CN109801306A - Tidal creek extracting method based on high score remote sensing image - Google Patents
Tidal creek extracting method based on high score remote sensing image Download PDFInfo
- Publication number
- CN109801306A CN109801306A CN201910059401.7A CN201910059401A CN109801306A CN 109801306 A CN109801306 A CN 109801306A CN 201910059401 A CN201910059401 A CN 201910059401A CN 109801306 A CN109801306 A CN 109801306A
- Authority
- CN
- China
- Prior art keywords
- tidal creek
- tidal
- creek
- tiny
- remote sensing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
A kind of tidal creek extracting method based on high score remote sensing image is disclosed, method includes the following steps: to acquire high-resolution remote sensing image, and remote sensing image eliminates atmospheric interference via radiation calibration and atmospheric correction.The strategy of context of methods is separately to extract broad tidal creek and tiny tidal creek, wherein the tidal creek that width is more than or equal to predetermined pixel is broad tidal creek, and the tidal creek that width is less than predetermined pixel is tiny tidal creek.Broad tidal creek is extracted using normalization water body index (NDWI) and maximum variance between clusters (OTSU).Using SEaTH algorithm, tiny tidal creek extraction is carried out using separability, the wave band for selecting tidal creek and tidal flat to differ greatly between J-M (Jeffries-Matusita) distance calculating class.It for the wave band that SEaTH algorithm filters out, is carried out after uniforming heterogeneous background using improved FCM Algorithms, there is the tiny tidal creek of gaussian-shape using the enhancing of multiple dimensioned Gauss matched filtering.Then it uses and tiny tidal creek is extracted based on the adaptive threshold fuzziness of global mean value and standard deviation.Then the clast patch in tiny tidal creek is removed.Finally tiny tidal creek is merged with broad tidal creek using ditch logic or operation, forms complete tidal creek.
Description
Technical field
The present invention relates to remote sensing technology field, especially a kind of tidal creek extracting method based on high score remote sensing image.
Background technique
Tidal creek is a kind of develop in intertidal zone (especially between flour sand Muddy Bottoms tide as geomorphologic factor typical on tidal flat
Band), the tide gate formed by drive marine (especially tidal action) is that tidal flat and Creek system itself and the external world are continuous
The important channel of substance, energy and information exchange is carried out, the supply of distribution and silt to tidewater all plays vital
Effect directly reflects the characteristic of tidal flat, is one of the important parameter that people study coastal region hydrology connectivity.With constructing
The development of the development activities such as engineering and mudflat aquaculture, the development of tidal creek are greatly affected.Therefore tidal creek is accurately extracted
Spatial distribution characteristic has very important theoretical significance and practical application value.
By the severe image of natural conditions, part coastal region accessibility is poor, and sea water intrusion is serious in addition, tidal creek shape
State variation is more frequent, and traditional field survey method can not effectively carry out.The characteristics of tidal creek is that 1) tidal creek width is different, tide
Ditch width is differed from the thousands of rice of the several tens cm on several intertidal zone tops to from the estuary of intertidal zone lower part, and width difference is significant.
2) tidal flat background is heterogeneous strong, and tidal creek is developed from intertidal zone lower part, flows through mudflat, shrub-marsh and three kinds of marsh
Different Wetland Types, in three regions, tidal flat water content, vegetation coverage and tidal creek silt content have larger difference.3)
Tidal creek anisotropy is strong, and different segmentation sequences can show the preferred orientations for deviateing random network.These features are to tidal creek
Fast accurate proposes very big challenge.The existing remote sensing image visual interpretation method used, which exists, to be taken time and effort and leads
The big disadvantage of the property seen, is not suitable for the large area repeated measures applied to tidal creek.In view of tidal creek growth course continue it is unstable
Property, the technical method for developing the long-term dynamics monitoring of a set of tidal creek evolutionary process faces urgent demand.
In geometric shape, tidal creek is similar to the land network of waterways, blood vessel and road in winding linear distribution, many
Scholar conducts in-depth research for the automatically or semi-automatically extraction of the classes tidal creek elements such as road, blood and the land network of waterways, related
Extracting method is generally divided into following several: at flow path modelling, supervised classification, object oriented classification method and high vision
Manage class method.Wherein, flow path modelling depends on high-precision dem data, vulnerable to the influence of structural fault and crack, and
It is only capable of obtaining the tidal creek of single pixel width, has ignored the width information of tidal creek；Supervised classification is based on data-driven, antinoise
Interference performance is poor, cannot keep the spatial continuity of geographic element, inevitably generates spiced salt phenomenon；Object oriented classification
Method is directed to high resolution image, is homogeney region by Image Segmentation, but is influenced by intertidal zone tidal flat water content,
Developing a large amount of tiny tidal creeks in this region can not effectively distinguish；High vision processing class method is conceived to the threadiness point of tidal creek
Cloth feature is generally divided into three linear enhancing, segmentation and post-processing steps, and produces preferably effect in many regions.
For coastal tidal creek dynamic monitoring, airborne LiDAR DEM is mostly used mutually to tie with aerophotogrammetry data at present
The method of conjunction extracts tidal creek, but the following great number observation cost limits the data and sees in based Interpretation of Remote Sensing Images
Widespread adoption and popularization in survey.Satellite remote-sensing image is relatively low by its high-timeliness, extensive repeated measures and price
Honest and clean characteristic can monitor for coastal tidal creek and provide the data supporting of long-term sequence.Particularly, not with space technology
Disconnected progress, the service band of sensor are constantly expanded, and spatial resolution is also quickly improving, distant to littoral zone dynamic
Sense monitoring is filled with new vitality.
Disclosed above- mentioned information are used only for enhancing the understanding to background of the present invention in the background section, therefore can
It can be comprising not constituting the information of the prior art known to a person of ordinary skill in the art in home.
Summary of the invention
Problems to be solved by the invention
As described above, the present invention needs to provide a kind of tidal creek extracting method based on high score remote sensing image, this paper presents
A kind of tidal creek extracting method based on digital image processing techniques uses the multispectral shadow such as domestic high score two (GF-2)
As basic data source.It is extracted first with normalization water body index (NDWI) and maximum variance between clusters (OTSU) broad
Tidal creek.Secondly, weakening tide using improved FCM Algorithms (MFCM) and multiple dimensioned Gauss matched filtering (MGMF)
Enhance tiny tidal creek on the basis of beach background is heterogeneous.Tiny tidal creek is extracted followed by adaptive threshold fuzziness.Finally merge
Tiny tidal creek and broad tidal creek form complete tidal creek network.The precision that automatically extracts of the invention significantly improves and realizes length
Time dynamic continuously automatically extracts.
Solution to problem
The inventors of the present invention have made intensive studies in order to achieve the above objectives, specifically, the present invention provides a kind of base
In the tidal creek extracting method of high score remote sensing image comprising following step:
First step, acquisition include the high-resolution remote sensing image of multispectral data, remote sensing image via radiation calibration and
Atmospheric correction eliminates atmospheric interference,
Second step separately extracts broad tidal creek and tiny tidal creek, wherein width is more than or equal to predetermined pixel
Tidal creek is broad tidal creek, and the tidal creek that width is less than predetermined pixel is tiny tidal creek.Using SEaTH algorithm, J-M is utilized
(Jeffries-Matusita) distance calculates separability between class, and the wave band for selecting tidal creek and tidal flat to differ greatly carries out tiny
Tidal creek extracts.SeaTH algorithmic formula is as follows:
J-M range formula is as follows:
J=2 (1-e-B), wherein
J represents the distance between two classifications, mi(i=1,2) and σi(i=1,2) mean value and side of two classes are respectively indicated
Difference,
Third step extracts broad tidal creek based on maximum variance between clusters, when meeting inter-class variance and reaching maximum, obtains
Obtain optimum segmentation threshold value k*, wherein maximum variance between clusters formula is as follows:
Wherein,Be threshold value be k when inter-class variance, mGIt is full figure average gray, ash when m (k) is threshold value k
Degree, L is the number of grey levels of image, P1(k) it is probability that pixel is classified into class 1,
Four steps uniforms heterogeneous background, wherein initialization cluster centre viWith biased field βj, update degree of membership letter
Number μij, cluster centre viWith biased field βj, as ‖ vb+1-vbWhen ‖ < ε, front and back twice calculated cluster centre be less than it is scheduled
Convergence threshold stops calculating, and obtains background and uniforms data, wherein i is cluster, and j is pixel, and ε is scheduled convergence threshold；b
The number of iterations, v are referred to b+1b+1It is iteration b+1 times cluster centre, vbIt is iteration b times cluster centre,
5th step, multiple dimensioned Gauss matched filtering enhancing have the tiny tidal creek of gaussian-shape, and filter rotation obtains not
The response of equidirectional tidal creek only retains the maximum response of its multiple directions for each pixel.By the matching of multiple scales
The response results of filter are multiplied after normalization,
6th step, the tiny tidal creek of adaptive threshold fuzziness divide tiny tidal creek based on global mean value and standard deviation
Cut, threshold formula: T=mean+k*std, wherein mean is global mean value；Std is that global criteria is poor；T is optimal threshold, ginseng
For the range of number k between 0.01-1, the pixel by value greater than T is divided into tiny tidal creek, obtains tiny tidal creek coarse segmentation result.
Finally go the isolated patch that area is less than given threshold to obtain the final segmentation result of tiny tidal creek.
7th step merges the tiny tidal creek of the 6th step with the broad tidal creek logic of third step or operation to be formed
Complete tidal creek.
In the tidal creek extracting method based on high score remote sensing image, in first step, the multispectral data via
Multispectral camera acquisition, the multispectral data includes the blue wave band that wavelength is 0.45-0.52 microns, wavelength 0.52-
0.59 micron of green light band, the red spectral band that wavelength is 0.63-0.69 microns and wavelength are 0.77-0.89 microns close red
Wave section, the spatial resolution of image are 4m.
In the tidal creek extracting method based on high score remote sensing image, in first step, selection is based on MODTRAN spoke
The FLAASH atmospheric correction algorithm for penetrating mode carries out atmospheric correction to image.
In the tidal creek extracting method based on high score remote sensing image, in second step, making a reservation for a pixel is 5.It adopts
Select tidal creek and tidal flat poor using separability between J-M (Jeffries-Matusita) distance calculating class with SEaTH algorithm
Different biggish wave band carries out tiny tidal creek extraction, mud bank regional choice green light band, sabkha regional choice NDWI wave band, wherein
NDWI=(GREEN-NIR)/(GREEN+NIR)), GREEN refers to that green light band, NIR refer near infrared band.
Observation r in the tidal creek extracting method based on high score remote sensing image, in four steps, at pixel jj
It is by true value xjWith biased field βjWhat addition obtained, by subtracting biased field βj, we can obtain true anti-at pixel j
Rate is penetrated, formula is as follows:
rj=xj+βj,j∈[1,N]
Global objective function J is,
Wherein C indicates the quantity of cluster centre；N indicates the sum of pixel number in image；The power of α control neighborhood effect；Indicate the average value of pixel in pixel j neighborhood window；μijIndicate that pixel j belongs to the subordinating degree function of class i；M is degree of membership
The factor is set as 2；viIndicate the cluster centre of class i；J is global objective function, when global objective function reaches extreme value, is obtained
Data are uniformed to background.
In the tidal creek extracting method based on high score remote sensing image, in the 5th step, is matched and filtered using dimensional Gaussian
Wave device effectively enhances linear tidal creek, and the dimensional Gaussian matched filter is defined as follows:
Wherein σ indicates the distribution of intensity,；L be convolution mask along y-axis length with smooth noise, x refers to convolution mould
The length of plate x-axis；M (x, y) is dimensional Gaussian matched filter；- g " (x) refers to one-dimensional Gauss matched filtering deviceIt is taken after x and y both direction derivation negative.
In the tidal creek extracting method based on high score remote sensing image, in the 5th step, the setting of filter rotation steps
It is 10 °, filter rotates 18 times, to cover the tidal creek in all possible directions.The length of matched filter x-axis is set as | x |=
3 σ, the width of filter y-axis are set as 5 to detect the long tidal creek of 5 pixels.Tidal creek width to be reinforced 1-5 pixel it
Between,W is the half of tidal creek width to be reinforced, and σ is the standard deviation of Gaussian function second dervative, indicates intensity
Distribution.By observing image to be reinforced, the width of the most tidal creek to be reinforced of discovery concentrates on 3 and 5 pixels, optimal σ
For 0.9 and 1.4, calculated to simplify, by the σ of mud bank tidal creek and sabkha tidal creek1And σ2Uniformly it is set as 1 and 1.5.
In the tidal creek extracting method based on high score remote sensing image, in four steps, cluster centre is according to original shadow
After carrying out selection screening as histogram, selected at the wave crest and trough of histogram in conjunction with the sample point of selection.MFCM algorithm pair
α is insensitive, and α is set as 0.1.The cluster centre in mud bank region is [0.1601；0.1846], the cluster centre in sabkha region is set
It is set to [0.3198；-0.2344].
In the tidal creek extracting method based on high score remote sensing image, the 6th step, in adaptive threshold fuzziness, ginseng
Number k is 0.2, and the scale that clast patch is rejected is set as 50.
Advantageous effects of the invention:
The two-dimensional geometry form and tidal flat background heterogeneity of tidal creek itself complexity are to cause to be difficult to mention from optical image
Take the key reason of tidal creek.Change width feature and section gaussian-shape feature based on tidal creek, the present invention are directed to No. two 4m of high score
Multispectral image proposes a set of coastal tidal creek automatic Extraction Algorithm: 1. dividing broad tidal creek by OTSU combination water body index；
2. being extracted using MFCM, MGMF and adaptive threshold fuzziness and dividing tiny tidal creek；3. merging broad tidal creek and tiny tidal creek extracting
As a result complete tidal creek network is formed.Overcome coastal the tidal creek apparent difficult point of scales' change, complete extraction under heterogeneous background
The tidal creek of different in width in Yellow River delta region.Quantitative precision evaluation shows that context of methods overall accuracy is greater than
97%, Kappa coefficient are greater than 0.8, and misclassification error is lower than 20%, and leakage divides error lower than 11%, compared to traditional supervised classification
Method has clear improvement, and while reduction mistake is divided, more completely remains tiny tidal creek.Most of parameter in the algorithm
It can be obtained by image self-characteristic, be reduced under the premise of guaranteeing algorithm automatization level for subjective experience
It relies on.
The above description is only an overview of the technical scheme of the present invention, in order to make technological means of the invention more clear
Chu Mingbai, reaches the degree that those skilled in the art can be implemented in accordance with the contents of the specification, and in order to allow this hair
Bright above and other objects, features and advantages can be more clearly understood, and be lifted below with a specific embodiment of the invention
Example explanation.
Detailed description of the invention
[Fig. 1] shows the step of tidal creek extracting method based on high score remote sensing image of one embodiment of the invention and shows
It is intended to.
[Fig. 2] shows the flow chart of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention.
[Fig. 3] shows the homogenization of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention
Background flow chart.
The background that [Fig. 4] shows the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention is equal
One changes display diagram.
[Fig. 5] shows the multiple dimensioned of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention
The tidal creek cross-sectional view of enhancing.
[Fig. 6] shows the multiple dimensioned of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention
Enhance display diagram.
[Fig. 7] shows the adaptive of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention
Threshold value relational graph.
[Fig. 8] shows the extraction of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention
Result schematic diagram.
[Fig. 9] shows the extraction of the tidal creek extracting method based on high score remote sensing image of another embodiment of the present invention
Result schematic diagram.
[Figure 10] shows the tidal creek extracting method and not Tongfang based on high score remote sensing image of one embodiment of the invention
Method extracts Comparative result schematic diagram.
Specific embodiment
The specific embodiment that the present invention will be described in more detail below with reference to accompanying drawings.Although showing the present invention in attached drawing
Specific embodiment, it being understood, however, that may be realized in various forms the present invention without should be by embodiments set forth here institute
Limitation.It is to be able to thoroughly understand the present invention on the contrary, providing these embodiments, and can be by the scope of the present invention
It is fully disclosed to those skilled in the art.
It should be noted that having used some vocabulary in the specification and claims to censure specific components.Ability
Field technique personnel it would be appreciated that, technical staff may call the same component with different nouns.This specification and right
It is required that not in such a way that the difference of noun is as component is distinguished, but with the difference of component functionally as differentiation
Criterion."comprising" or " comprising " as mentioned throughout the specification and claims are an open language, therefore are answered
It is construed to " including but not limited to ".Specification subsequent descriptions are to implement better embodiment of the invention, the right description
It is the range that is not intended to limit the invention for the purpose of the rule of specification.Protection scope of the present invention is worked as appended by view
Subject to as defined in claim.
In order to facilitate understanding of embodiments of the present invention, it is done by taking several specific embodiments as an example into one below in conjunction with attached drawing
The explanation of step, and each attached drawing does not constitute the restriction to the embodiment of the present invention.
Specifically, the step schematic diagram of the tidal creek extracting method based on high score remote sensing image as shown in Figure 1.A kind of base
Include the following steps: in the tidal creek extracting method of high score remote sensing image
First step S1, acquisition include the high-resolution remote sensing image of multispectral data, and remote sensing image is via radiation calibration
And atmospheric correction eliminates atmospheric interference,
Second step S2, sense image eliminate atmospheric interference via radiation calibration and atmospheric correction,
Second step (S2), broad tidal creek and tiny tidal creek are separately extracted, wherein width is more than or equal to predetermined picture
The tidal creek of element is broad tidal creek, and the tidal creek that width is less than predetermined pixel is tiny tidal creek.Using SEaTH algorithm, J-M is utilized
(Jeffries-Matusita) distance calculates separability between class, and the wave band for selecting tidal creek and tidal flat to differ greatly carries out tiny
Tidal creek extracts.SEaTH algorithmic formula is as follows:
J-M range formula is as follows:
J=2 (1-e-B), wherein
J represents the distance between two classifications, mi(i=1,2) and σi(i=1,2) mean value and side of two classes are respectively indicated
Difference,
Third step S3 extracts broad tidal creek based on maximum variance between clusters, when meeting inter-class variance and reaching maximum,
Obtain optimum segmentation threshold value k*, wherein maximum variance between clusters formula is as follows:
Wherein,Be threshold value be k when inter-class variance, mGIt is full figure average gray, ash when m (k) is threshold value k
Degree, L is the number of grey levels of image, P1(k) it is probability that pixel is classified into class 1,
Four steps S4 uniforms heterogeneous background, wherein initialization cluster centre viWith biased field βj, update degree of membership
Function muij, cluster centre viWith biased field βj, as ‖ vb+1-vbWhen ‖ < ε, front and back twice calculated cluster centre be less than it is predetermined
Convergence threshold stop calculate, obtain background uniform data, wherein i be cluster, j is pixel, and ε is scheduled convergence threshold；
B and b+1 refers to the number of iterations, vb+1It is iteration b+1 times cluster centre, vbIt is iteration b times cluster centre,
5th step S5, multiple dimensioned Gauss matched filtering enhancing have the tiny tidal creek of gaussian-shape, and filter rotation obtains
The response of different directions tidal creek only retains the maximum response of its multiple directions for each pixel.By of multiple scales
Response results with filter are multiplied after normalization,
6th step S6, the tiny tidal creek of adaptive threshold fuzziness carry out tiny tidal creek based on global mean value and standard deviation
Segmentation, threshold formula: T=mean+k*std, wherein mean is global mean value；Std is that global criteria is poor；T is optimal threshold,
For the range of parameter k between 0.01-1, the pixel by value greater than T is divided into tiny tidal creek, obtains tiny tidal creek coarse segmentation knot
Fruit.Finally go the isolated patch that area is less than given threshold to obtain the final segmentation result of tiny tidal creek.
7th step S7 closes the tiny tidal creek of the 6th step S6 with the broad tidal creek logic of third step S3 or operation
And form complete tidal creek.
The tidal creek of the homogenization of tidal creek extracting method combination background and linear enhancing based on high score remote sensing image of the invention
Extraction algorithm has excellent expansibility, and cluster centre can be to combine sample point in histogram wave crest and trough after coarse sizing
Place's selection；The parameters such as the size and intensity distribution of Gauss matched filtering device are according to can be according to the width and length of tidal creek to be reinforced
Degree is chosen；Optimal adaptive threshold is calculated by the method for being stepped up step-length, is improved in different sensors shadow
The applicability of picture and different zones.
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, in first step S1, institute
It states multispectral data to acquire via multispectral camera, the multispectral data includes the blue light wave that wavelength is 0.45-0.52 microns
Section, the green light band that wavelength is 0.52-0.59 microns, the red spectral band that wavelength is 0.63-0.69 microns and wavelength are 0.77-
0.89 micron of near infrared band, the spatial resolution of image are 4m.
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, in first step S1, choosing
It selects the FLAASH atmospheric correction algorithm based on MODTRAN radiative transfer model and atmospheric correction is carried out to image.
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, in second step S2, in advance
Fixed pixel is 5.Using SEaTH algorithm, separability between class is calculated using J-M (Jeffries-Matusita) distance,
The wave band that selection tidal creek and tidal flat differ greatly carries out tiny tidal creek extraction, mud bank regional choice green light band, the choosing of sabkha region
Select NDWI wave band, wherein NDWI=(GREEN-NIR)/(GREEN+NIR)), GREEN refers to green light band, what NIR referred to
It is near infrared band.
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, in four steps S4, as
Observation r at plain jjIt is by true value xjWith biased field βjWhat addition obtained, by subtracting biased field βj, we can obtain
Real reflectance at pixel j, formula are as follows:
rj=xj+βj,j∈[1,N]
Global objective function J is,
Wherein C indicates the quantity of cluster centre；N indicates the sum of pixel number in image；The power of α control neighborhood effect；Indicate the average value of pixel in pixel j neighborhood window；μijIndicate that pixel j belongs to the subordinating degree function of class i；M is degree of membership
The factor is set as 2；viIndicate the cluster centre of class i；J is global objective function, when global objective function reaches extreme value, is obtained
Data are uniformed to background.
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, in the 5th step S5, make
Effectively enhance linear tidal creek with dimensional Gaussian matched filter, the dimensional Gaussian matched filter is defined as follows:
Wherein σ indicates the distribution of intensity,；L be convolution mask along y-axis length with smooth noise, x refers to convolution mould
The length of plate x-axis；M (x, y) is dimensional Gaussian matched filter；- g " (x) is referred to one-dimensional Gauss matched filtering deviceIt is taken after x and y both direction derivation negative.
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, in the 5th step S5, filter
Wave device rotation steps are set as 10 °, and filter rotates 18 times, to cover the tidal creek in all possible directions.Matched filter x-axis
Length be set as | x |=3 σ, the width of filter y-axis are set as 5 to detect the long tidal creeks of 5 pixels.Tidal creek to be reinforced is wide
It spends between 1-5 pixel,W is the half of tidal creek width to be reinforced, and σ is Gaussian function second dervative
Standard deviation indicates the distribution of intensity.By observing image to be reinforced, the width of the most tidal creek to be reinforced of discovery concentrates on 3
With 5 pixels, optimal σ is 0.9 and 1.4, is calculated to simplify, by the σ of mud bank tidal creek and sabkha tidal creek1And σ2Uniformly it is set as
1 and 1.5.
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, in four steps S4, gather
After class center carries out selection screening according to raw video histogram, in conjunction with selection sample point histogram wave crest and trough
Place's selection.MFCM algorithm is insensitive to α, and α is set as 0.1.The cluster centre in mud bank region is [0.1601；0.1846], sabkha
The cluster centre in region is set as [0.3198；-0.2344].
In the preferred embodiment of the tidal creek extracting method based on high score remote sensing image, the 6th step S6 is adaptive
It answers in Threshold segmentation, parameter k is 0.2, and the scale that clast patch is rejected is set as 50.
For a further understanding of the present invention, in one embodiment, high score two (GF-2) multispectral data conducts are chosen
Data, imaging time are on August 13rd, 2016, and it can be subsequent that the period tidal level is lower and salt marsh vegetation development is vigorous
Influence of the salt marsh vegetation development to tidal creek provides reference.This is PMS2, band class information and space using the sensor of image
Resolution ratio is shown in Table No. two image parameters information of 1 high score.Pre-treatment step includes radiation calibration atmospheric correction, dry to eliminate atmosphere
It disturbs, restores true Reflectivity for Growing Season.Absolute calibration coefficient used in pretreatment is from China Resource Satellite Applied Center official
Net obtains, and atmospheric correction models select the FLASSH atmospheric correction based on MODTRAN model, and pre-treatment step is in ENVI5.3
It completes.
Table No. two image parameters information of 1 high score
Fig. 2 shows the flow charts of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention.From
Tidal creek is extracted in No. two images of high score, it is as shown in Figure 2 that broad tidal creek with tiny tidal creek is separately extracted flow chart.According on the spot
Investigation discovery, width is densely distributed less than the tidal creek of 20m and is difficult to extract, therefore the tidal creek by width less than 5 pixels herein
It is defined as tiny tidal creek.Tiny tidal creek extraction step is as follows: (1) complex background uniforms, and sabkha tidal creek region is merely with normalizing
Tidal creek and tidal flat can preferably be distinguished by changing water body index (NDWI, NDWI=(GREEN-NIR)/(GREEN+NIR)).Mud
Beach tidal creek region uses SEaTH algorithm, utilizes separability, selection between J-M (Jeffries-Matusita) distance calculating class
The green light band that tidal creek and tidal flat differ greatly.Herein on basis, for the wave band of selection, improved fuzzy C-mean algorithm is utilized
Algorithm (MFCM), the otherness comparison caused by inhibiting tidal flat background heterogeneous between target and background.(2) multiple dimensioned linear increasing
By force, there is the tiny tidal creek of gaussian-shape using multiple dimensioned Gauss matched filtering (MGMF) enhancing, determined change width with solving tidal creek
Problem.(3) convolutional filter weakens the influence that the strong anisotropy of tidal creek is brought to cover all possible direction.(4) it uses
The tiny tidal creek of adaptive threshold fuzziness.(5) clast patch removes.Seawater is full of inside broad tidal creek, it can by NDWI and most
Big Ostu method (OTSU) is divided.Finally tiny tidal creek is merged with broad tidal creek result using logic or operation, shape
At complete tidal creek.
When carrying out Clean water withdraw, traditional hard clustering method, which is subjected to the probabilistic influence of remote sensing image, to be caused point
There are ambiguities for class result.Actually under this complex background of tidal flat, the boundary between water body and non-water body be it is fuzzy, because
Fuzzy clustering method can be used to solve this ambiguity in this.Fuzzy C-mean algorithm (FCM) algorithm of standard is very quick for noise
Sense, and the spatial distribution of pixel is not accounted for, which prevent it in Remote Sensing Image Segmentation.
Fig. 3 shows the homogenization back of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention
Scape flow chart assumes initially that the observation r at pixel jjIt is by true value xjWith biased field βjWhat addition obtained, by subtracting
Biased field βj, we can obtain the real reflectance at pixel j, and formula is as follows:
rj=xj+βj,j∈[1,N]
Then, adjacent space information is integrated into standard FCM algorithm, to reduce the influence of isolated noise bring.
Wherein C indicates the quantity of cluster centre；N indicates the sum of pixel number in image；The power of α control neighborhood effect；Indicate the average value of pixel in pixel j neighborhood window；μijIndicate that pixel j belongs to the subordinating degree function of class i；M is degree of membership
The factor is usually arranged as 2；viIndicate the cluster centre of class i；J is global objective function.When objective function reaches extreme value, to
Image of classifying obtains optimal classification scheme.Subordinating degree function μ is introduced thusijConstraint condition:
The background that Fig. 4 shows the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention is uniform
Change display diagram, wherein referring specifically to Fig. 4, (a) green light band raw video；(b) green light band histogram；(c) background uniforms
As a result；(d) background uniforms histogram.
Longitudinally varying and ambient background pixel mixed spectra effect by tidal creek is influenced, and tiny tidal creek and image are carried on the back
Scape contrast is poor, can not effectively identify merely with OTSU, although more small tidal creeks can be obtained by reducing threshold value, can also lead
Cause more noises.By observing the cross section of tidal creek, it can be found that the cross section of tidal creek is in down U or the shape of falling V, Ke Yiyong more
Gaussian curve is next approximate, therefore, can effectively enhance linear tidal creek using dimensional Gaussian matched filter.Fig. 5 shows this hair
The tidal creek cross-sectional view of the multi-scale enhancement of the tidal creek extracting method based on high score remote sensing image of bright one embodiment, referring to
Fig. 5.
Dimensional Gaussian matched filter is defined as follows:
Wherein σ indicates the distribution of intensity, its value is closely related with the width of target tidal creek；L is convolution mask along y-axis
Length with smooth noise.
In view of the strong anisotropy of Creek system, need to rotate filter into possible angle, to obtain different directions
The response of tidal creek only retains the maximum response of its multiple directions for each pixel.Rotation steps are set as herein
10 °, filter rotates 18 times, to cover the tidal creek in all possible directions.In view of tidal creek width of span is very big, use is single
The matched filter of scale can not enhance all tiny tidal creeks, may cause many tiny tidal creeks by Gaussian noise institute
It floods.For this purpose, solving the problems, such as this using multi-scale enhancement method, the response results of the matched filter of multiple scales are existed
It is multiplied after normalization, line feature can be further enhanced while suppressing noise.Fig. 6 shows a reality of the invention
Apply the multi-scale enhancement display diagram of the tidal creek extracting method based on high score remote sensing image of example.Referring specifically to fig. 6, wherein (a)
Scale 1 enhances result；(b) scale 2 enhances result；(c) multi-scale enhancement result.
It is higher by the calculated optimal threshold of OTSU in tiny tidal creek cutting procedure, lead to many fuzzy tiny tides
Ditch is ignored, for this purpose, threshold value is public using combining the adaptive threshold of global mean value and standard deviation to be split tiny tidal creek
Formula is as follows:
T=mean+k*std
Wherein mean is global mean value；Std is that global criteria is poor；T is optimal threshold；Pixel by value greater than T is divided into
Tidal creek.
In one embodiment, the parameter of tidal creek extraction algorithm is divided into three parts (1): improved FCM Algorithms
Cluster centre；(2) length, width and the σ of multiple dimensioned Gauss matched filtering device；(3) threshold value in adaptive threshold fuzziness
k。
After cluster centre carries out selection screening according to raw video histogram, in conjunction with selection sample point in histogram
It is selected at wave crest and trough.The cluster centre of mud bank tidal creek herein is [0.1601；0.1846], in the cluster of sabkha tidal creek
The heart is set as [0.3198；-0.2344]；MFCM algorithm is insensitive to α, and α is uniformly set as 0.1.
Since below Gaussian curve 99% or more area is between [- 3 σ, 3 σ], therefore by matched filter x-axis
Length is set as | x |=3 σ, the width of filter y-axis are set as 5 to detect the long tidal creek of 5 pixels, and 4 test sections wait increasing
Strong tidal creek width is between 1-5 pixel.In multi-scale enhancement,The corresponding optimal σ of tiny tidal creek is
Between 0.3-1.4, by observing image to be reinforced, the width of the most tidal creek to be reinforced of discovery concentrates on 3 and 5 pixels, most
Excellent σ is 0.9 and 1.4, is calculated to simplify, by the σ of mud bank tidal creek and sabkha tidal creek1And σ2Uniformly it is set as 1 and 1.5.
In adaptive threshold fuzziness, the setting of parameter k is affected to extraction accuracy, how between mistake point and leakage mention
Keeping balance is the key that choose optimal threshold, and the range of k is arranged thus between 0.01-1, is step-length with 0.01, when
When Kappa coefficient highest, global optimum's threshold value is obtained.As shown in Figure 10, in the maximum of four test zone Kappa coefficients
At 0.2,0.25,0.23 and 0.22, all near 0.2, therefore 0.2 is set by k.The scale setting that clast patch is rejected
It is 50.Fig. 7 shows the adaptive threshold of the tidal creek extracting method based on high score remote sensing image of one embodiment of the invention
Relational graph, wherein the relationship of threshold value k and Kappa coefficient is as shown in Figure 7.
In order to examine reliability of the invention, the tide of complex background homogenization and the enhancing of multiple dimensioned line feature will be combined
The same maximum likelihood method of ditch extracting method (Maximum Likelihood, ML) and support vector machines (Support Vector
Machine, SVM) it compares when selecting training sample, mud bank tidal creek is divided into intertidal zone, supratidal zone, tiny tidal creek and width
Wealthy four class of tidal creek；Sabkha tidal creek is divided into cloud, tidal flat, five class of vegetation, tiny tidal creek and broad tidal creek.Use Radial basis kernel function
The kernel function of (Radial Basis Function, RBF) as SVM, parameters setting in, select Kappa coefficient highest
Final result of the person as two kinds of control methods.
Square is obscured by the Confusion Matrix-Using Ground Truth Image module calculating of ENVI5.3
Battle array (Confusion Matrix) carries out precision evaluation to result is extracted.This chooses overall accuracy (Overall
Accuracy, OA), kappa coefficient (Kappa Coefficient, KC), leakage divide error (Omission, EO) and misclassification error
(Commission, EC) is used as evaluation index.
Context of methods realizes that the complete extraction result of mud bank tidal creek and sabkha tidal creek is such as in MATLAB 2017a platform
Shown in Fig. 8 and Fig. 9, four kinds of methods extraction Comparative results of regional area are as shown in Figure 10, and regional area distinct methods extract result
Quantitative assessment be shown in Table 2.
Interpretation comparison is as can be seen that context of methods is more fully extracted most of tidal creeks, especially width by visual observation
Tiny tidal creek between 1-5 pixel, extraction effect is preferable, and the integrality of tidal creek and continuous holding are preferable.
In four test zones, only for visual interpretation angle, ML and SVM are not very in terms of tiny tidal creek extraction
Ideal, four test zones all produce serious leakage point phenomenon.Especially in mud bank tidal creek region, a large amount of tidal flats are by mistake point
For tidal creek.
Quantitative accuracy evaluation shows combination complex background homogenization proposed by the present invention and the enhancing of multiple dimensioned line feature
Tidal creek extracting method in Kappa coefficient, overall accuracy and leakage divide in terms of error better than other three kinds of methods, and Kappa coefficient is equal
Greater than 0.8, show that the result of this method extraction and ground measured data degree of agreement are more preferable.
In two pieces of mud bank tidal creek regions, water content is lower in tiny tidal creek, tiny tidal creek and tidal flat for spectrum angle
The thin difference of the two is smaller, produces same object different images phenomenon, has seriously affected the extraction of linear ground object and tiny atural object.Cause
When choosing sample, degree of isolation is not satisfactory between class, and supervised classification can not effectively identify tiny tidal creek, tidal flat and tiny tidal creek
It can not efficiently separate, every evaluation index is generally relatively low.From the point of view of geometric shape angle, tidal flat is in block distribution, tidal creek performance
Apparent line feature out can greatly be inhibited after reducing background interference using MFCM using line feature enhancing
The response of non-linear shape element reduces error.
The evaluation of 2 tidal creek extraction accuracy of table
Industrial applicibility
Tidal creek extracting method based on high score remote sensing image of the invention can use in remote sensing detection field.
Although embodiment of the present invention is described in conjunction with attached drawing above, the invention is not limited to above-mentioned
Specific embodiments and applications field, above-mentioned specific embodiment is only schematical, directiveness, rather than is limited
Property processed.Those skilled in the art are protected under the enlightenment of this specification and not departing from the claims in the present invention
Range in the case where, a variety of forms can also be made, these belong to the column of protection of the invention.
Claims (9)
1. a kind of tidal creek extracting method based on high score remote sensing image comprising following step:
First step (S1), acquisition include the high-resolution remote sensing image of multispectral data, remote sensing image via radiation calibration and
Atmospheric correction eliminates atmospheric interference,
Second step (S2), broad tidal creek and tiny tidal creek are separately extracted, wherein width is more than or equal to the tide of predetermined pixel
Ditch is broad tidal creek, and the tidal creek that width is less than predetermined pixel is tiny tidal creek.Using SEaTH algorithm, J-M is utilized
(Jeffries-Matusita) distance calculates separability between class, and the wave band for selecting tidal creek and tidal flat to differ greatly carries out tiny
Tidal creek extracts.SEaTH algorithmic formula is as follows:
J-M range formula is as follows:
J=2 (1-e-B), wherein
J represents the distance between two classifications, mi(i=1,2) and σi(i=1,2) respectively indicates the mean value and variance of two classes,
Third step (S3) extracts broad tidal creek based on maximum variance between clusters, when meeting inter-class variance and reaching maximum, obtains
Optimum segmentation threshold value k*, wherein maximum variance between clusters formula is as follows:
Wherein,Be threshold value be k when inter-class variance, mGIt is full figure average gray, gray scale when m (k) is threshold value k, L is
The number of grey levels of image, P1(k) it is probability that pixel is classified into class 1,
Four steps (S4) uniforms heterogeneous background, wherein initialization cluster centre viWith biased field βj, update subordinating degree function
μij, cluster centre viWith biased field βj, as ‖ vb+1-vbWhen ‖ < ε, calculated cluster centre is less than scheduled convergence twice for front and back
Threshold value stops calculating, and obtains background and uniforms data, wherein i is cluster, and j is pixel, and ε is scheduled convergence threshold；B and b+1
Refer to the number of iterations, vb+1It is iteration b+1 times cluster centre, vbIt is iteration b times cluster centre,
5th step (S5), multiple dimensioned Gauss matched filtering enhancing have the tiny tidal creek of gaussian-shape, and filter rotation obtains not
The response of equidirectional tidal creek only retains the maximum response of its multiple directions for each pixel.The matching of multiple scales is filtered
The response results of wave device are multiplied after normalization,
6th step (S6), the tiny tidal creek of adaptive threshold fuzziness divide tiny tidal creek based on global mean value and standard deviation
Cut, threshold formula: T=mean+k*std, wherein mean is global mean value；Std is that global criteria is poor；T is optimal threshold, parameter
For the range of k between 0.01-1, the pixel by value greater than T is divided into tiny tidal creek, obtains tiny tidal creek coarse segmentation result.Finally
It goes the isolated patch that area is less than given threshold to obtain the final segmentation result of tiny tidal creek.
7th step (S7) closes the broad tidal creek logic of the same third step of tiny tidal creek (S3) of the 6th step (S6) or operation
And form complete tidal creek.
2. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that first step
(S1) in, the multispectral data is acquired via multispectral camera, and the multispectral data includes that wavelength is 0.45-0.52 microns
Blue wave band, the green light band that wavelength is 0.52-0.59 microns, wavelength be 0.63-0.69 microns red spectral band and wavelength
For 0.77-0.89 microns of near infrared band, the spatial resolution of image is 4m.
3. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that first step
(S1) in, the FLAASH atmospheric correction algorithm based on MODTRAN radiative transfer model is selected to carry out atmospheric correction to image.
4. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that second step
(S2) in, making a reservation for a pixel is 5.It, can between calculating class using J-M (Jeffries-Matusita) distance using SEaTH algorithm
Separation property, the wave band for selecting tidal creek and tidal flat to differ greatly carry out tiny tidal creek extraction, mud bank regional choice green light band, sabkha
Regional choice NDWI wave band, wherein NDWI=(GREEN-NIR)/(GREEN+NIR)), GREEN refers to green light band, NIR
Refer near infrared band.
5. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that four steps
(S4) the observation r in, at pixel jjIt is by true value xjWith biased field βjWhat addition obtained, by subtracting biased field βj, we
The real reflectance at pixel j can be obtained, formula is as follows:
rj=xj+βj,j∈[1,N]
Global objective function J is,
Wherein C indicates the quantity of cluster centre；N indicates the sum of pixel number in image；The power of α control neighborhood effect；It indicates
The average value of pixel in pixel j neighborhood window；μijIndicate that pixel j belongs to the subordinating degree function of class i；M is the degree of membership factor, setting
It is 2；viIndicate the cluster centre of class i；J is global objective function, and when global objective function reaches extreme value, it is uniform to obtain background
Change data.
6. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that the 5th step
(S5) in, effectively enhance linear tidal creek using dimensional Gaussian matched filter, the dimensional Gaussian matched filter is defined as follows:
Wherein σ indicates the distribution of intensity,；L be convolution mask along y-axis length with smooth noise, x refers to convolution mask x-axis
Length；M (x, y) is dimensional Gaussian matched filter；- g " (x) is referred to one-dimensional Gauss matched filtering deviceIt is taken after x and y both direction derivation negative.
7. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that the 5th step
(S5) in, filter rotation steps are set as 10 °, and filter rotates 18 times, to cover the tidal creek in all possible directions.Matching filter
The length of wave device x-axis is set as | x |=3 σ, the width of filter y-axis are set as 5 to detect the long tidal creek of 5 pixels.It is to be reinforced
Tidal creek width between 1-5 pixel,W is the half of tidal creek width to be reinforced, and σ is that Gaussian function second order is led
Several standard deviations indicates the distribution of intensity.By observing image to be reinforced, the width of the most tidal creek to be reinforced of discovery is concentrated on
3 and 5 pixels, optimal σ are 0.9 and 1.4, are calculated to simplify, by the σ of mud bank tidal creek and sabkha tidal creek1And σ2Uniformly it is set as
1 and 1.5.
8. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that four steps
(S4) in, after cluster centre carries out selection screening according to raw video histogram, in conjunction with selection sample point histogram wave
It is selected at peak and trough.MFCM algorithm is insensitive to α, and α is set as 0.1.The cluster centre in mud bank region is [0.1601；
0.1846], the cluster centre in sabkha region is set as [0.3198；-0.2344].
9. the tidal creek extracting method according to claim 1 based on high score remote sensing image, which is characterized in that the 6th step
(S6), in adaptive threshold fuzziness, parameter k is 0.2, and the scale that clast patch is rejected is set as 50.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910059401.7A CN109801306B (en) | 2019-01-22 | 2019-01-22 | Tidal ditch extraction method based on high resolution remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910059401.7A CN109801306B (en) | 2019-01-22 | 2019-01-22 | Tidal ditch extraction method based on high resolution remote sensing image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109801306A true CN109801306A (en) | 2019-05-24 |
| CN109801306B CN109801306B (en) | 2020-10-27 |
Family
ID=66559980
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910059401.7A Active CN109801306B (en) | 2019-01-22 | 2019-01-22 | Tidal ditch extraction method based on high resolution remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109801306B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110689549A (en) * | 2019-09-30 | 2020-01-14 | 北京航天宏图信息技术股份有限公司 | Object extraction method, device and equipment |
| CN110991019A (en) * | 2019-11-22 | 2020-04-10 | 首都师范大学 | Automatic classification algorithm for coastal tidal ditches |
| CN117346744A (en) * | 2023-12-04 | 2024-01-05 | 山东科技大学 | Method for inverting measured water depth based on satellite-borne active and passive remote sensing information during rising and falling tide |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101634705A (en) * | 2009-08-19 | 2010-01-27 | 西安电子科技大学 | Method for detecting target changes of SAR images based on direction information measure |
| KR20180094702A (en) * | 2017-02-16 | 2018-08-24 | 목포대학교산학협력단 | Apparatus and method for extraction of tidal creeks in tidal flat zones using drone |
-
2019
- 2019-01-22 CN CN201910059401.7A patent/CN109801306B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101634705A (en) * | 2009-08-19 | 2010-01-27 | 西安电子科技大学 | Method for detecting target changes of SAR images based on direction information measure |
| KR20180094702A (en) * | 2017-02-16 | 2018-08-24 | 목포대학교산학협력단 | Apparatus and method for extraction of tidal creeks in tidal flat zones using drone |
Non-Patent Citations (2)
| Title |
|---|
| LIU MEIJIE等: "PCA-based sea-ice image fusion of optical data by HIS transform and SAR data by wavelet transform", 《ACTA OCEANOL.SIN》 * |
| 孙超等: "近25a来江苏中部沿海盐沼分布时空演变及围垦影响分析", 《自然资源学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110689549A (en) * | 2019-09-30 | 2020-01-14 | 北京航天宏图信息技术股份有限公司 | Object extraction method, device and equipment |
| CN110689549B (en) * | 2019-09-30 | 2020-12-18 | 北京航天宏图信息技术股份有限公司 | Object extraction method, device and equipment |
| CN110991019A (en) * | 2019-11-22 | 2020-04-10 | 首都师范大学 | Automatic classification algorithm for coastal tidal ditches |
| CN110991019B (en) * | 2019-11-22 | 2023-05-05 | 首都师范大学 | Automatic grading algorithm for coastal tidal ditches |
| CN117346744A (en) * | 2023-12-04 | 2024-01-05 | 山东科技大学 | Method for inverting measured water depth based on satellite-borne active and passive remote sensing information during rising and falling tide |
| CN117346744B (en) * | 2023-12-04 | 2024-03-19 | 山东科技大学 | Method for inverting measured water depth based on satellite-borne active and passive remote sensing information during rising and falling tide |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109801306B (en) | 2020-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cheng et al. | SeNet: Structured edge network for sea–land segmentation | |
| Gavazzi et al. | Evaluation of seabed mapping methods for fine-scale classification of extremely shallow benthic habitats–application to the Venice Lagoon, Italy | |
| Witharana et al. | Understanding the synergies of deep learning and data fusion of multispectral and panchromatic high resolution commercial satellite imagery for automated ice-wedge polygon detection | |
| Son et al. | Mangrove mapping and change detection in Ca Mau Peninsula, Vietnam, using Landsat data and object-based image analysis | |
| CN109190538A (en) | A kind of fully-loaded stream river deposition coast band evolution analysis method based on remote sensing technology | |
| Hou et al. | Marine floating raft aquaculture extraction of hyperspectral remote sensing images based decision tree algorithm | |
| Smeeckaert et al. | Large-scale classification of water areas using airborne topographic lidar data | |
| CN109448016A (en) | It is a kind of based on object-oriented and its be subordinate to rule remote sensing image tidal saltmarsh method | |
| CN109801306A (en) | Tidal creek extracting method based on high score remote sensing image | |
| CN109740485B (en) | Reservoir or small reservoir identification method based on spectral analysis and deep convolutional neural network | |
| CN113408460B (en) | Method for detecting spartina alterniflora distribution based on remote sensing big data and cloud platform | |
| CN110414509A (en) | Stop Ship Detection in harbour based on the segmentation of extra large land and feature pyramid network | |
| Kaiser | Environmental changes, remote sensing, and infrastructure development: The case of Egypt's East Port Said harbour | |
| CN114266958A (en) | Cloud platform based mangrove remote sensing rapid and accurate extraction method | |
| CN114612769A (en) | Integrated sensing infrared imaging ship detection method integrated with local structure information | |
| CN111597930A (en) | Coastline extraction method based on remote sensing cloud platform | |
| Moniruzzaman et al. | Imaging and classification techniques for seagrass mapping and monitoring: A comprehensive survey | |
| CN109886221A (en) | Sand dredger recognition methods based on saliency detection | |
| Liu et al. | Mapping China’s offshore mariculture based on dense time-series optical and radar data | |
| Zhao et al. | Stability evaluation of tidal flats based on time-series satellite images: A case study of the Jiangsu central coast, China | |
| Zhao et al. | Identifying large-area mangrove distribution based on remote sensing: A binary classification approach considering subclasses of non-mangroves | |
| Rajendiran et al. | Pixel level feature extraction and machine learning classification for water body extraction | |
| CN106290389B (en) | The algal tufa and non-algal tufa condition classification method of a kind of eutrophic lake MODIS images | |
| Urbański et al. | Object-oriented classification of QuickBird data for mapping seagrass spatial structure | |
| Zhu et al. | Spatiotemporal variations of total suspended matter in complex archipelagic regions using a sigmoid model and Landsat-8 imagery |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |