical remote sensing ship detection
The flow chart of extra large land separation method.
Land and sea junction region essences of the Fig. 2 based on gradient integrogram divides flow chart.
Fig. 3 normal images generate look-up table schematic diagram.
Specific implementation mode
Below in conjunction with point strategy coarse-fine combination of the description of the drawings application according to the present invention in optical remote sensing ship detection
The embodiment of extra large land separation method.Point strategy according to an embodiment of the invention applied in optical remote sensing ship detection
The main-process stream of coarse-fine combination sea land separation method is as shown in Figure 1 comprising：
The first step：Divide the extra large land crude separation of look-up table based on extra large land, including：
(1.1) step divides the extra large land classification of type of look-up table based on extra large land.
The present invention is used establishes extra large land segmentation lookup by digital complex demodulation (Digital Elevation Model)
Table carries out the extra large land classification of type for inputting big visual field remote sensing images.This step is generated by extra large land look-up table and extra large land classification of type
Handle two steps composition.First, using the elevation information of each latitude coordinates point in DEM, height height above sea level is less than the big model equal to 0
It is ocean to enclose area identification, and extensive area of the height height above sea level more than 0 is identified as land, and carries out expansive working, to land area
Small " cavity " is filled caused by low-lying in domain.Thus generate extra large land look-up table, wherein the coordinate and longitude and latitude of look-up table
It is a linear transformation relationship between coordinate.Classification of type processing, is by pending image-region geography latitude coordinates range
As input, and determine the corresponding region of four vertex longitudes and latitudes of Region Of Interest in a lookup table.Then it is searched according to extra large land
Based on corresponding data in table, the result of extra large land classification of type is obtained.Output the result is that for this region Land-sea Distributions
Description.
(1.2) step judges the area type of present image according to matching result.
The remote sensing images of current matching can be divided into three categories by the matching result based on previous step according to geographical location：The
One kind is the full sea chart picture of remote bank, and the second class is the full land image of remote bank, and third class is land and sea junction area image.According to region class
Type, the present invention point strategy carry out subsequent operation to image.The concrete operations of three kinds of extra large land types are：When image is judged as remote bank
When the image in full sea, it will be directly entered the follow-up ship Objective extraction stage；When image is judged as the image on the remote full land of bank
When, it will the ship Objective extraction stage is not entered back into；When image is judged as land and sea junction area image, it will be directly entered
The stage of subsequent fine segmentation carries out the accurate extra large land segmentation based on integral nomography.
Second step：Land and sea junction region essence segmentation based on gradient integrogram
To solve to utilize intensity slicing method, accuracy to be easy by land complexity atural object grey-scale more, the big shadow of difference
Loud problem.The present invention proposes the land and sea junction region essence dividing method based on gradient integrogram.Crude separation on last stage is sentenced
Break and be used as input for region (simultaneously depositing region in extra large land) image of land and sea junction, carries out the Accurate Segmentation based on gradient integrogram.Its
Algorithm flow according to the method for the present invention is as shown in Figure 2.It specifically includes：
The gradient integrogram in (2.1) step land and sea junction area calculates
First, difference processing is made to the land and sea junction image after mean value down-sampling, finds out Gradient Features image；Then, exist
Gradient integrogram is obtained on the basis of Gradient Features image, series steps are redrawn including generation look-up table, gradient integrogram.
(2.1.1) step：Mean value down-sampling
Input picture is subjected to mean value down-sampling processing.Extra large land information belongs to the big structure information in image, can be to original
After image in different resolution carries out down-sampling operation, then carry out the extraction of extra large land separation information.It in this way can be in downscaled images size
While effectively inhibit spiced salt noise like.
(2.1.2) step：Calculate gradient image
To the image after previous step mean value down-sampling, corresponding Gradient Features are found out as difference by differential mode pixel-by-pixel
Image GFM (Gradient Feature Map):
GFM=max (Ghorz,Gvert)(3)
R and c indicates that the row and column of an image, x and y indicate the transverse and longitudinal coordinate position of each pixel respectively respectively, ginseng
Number IX, yIndicate the gray value of the position pixel (x, y), parameter GhorzIt is the approximation of horizontal gradient, parameter GvertIt is vertical gradient
Approximation.Formula (3) defines the maximum value in horizontal and vertical gradient as Gradient Features figure GFM.Gradient Features figure is main
What is characterized is the local grain and structure feature statistics in optical imagery.
(2.1.3) step：Generate look-up table
Gross area look-up table SAT (SumAreaTable) is generated from Gradient Features figure, the foundation of gross area look-up table SAT is
One step of key of integrogram is generated, and in any visual range, it is different that gross area look-up table SAT can quickly establish selection
Filter window size integrogram.Its according to the method for the present invention SAT look-up table operations such as Fig. 3 is generated by normal image
It is shown.The formula for generating SAT look-up tables is as follows：
Stexture=SX+a, y+b-SX+a, y-SX, y+b+SX, y (5)
SX, yIt is to indicate to be somebody's turn to do respectively from the sum of the texture in the region obtained in Gradient Features figure, x and y based on area size
The width and length in region, StextureIt indicates local grain integral fusion, is to pass through formula from neighbouring in SAT generation tables
(5) calculating is got.
(2.1.4) step：Integrogram is redrawn
By the following SAT matrix (6) of generation, the integrogram based on filtering is carried out to optical gradient characteristic image and is redrawn,
Scene is redrawn into region and the smoother region of texture for coarse texture, to obtain gradient integrogram.According to the line in marine site
Feature is managed, it is water area that the more smooth region of texture is given tacit consent in integrogram.Specific SAT integrogram redrawing process formula is such as
Under：
Formula (6) and formula (7) are accomplished that local grain integral fused filtering operation, the texture integrated value of regional area
It will enter in the center pixel of filter window.SAT_map in formula (6) indicates SAT look-up tables, is gradient integrogram
Summation in a region, m and n are the size in the region respectively, and T (i, j) is the texture information of Grad summation, public
Integral_image in formula (7) indicates that integrogram is calculated by (5) and (6) and gets that indicate integrogram redraws step
Suddenly, smooth and coarse texture region integrogram can be obtained.Parameter n indicates filter window size.
Extra large land segmentation template of (2.2) step based on gradient integrogram obtains
Using bimodal mean value adaptivenon-uniform sampling algorithm, the integrogram after redrawing is split, obtains land mask, then
The extra large land essence segmentation two-value template result in the region of land and sea junction is obtained by a series of morphological operations.
(2.2.1) step：Adaptivenon-uniform sampling
To the gradient integrogram that previous step obtains, the binarization operation of bimodal mean value adaptivenon-uniform sampling algorithm is carried out, is obtained
Land area is set to 1 by the two-value original template that can identify Hai Lu, and sea area is set to 0.
(2.2.2) step Morphological scale-space
A series of morphological operations such as expanded, corroded to the integrogram after segmentation.It realizes in land area due to office
" cavity " and " fracture " phenomenon caused by portion's texture smooth region, to obtain capable of more characterizing the extra large Lu Mo of extra large land big structure information
Plate.
(2.2.3) step dissipates zonule and rejects
Because the original intention of this method is to obtain the big structure information that land and ocean are distinguished, therefore, use in two-value here
Value filtering, which is rejected, dissipates small land area.
Land and sea junction of (2.3) step based on extra large land template distinguishes strategy processing
According to the two-value template of the Hai Lu of acquisition the extraction of ship target is no longer carried out to being identified as the region on land；It is right
It is identified as sea area, into the extraction flow of subsequent ship target.
Third walks：The extra large land information of comprehensive crude separation and essence segmentation
The extra large land information that the comprehensive crude separation stage generates, the extra large land information obtained with the smart segmentation stage can be distant by big visual field
Sense image is fast and accurately divided into ocean and land area, and guiding is provided to follow-up ship detection task.We are crude separation
In be judged to the remote full sea chart picture of bank and essence segmentation in be judged to sea area image carry out anastomosing and splicing, be uniformly input to follow-up ship
The Objective extraction stage.We are the image for being judged to be judged to land area in crude separation in the remote full land image of bank and essence segmentation simultaneously
Anastomosing and splicing is carried out, it is unified no longer to carry out ship Objective extraction step.
The present invention has the following advantages compared with existing detection method:
Since the separation of current remote sensing images sea land is all based on greatly the extra large land separation method of coastline database, although calculating
Efficiency is higher, but segmentation precision is not high, and littoral false-alarm or coastal-going ship missing inspection are also easy to produce in ship detection.On the other hand,
Image segmentation algorithm is all based on greatly gray feature realization, and the accuracy of segmentation is easy by land complexity atural object grey-scale
Influence more, difference is big.Further, since partitioning algorithm is the operation of Pixel-level, need to traverse in the case of no guidance information
Full figure all pixels, operand is higher, influences the timeliness of practical application.Therefore, for problem in terms of three above, the present invention
Point strategy, the more means and methods that coarse-fine combination is proposed in big visual field remote sensing ship detection flow scheme design, specifically include and are based on
The extra large land crude separation of coastline database, the land and sea junction region essence segmentation based on gradient integrogram and comprehensive crude separation and essence point
Three major parts of extra large land information cut.
There are three prominent features for present invention tool：
First, this method proposes crude separation and the essence strategy that is combined of segmentation reaching more accurate, highly efficient
Extra large land segmentation effect.By coarse segmentation, big visual field remote sensing images can be fast and accurately divided into remote bank, and sea, remote bank are complete entirely
Land and extra large land and the three kinds of extra large land type areas deposited, wherein the above two directly carry out subsequent operation, it is not necessary to carry out extra large land information again
Fine differentiation；And it only need to be to extra large land and the extra large land cutting operation of the type deposited progress finely.The strategy is preferably utilized thick
The advantages of separation and essence segmentation, also preferably compensates for crude separation and essence divides respective disadvantage.
Secondly, crude separation realizes quick rough grade sea land separation plan using the method for establishing the seas DEM land segmentation look-up table
Slightly, which can reduce unnecessary processing (the extra large land class in large stretch of remote full sea of bank and large stretch of remote full land of bank in succeeding target extraction
Type), the timeliness of entire extra large land division processing method can be substantially improved.
Finally, in the smart segmentation stage, to overcome extra large land image segmentation algorithm most of at present to be realized based on gray feature,
The accuracy of segmentation is easy to be influenced by land complexity atural object grey-scale is more, difference is big.In view of large stretch of marine site and land
Significant difference of the region in terms of texture, the present invention is using the land and sea junction region essence segmentation based on gradient integrogram, effectively
Extra large land separation in the case of realizing extra large land and depositing.
Claims (2)
1. a kind of point tactful coarse-fine combination sea land separation method applied in optical remote sensing ship detection, for fast and accurately
Carry out extra large land segmentation, including following series steps：
A) divide the extra large land crude separation of look-up table based on extra large land, including：
A1) first, the extra large land classification of type that look-up table is divided in extra large land is established based on digital elevation model；
A2) secondly, according to matching result, judge the area type of present image, big visual field remote sensing images are fast and accurately drawn
The three kinds of extra large land type areas for being divided into the pure sea of large stretch of remote bank, large stretch of remote pure land of bank and extra large land and depositing,
Extra large, the large stretch of remote pure land type area of bank wherein pure to large stretch of remote bank directly carries out subsequent operation, it is not necessary to carry out extra large land letter again
The fine differentiation of breath, and fine extra large land cutting operation need to be only carried out to extra large land and the case where deposit, significant increase efficiency of algorithm；
B) the land and sea junction region essence segmentation based on gradient integrogram, by the region for being judged as land and sea junction type to previous step
Image carries out the segmentation based on gradient integrogram：The gradient integrogram for finding out land and sea junction area first, is accumulated secondly based on to gradient
Component processing obtains extra large land and divides two-value template, and the land and sea junction for being finally based on extra large land template distinguishes strategy processing；
C) the extra large land information of comprehensive crude separation strategy and smart segmentation strategy, final merging obtain efficiently and accurately sea land segmentation result,
The extra large land information that the extra large land information that the crude separation stage generates is obtained with the smart segmentation stage is integrated, it can be by big visual field remote sensing figure
As being fast and accurately divided into ocean and land area,
Wherein：
Step A1) include：
Divide the extra large land classification of type of look-up table based on extra large land；Including using by digital complex demodulation (Digital
Elevation Model) extra large land segmentation look-up table is established, the extra large land classification of type for inputting big visual field remote sensing images is carried out, including
Using the elevation information of each latitude coordinates point in digital elevation model (DEM), height height above sea level is less than the big model equal to 0
It is ocean to enclose area identification, and extensive area of the height height above sea level more than 0 is identified as land, and carries out expansive working to land area
In it is low-lying caused by small " cavity " be filled, thus generate extra large land look-up table, wherein the coordinate of look-up table is sat with longitude and latitude
A linear transformation relationship between mark, classification of type processing be using pending image-region geography latitude coordinates range as
Input, and determine the corresponding region of four vertex longitudes and latitudes of Region Of Interest in a lookup table,
Based on the corresponding data in extra large land look-up table, the result of extra large land classification of type is obtained.Output the result is that for
The description of this region Land-sea Distributions,
Step A2) include：
According to matching result, the area type of present image is judged, establishing extra large land based on digital elevation model divides look-up table
The remote sensing images of current matching can be divided into three categories by the matching result of extra large land classification of type according to geographical location：The first kind is
The full sea chart picture of remote bank, the second class are the full land image of remote bank, and third class is land and sea junction area image, according to area type, is divided
Strategy carries out subsequent operation to image, including：
When image is judged as the image in the full sea of remote bank, it will be directly entered the follow-up ship Objective extraction stage；
When image is judged as the image on the remote full land of bank, it will do not enter back into the ship Objective extraction stage；
When image is judged as land and sea junction area image, it will the stage for being directly entered subsequent fine segmentation carries out based on integral
The accurate extra large land segmentation of nomography.
2. according to claim 1 point of strategy coarse-fine combination sea land separation method, it is characterised in that, will in the smart segmentation stage
The extra large land crude separation stage is judged as that the area image of land and sea junction as input, carries out the Accurate Segmentation based on gradient integrogram,
Including：
The gradient integrogram in land and sea junction area calculates, wherein first, making at difference to the land and sea junction image after mean value down-sampling
Reason, finds out Gradient Features image, and gradient integrogram is then obtained on the basis of Gradient Features image, and series steps include generating
Look-up table, gradient integrogram are redrawn,
Extra large land segmentation template based on gradient integrogram obtains, including the use of bimodal mean value adaptivenon-uniform sampling algorithm, after redrawing
Integrogram be split, obtain land mask, the sea in the region of land and sea junction then obtained by a series of morphological operations
Land essence segmentation two-value template as a result,
Land and sea junction based on extra large land template distinguishes strategy processing, includes the two-value template according to the Hai Lu of acquisition, to being identified as
The region on land no longer carries out the extraction of ship target；To being identified as sea area, into the extraction stream of subsequent ship target
Journey.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810113117.9A CN108376247A (en) | 2018-02-05 | 2018-02-05 | Strategic coarse-fine combined sea-land separation method applied to optical remote sensing ship detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810113117.9A CN108376247A (en) | 2018-02-05 | 2018-02-05 | Strategic coarse-fine combined sea-land separation method applied to optical remote sensing ship detection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108376247A true CN108376247A (en) | 2018-08-07 |
Family
ID=63017296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810113117.9A Pending CN108376247A (en) | 2018-02-05 | 2018-02-05 | Strategic coarse-fine combined sea-land separation method applied to optical remote sensing ship detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108376247A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109886134A (en) * | 2019-01-25 | 2019-06-14 | 北京市遥感信息研究所 | The extra large land information index coding method of high-resolution global digital elevation model image |
| CN110221257A (en) * | 2019-07-02 | 2019-09-10 | 中国航空工业集团公司雷华电子技术研究所 | A kind of sea-surface target detection false alarm rejection method, apparatus and radar system |
| CN110648347A (en) * | 2019-09-24 | 2020-01-03 | 北京航天宏图信息技术股份有限公司 | Coastline extraction method and device based on remote sensing image |
| CN110866926A (en) * | 2019-10-24 | 2020-03-06 | 北京航空航天大学 | Infrared remote sensing image rapid and fine sea-land segmentation method |
| CN112085656A (en) * | 2020-08-31 | 2020-12-15 | 厦门蓝海天信息技术有限公司 | Ocean mode data shoreline extension method and system, mobile terminal and storage medium |
| CN112150482A (en) * | 2020-09-17 | 2020-12-29 | 北京航空航天大学 | Sea-land segmentation method based on elevation information and normalized sea water index |
| CN112434642A (en) * | 2020-12-07 | 2021-03-02 | 北京航空航天大学 | Sea-land segmentation method suitable for processing large-scene optical remote sensing data |
| CN114742854A (en) * | 2022-04-02 | 2022-07-12 | 西安电子科技大学 | SAR image sea-land segmentation method based on scene prior knowledge and region combination |
| CN115830448A (en) * | 2022-11-30 | 2023-03-21 | 广州市地质调查院(广州市地质环境监测中心) | Remote sensing image contrast analysis method based on multi-view fusion |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8170282B1 (en) * | 2009-09-17 | 2012-05-01 | Sandia Corporation | Technique for ship/wake detection |
| CN102663348A (en) * | 2012-03-21 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Marine ship detection method in optical remote sensing image |
| CN105513041A (en) * | 2015-10-28 | 2016-04-20 | 深圳大学 | Large-scale remote sensing image sea-land segmentation method and system |
| CN106709914A (en) * | 2017-01-05 | 2017-05-24 | 北方工业大学 | SAR image ship detection false alarm eliminating method based on two-stage DEM sea-land reservoir |
| CN107230210A (en) * | 2017-06-19 | 2017-10-03 | 长光卫星技术有限公司 | A kind of fast partition method of remote sensing images harbour waterborne target |
-
2018
- 2018-02-05 CN CN201810113117.9A patent/CN108376247A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8170282B1 (en) * | 2009-09-17 | 2012-05-01 | Sandia Corporation | Technique for ship/wake detection |
| CN102663348A (en) * | 2012-03-21 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Marine ship detection method in optical remote sensing image |
| CN105513041A (en) * | 2015-10-28 | 2016-04-20 | 深圳大学 | Large-scale remote sensing image sea-land segmentation method and system |
| CN106709914A (en) * | 2017-01-05 | 2017-05-24 | 北方工业大学 | SAR image ship detection false alarm eliminating method based on two-stage DEM sea-land reservoir |
| CN107230210A (en) * | 2017-06-19 | 2017-10-03 | 长光卫星技术有限公司 | A kind of fast partition method of remote sensing images harbour waterborne target |
Non-Patent Citations (1)
| Title |
|---|
| 周浩天 等: "多尺度积分图重建的红外遥感图像海陆分割算法", 《计算机工程与应用》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109886134A (en) * | 2019-01-25 | 2019-06-14 | 北京市遥感信息研究所 | The extra large land information index coding method of high-resolution global digital elevation model image |
| CN110221257A (en) * | 2019-07-02 | 2019-09-10 | 中国航空工业集团公司雷华电子技术研究所 | A kind of sea-surface target detection false alarm rejection method, apparatus and radar system |
| CN110648347A (en) * | 2019-09-24 | 2020-01-03 | 北京航天宏图信息技术股份有限公司 | Coastline extraction method and device based on remote sensing image |
| CN110866926B (en) * | 2019-10-24 | 2022-12-20 | 北京航空航天大学 | Infrared remote sensing image rapid and fine sea-land segmentation method |
| CN110866926A (en) * | 2019-10-24 | 2020-03-06 | 北京航空航天大学 | Infrared remote sensing image rapid and fine sea-land segmentation method |
| CN112085656A (en) * | 2020-08-31 | 2020-12-15 | 厦门蓝海天信息技术有限公司 | Ocean mode data shoreline extension method and system, mobile terminal and storage medium |
| CN112085656B (en) * | 2020-08-31 | 2023-12-05 | 厦门蓝海天信息技术有限公司 | Ocean mode data shoreline extension method, system, mobile terminal and storage medium |
| CN112150482A (en) * | 2020-09-17 | 2020-12-29 | 北京航空航天大学 | Sea-land segmentation method based on elevation information and normalized sea water index |
| CN112434642A (en) * | 2020-12-07 | 2021-03-02 | 北京航空航天大学 | Sea-land segmentation method suitable for processing large-scene optical remote sensing data |
| CN112434642B (en) * | 2020-12-07 | 2022-11-11 | 北京航空航天大学 | Sea-land segmentation method suitable for processing large-scene optical remote sensing data |
| CN114742854A (en) * | 2022-04-02 | 2022-07-12 | 西安电子科技大学 | SAR image sea-land segmentation method based on scene prior knowledge and region combination |
| CN115830448A (en) * | 2022-11-30 | 2023-03-21 | 广州市地质调查院(广州市地质环境监测中心) | Remote sensing image contrast analysis method based on multi-view fusion |
| CN115830448B (en) * | 2022-11-30 | 2024-02-09 | 广州市地质调查院(广州市地质环境监测中心) | Remote sensing image comparison analysis method based on multi-view fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108376247A (en) | Strategic coarse-fine combined sea-land separation method applied to optical remote sensing ship detection | |
| CN109934163B (en) | Aerial image vehicle detection method based on scene prior and feature re-fusion | |
| CN105513041B (en) | A kind of method and system of large format remote sensing images sea land segmentation | |
| CN112084869B (en) | Compact quadrilateral representation-based building target detection method | |
| CN109389163B (en) | Unmanned aerial vehicle image classification system and method based on topographic map | |
| CN107248159A (en) | A kind of metal works defect inspection method based on binocular vision | |
| CN112149594B (en) | Urban construction assessment method based on deep learning and high-resolution satellite images | |
| CN111027511B (en) | Remote sensing image ship detection method based on region of interest block extraction | |
| CN108121991A (en) | A kind of deep learning Ship Target Detection method based on the extraction of edge candidate region | |
| Turner et al. | CZM applications of Argus coastal imaging at the Gold Coast, Australia | |
| CN112287832A (en) | High-resolution remote sensing image-based urban illegal building detection method | |
| CN107818303B (en) | Unmanned aerial vehicle oil and gas pipeline image automatic contrast analysis method, system and software memory | |
| CN107451982A (en) | A kind of high canopy density standing forest tree crown area acquisition methods based on unmanned plane image | |
| CN109271928A (en) | A kind of road network automatic update method based on the fusion of vector road network with the verifying of high score remote sensing image | |
| CN109919944A (en) | A kind of joint super-pixel figure of complex scene building variation detection cuts optimization method | |
| CN109377480A (en) | Arable land use change detection method based on deep learning | |
| CN104992172B (en) | Port remote sensing image shore ship detection method based on coastal outburst and different area scanning | |
| CN113505842B (en) | Automatic urban building extraction method suitable for large-scale regional remote sensing image | |
| CN101976436A (en) | Pixel-level multi-focus image fusion method based on correction of differential image | |
| CN103871072A (en) | Method for automatic extraction of orthoimage embedding line based on projection digital elevation models | |
| CN113920436A (en) | Remote sensing image marine vessel recognition system and method based on improved YOLOv4 algorithm | |
| CN108537816A (en) | A kind of obvious object dividing method connecting priori with background based on super-pixel | |
| AU2020103470A4 (en) | Shadow Detection for High-resolution Orthorectificed Imagery through Multi-level Integral Relaxation Matching Driven by Artificial Shadows | |
| CN103700063B (en) | Topography integration quick mapping method based on high definition satellite image | |
| CN113610873B (en) | Noctilucent remote sensing image data processing method and device |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180807 |
|
| RJ01 | Rejection of invention patent application after publication |