CN110308097A - A kind of satellite image cloud detection method of optic and system - Google Patents
A kind of satellite image cloud detection method of optic and system Download PDFInfo
- Publication number
- CN110308097A CN110308097A CN201910602938.3A CN201910602938A CN110308097A CN 110308097 A CN110308097 A CN 110308097A CN 201910602938 A CN201910602938 A CN 201910602938A CN 110308097 A CN110308097 A CN 110308097A
- Authority
- CN
- China
- Prior art keywords
- cloud
- pixel
- earth
- reflectivity
- ground surface
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N2021/1793—Remote sensing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Chemical & Material Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Image Processing (AREA)
Abstract
The present invention discloses a kind of satellite image cloud detection method of optic and system.The cloud detection method of optic includes: to obtain satellite image to be detected；The satellite image is divided by ground surface type, obtains the satellite image of different ground surface types；Cloud detection is carried out according to threshold value corresponding to ground surface type in the satellite image of each ground surface type, obtains the cloud atlas picture under each ground surface type.Corresponding threshold value is arranged for different ground surface types in cloud detection method of optic and system of the invention, and the accuracy for improving whole cloud detection result reduces the difference degree of different zones cloud detection accuracy simultaneously.
Description
Technical field
The present invention relates to cloud detection fields, more particularly to a kind of satellite image cloud detection method of optic and system.
Background technique
Cloud shows white or linen light tone in the visible light and near infrared band of remote sensing images, and has higher
Reflectivity Characteristics.The principle of cloud detection is exactly to utilize cloud in visible light and near infrared band and vegetation, soil, rock and waters
What the difference in reflectivity of equal underlying surfaces was detected, cloud reflectivity with higher.Remote sensing images cloud detection method of optic is numerous, main
Want method that can be divided into threshold method, statistical method and artificial neural network method three categories.Wherein, threshold method because its algorithm is simple,
Easy to operate, the higher feature of precision, is widely used in the pretreatment of all kinds of remote sensing images, and principle is to pass through Yun Yudian
The type object spectrum difference or bright temperature difference is different realizes cloud detection.Threshold method common at present has ISCCP algorithm, APOLLO to calculate
Method, CLAVR algorithm, mainly rule of thumb given threshold carries out cloud detection to these methods.ISCCP method assumes observation radiation
Two kinds of atmospheric conditions of clear sky and cloud only are from, the amount characterized by 0.6 μm of visible light and 11 mu m waveband data of Infrared window uses
Threshold method carries out cloud detection.The pixel that CLAVR method then uses 2*2 differentiated, and when according to underlay surface properties and observation
Between difference, algorithm is divided into ocean on daytime, four class of land on daytime, night ocean and night land.APPOLLO method is used
Five full resolution detection channels data of AVHRR realize that cloud pixel is known in each path setting threshold value according to object feature
Not.For LANDSAT series data, Zhu et al. proposes Fmask algorithm, which utilizes the physical characteristic of cloud, utilizes one
Serial threshold testing extracts potential cloud layer, isolates cloud, cloud shade and snow.Existing threshold method does not consider ground surface type
Influence to threshold value causes the accuracy of the cloud detection under different ground surface types to differ greatly, the cloud detection in respective regions
Accuracy is lower.
Summary of the invention
The object of the present invention is to provide a kind of satellite image cloud detection method of optic and systems, set for different ground surface types
Corresponding threshold value is set, the accuracy for improving whole cloud detection result reduces the difference journey of different zones cloud detection accuracy simultaneously
Degree.
To achieve the above object, the present invention provides following schemes:
A kind of satellite image cloud detection method of optic, comprising:
Obtain satellite image to be detected；
The satellite image is divided by ground surface type, obtains the satellite image of different ground surface types；
Cloud detection is carried out according to threshold value corresponding to ground surface type in the satellite image of each ground surface type, obtains various regions
Cloud atlas picture under table type.
Optionally, described that the satellite image is divided by ground surface type, obtain the satellite mapping of different ground surface types
Picture specifically includes:
The ground surface type is divided into wetland, water body, artificial earth's surface, bare area, sea according to history ground surface type data
Domain, arable land, forest, meadow and shrub；The wherein wetland, the water body, the artificial earth's surface, the bare area and the sea
Domain belongs to constant attribute earth's surface；The arable land, the forest, the meadow and the shrub belong to change to attributes earth's surface.
Optionally, described to carry out cloud inspection according to threshold value corresponding to ground surface type in the satellite image of each ground surface type
It surveys, obtains the cloud atlas picture under each ground surface type, specifically include:
For each constant attribute earth's surface, chooses dark blue-cirrus band index and ground surface type and cloud sector is not spent most
Big wave band carries out cloud detection；Dark blue-cirrus the band index is
Wherein CCI is dark blue-cirrus band index, and Coastal is dark blue wave band reflectivity, and Cirrus is cirrus wave band
Reflectivity；
For each change to attributes earth's surface, the threshold of each wave band is determined according to latitude locating for each ground surface type and season
It is worth and carries out cloud detection in conjunction with the dark blue-cirrus band index.
Optionally, described to be directed to each constant attribute earth's surface, choose dark blue-cirrus band index and ground surface type
Maximum wave band is not spent with cloud sector and carries out cloud detection, is specifically included:
For the sea area, 0.06 ∩ NIR > of formula (0.12 ∪ Green > 0.08 of Blue >) ∩ Red > will be met
The pixel of 0.06 CCI≤0.95 ∪ is determined as cloud pixel；
For the water body, 0.08 CCI≤0.95 ∪ 0.13 ∪ Green > of formula Blue >, 0.10 ∪ Red > will be met
Pixel be determined as cloud pixel；
For the wetland, 0.13 CCI≤0.95 ∪ 0.13 ∪ Green > of formula Blue >, 0.15 ∪ Red > will be met
Pixel be determined as cloud pixel；
For the bare area, the pixel for meeting 0.95 ∪ Blue > 0.25 of formula CCI < is determined as cloud pixel；
For the artificial earth's surface, formula (298 ∩ NDVI of (0.25 ∪ Green > 0.2 of Blue >) ∩ BT < will be met
< 0.3) ∩ BT < Tcor_artiThe pixel of ∪ Cirrus > 0.0025 is determined as cloud pixel；
Wherein Blue indicates that the reflectivity of blue wave band, Green indicate that the reflectivity of green light band, Red indicate feux rouges wave
The reflectivity of section, NIR indicate that the reflectivity of near infrared band, BT indicate bright temperature value, and NDVI indicates normalized differential vegetation index,
Tcor_artiIndicate that the bright temperature of dynamic corrects threshold value.
Optionally, described to be directed to each change to attributes earth's surface, it is determined according to latitude locating for each ground surface type and season
The threshold value of each wave band simultaneously carries out cloud detection in conjunction with the dark blue-cirrus band index, specifically includes:
For the forest, the meadow and the shrub:
Utilize formulaIt calculates mixed under different-waveband, different latitude and Various Seasonal
Close pixel reflectivity threshold value；Wherein P is the reflectivity threshold value of mixed pixel, ciFor each component eiAccounting, n be it is uncertain because
Son；
Reflectivity will be met greater than the mixed pixel reflectivity threshold value and dark blue-cirrus band index less than 0.95
Pixel is determined as cloud pixel；
For the arable land:
It is lower than the region of preset threshold using visible light and vegetation index removal reflectivity, obtains cloud and bright exposed soil region；
The bright exposed soil region, which is removed, using normalization building index and normalized differential vegetation index obtains cloud sector domain；
The pixel in the cloud sector domain and the dark blue-pixel of the cirrus band index less than 0.95 are determined as cloud pixel.
Invention additionally discloses a kind of satellite image cloud detection systems, comprising:
Satellite image obtains module, for obtaining satellite image to be detected；
Ground surface type division module obtains different earth's surface classes for dividing to the satellite image by ground surface type
The satellite image of type；
Cloud detection module, for being carried out in the satellite image of each ground surface type according to threshold value corresponding to ground surface type
Cloud detection obtains the cloud atlas picture under each ground surface type.
Optionally, the ground surface type division module includes:
Submodule is divided, for the ground surface type to be divided into wetland, water body, people according to history ground surface type data
Make earth's surface, bare area, sea area, arable land, forest, meadow and shrub；The wherein wetland, the water body, the artificial earth's surface, institute
It states bare area and the sea area belongs to constant attribute earth's surface；The arable land, the forest, the meadow and the shrub belong to change
Change attribute earth's surface.
Optionally, the cloud detection module includes:
Constant attribute earth's surface cloud detection submodule chooses dark blue-cirrus wave for being directed to each constant attribute earth's surface
Section index and ground surface type and cloud sector do not spend maximum wave band and carry out cloud detection；Dark blue-cirrus the band index is
Wherein CCI is dark blue-cirrus band index, and Coastal is dark blue wave band reflectivity, and Cirrus is cirrus wave band
Reflectivity；
Change to attributes earth's surface cloud detection submodule, for being directed to each change to attributes earth's surface, according to each ground surface type institute
The latitude at place and season determine the threshold value of each wave band and carry out cloud detection in conjunction with the dark blue-cirrus band index.
Optionally, the constant attribute earth's surface cloud detection submodule includes:
Sea area earth's surface cloud detection unit will meet formula (0.12 ∪ Green > of Blue > for being directed to the sea area
0.08) pixel of 0.06 CCI≤0.95 ∪ 0.06 ∩ NIR > of ∩ Red > is determined as cloud pixel；
Water body earth's surface cloud detection unit will meet 0.13 ∪ Green > of formula Blue > for being directed to the water body
The pixel of 0.08 CCI≤0.95 ∪ 0.10 ∪ Red > is determined as cloud pixel；
Wetland earth's surface cloud detection unit will meet 0.13 ∪ Green > of formula Blue > for being directed to the wetland
The pixel of 0.13 CCI≤0.95 ∪ 0.15 ∪ Red > is determined as cloud pixel；
Bare area earth's surface cloud detection unit will meet 0.95 ∪ Blue > 0.25 of formula CCI < for being directed to the bare area
Pixel be determined as cloud pixel；
Artificial earth's surface cloud detection unit will meet formula ((0.25 ∪ of Blue > for being directed to the artificial earth's surface
Green > 0.2) 298 ∩ NDVI < 0.3 of ∩ BT <) ∩ BT < Tcor_artiThe pixel of ∪ Cirrus > 0.0025 is determined as cloud picture
Member；
Wherein Blue indicates that the reflectivity of blue wave band, Green indicate that the reflectivity of green light band, Red indicate feux rouges wave
The reflectivity of section, NIR indicate that the reflectivity of near infrared band, BT indicate bright temperature value, and NDVI indicates normalized differential vegetation index,
Tcor_artiIndicate that the bright temperature of dynamic corrects threshold value.
Optionally, the change to attributes earth's surface cloud detection submodule includes: green plant earth's surface cloud detection unit and land surface
Cloud detection unit；
The green satellite planted earth's surface cloud detection unit and be used to be directed to the forest, the meadow and the shrub type
Image carries out cloud detection；
The land surface cloud detection unit is used to carry out cloud detection for the satellite image of the farmland types；
The green plant earth's surface cloud detection unit includes:
Mixed pixel reflectivity threshold calculations subelement, for utilizing formulaIt calculates not
Mixed pixel reflectivity threshold value under same wave band, different latitude and Various Seasonal；Wherein P is the reflectivity threshold value of mixed pixel,
ciFor each component eiAccounting, n is uncertain factor；
Green growing area domain cloud pixel determines subelement, for that will meet reflectivity greater than the mixed pixel reflectivity threshold value
It is determined as cloud pixel with dark blue-pixel of the cirrus band index less than 0.95；
The land surface cloud detection unit includes:
Low reflectivity regions remove subelement, for being lower than default threshold using visible light and vegetation index removal reflectivity
The region of value obtains cloud and bright exposed soil region；
Bright exposed soil region removes subelement, for using described in normalization building index and normalized differential vegetation index removal
Bright exposed soil region obtains cloud sector domain；
Arable land region cloud pixel determines subelement, for by the pixel in the cloud sector domain and the land surface type
Dark blue described in satellite image-pixel of the cirrus band index less than 0.95 is determined as cloud pixel.
The specific embodiment provided according to the present invention, the invention discloses following technical effects: disclosed in this invention to defend
Corresponding threshold value is arranged as cloud detection method of optic and system, for different ground surface types in star chart, to make satellite mapping of the invention
As cloud detection method of optic and system consider influence of the ground surface type to threshold value, the standard of the whole cloud detection result of raising in cloud detection
Exactness reduces the difference degree of different zones cloud detection accuracy simultaneously.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be in embodiment
Required attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is only some realities of the invention
Example is applied, it for those of ordinary skill in the art, without any creative labor, can also be according to these
Attached drawing obtains other attached drawings.
Fig. 1 is the method flow diagram of 1 satellite image cloud detection method of optic of the embodiment of the present invention；
Fig. 2 is vegetation CCI value and vegetation+cirrus CCI value comparison diagram；
Fig. 3 is water body CCI value and water body+cirrus CCI value comparison diagram；
Fig. 4 is bare area CCI value and bare area+cirrus CCI value comparison diagram；
Fig. 5 is the cloud accounting statistical result regression analysis figure of cloud detection method of optic and visual interpretation method of the invention；
Fig. 6 indicates the precision figure of six sample areas of each ground surface type；
Fig. 7 is the statistical results chart of each ground surface type totality accuracy of cloud detection method of optic of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts it is all its
His embodiment, shall fall within the protection scope of the present invention.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Embodiment 1:
Satellite image cloud detection method of optic of the invention is for 8 sensor of Landsat based on 30m global seismic type
The threshold method cloud detection method of optic that data (GlobeLand30) are supported, the present invention set different clouds for different ground surface types
Corresponding threshold value is set separately mainly for constant attribute earth's surface and change to attributes earth's surface in detection threshold value.
(1) constant attribute earth's surface threshold value is arranged
Wetland, water body, artificial earth's surface, bare area and sea area in GlobeLand30 ground surface type data is classified as constant category
Property earth's surface because its Reflectivity is relatively stable with the variation of season and longitude and latitude, be not present biggish fluctuation and variation,
Ground Table Properties are constant.For constant attribute earth's surface, it is based on ground-object spectrum feature, it is respectively maximum with cloud distinction to choose each earth's surface
Wave band carry out cloud detection, while also using it is dark blue-cirrus band index progress cirrus detection.
(2) change to attributes earth's surface threshold value is arranged
For arable land, forest, meadow and shrub, Various Seasonal, diverse geographic location will cause vegetation growth shape
The difference of state, this species diversity have and can cause reflectivity changes.In addition to arable land, it is considered to the vegetation of Various Seasonal, different latitude
The influence of mixed pixel, assigns the mixing accounting of different bare areas and vegetation, and then obtains corresponding cloud detection threshold value.And for
Land surface, different crops cause the uncertainty of ground mulching is irregular to follow, and the present invention utilizes reflectivity, normalization
Building index and normalized differential vegetation index are successively excluded, and cloud pixel is finally obtained.
The embodiment of the invention further relates to the modified result step based on ground surface type, is obtained using reflectivity Characteristics
Cloud detection result will appear situations such as scrappy broken spot, the erroneous judgement of highlighted earth's surface and thin cloud are failed to judge, therefore, the present invention first into
Go the removal of scrappy pixel, it is different to accidentally mentioning in the bright temperature difference of the artificial earth's surface overhead based on image statistics cloud and earth's surface later
The highlighted artificial earth's surface of temperature is removed.
Fig. 1 is the method flow diagram of 1 satellite image cloud detection method of optic of the embodiment of the present invention.
Referring to Fig. 1, the satellite image cloud detection method of optic, comprising:
Obtain satellite image to be detected；The satellite image is divided by ground surface type, obtains different earth's surface classes
The satellite image of type.GlobeLand30 is to utilize 30 meters of multispectral image, including US Terrestrial landsat (Landsat)
The image of TM5, ETM+ and Chinese environmental mitigation satellite (HJ-1), and a large amount of auxiliary datas and reference are combined, such as whole world
Ecogeography partition data, global basis geographic information data and global dem data etc., the global seismic covering being made
Categorical data collection product.Wetland, water body, artificial earth's surface, bare area and sea area in GlobeLand30 ground surface type data is returned
For constant attribute earth's surface, arable land, forest, meadow and shrub are classified as change to attributes earth's surface.
Cloud detection is carried out according to threshold value corresponding to ground surface type in the satellite image of each ground surface type, obtains various regions
Cloud atlas picture under table type.It specifically includes:
For each constant attribute earth's surface, chooses dark blue-cirrus band index and ground surface type and cloud sector is not spent most
Big wave band carries out cloud detection；Dark blue-cirrus the band index is
For each change to attributes earth's surface, the threshold of each wave band is determined according to latitude locating for each ground surface type and season
It is worth and carries out cloud detection in conjunction with the dark blue-cirrus band index.
About dark blue-cirrus band index:
The height of general cloud layer is distributed in 0.2km~12km, wherein cirrus is as one of high cloud, the height of cloud base
It generally in 4.5km~10km, is made of the tiny ice crystal in high-altitude, and ice crystal is than sparse, therefore cloud body is relatively thin.
For the land Landsat-8 OLI imager, cirrus band center wavelength is 1.375 μm, which is that steam is strong
Absorption bands.When radiation is across atmosphere, in the wave band, vapour molecule has strong absorption to radiation, until making radiation energy
Change to internal energy of molecular, and then cause the decaying of the wave band intensity of solar radiation, radiation is few even fully can not be through big
Gas.By atmosphere vapour profile it is found that the higher moisture content of height above sea level is lower, intermediate altitude within 3km height moisture content compared with
Height, 3km or more moisture content are very low.And solar radiation reaches ground, atmosphere is needed guiding through, at 1.375 μm, due to steam
Absorption keeps the energy for reaching earth's surface seldom；However cirrus leads to moisture content very little because of its height in the range, because
This is acted on weak by water vapor absorption.To sum up, the thin cirrus in high-altitude can be effectively detected using cirrus wave band.
By Experimental Comparison, calculating is normalized using dark blue wave band and cirrus wave band, difference circle between wave band can be made
Obviousization is limited, to obtain dark blue-cirrus band index CCI, i.e. formula (1).
Wherein, CCI is dark blue-cirrus band index；Coastal be Landsat8 dark blue wave band (0.433 μm-
0.453 μm) reflectivity, Cirrus be cirrus wave band (1.360 μm -1.390 μm) reflectivity.Since Bao Yun is to visible light
The influence of apparent reflectance is smaller, therefore the previous item of formula (dark blue wave band reflectivity) influences less in ratio formula.It plays
What main region was allocated as is cirrus wave band: general atural object is only 10 in the reflectivity quantized level of cirrus wave band-4Left and right, but exist
When cirrus, the reflectivity of wave band can reach 10-3, the reflectivity of water body is to basically reach with decreasing wavelengths in dark blue wave band
Maximum value, and the reflectivity of its all band is very low, will affect calculated result when normalizing and calculating；For other atural objects, in depth
The reflectivity quantized level of blue wave band is also substantially 10-1Left and right, therefore the result has stronger robustness.
Fig. 2 is vegetation CCI value and vegetation+cirrus CCI value comparison diagram.
Fig. 3 is water body CCI value and water body+cirrus CCI value comparison diagram.
Fig. 4 is bare area CCI value and bare area+cirrus CCI value comparison diagram.
Referring to fig. 2~Fig. 4, the results showed that CCI value can more accurately distinguish the ground of clear sky ground and Bao Yun covering
Table, when pixel is clear sky, which is basically stable at 95% or more, and when there are Bao Yunshi, cirrus wave band numerical value expands 10
Times, CCI value overall reduction can be made.Therefore, as CCI≤0.95, which is considered comprising cirrus.
The CCI index being calculated using normalization, calculating parameter can be standardized, and measured, increased for range with 0~1
Both strong contrast, keeps the difference of clear sky and Bao Yun pixel more intuitive.In addition, carrying out cloud detection using CCI value can eliminate
Error caused by the conditions such as most of calibration, angle, landform and atmosphere, enhances the response power to thin cloud.
About constant attribute earth's surface:
The threshold value of constant attribute earth's surface is arranged, and is based on earth's surface spectrum database in conjunction with spectral information, according to different earth's surfaces
The difference of different Spectral Characteristics and cloud Spectral Characteristic selects each earth's surface preferably wave band and corresponding reasonable threshold value is arranged.
1, water body and sea area ground surface type:
It is slightly higher in blue green light since water body reflectivity is integrally relatively low, then gradually decrease, in infrared band close to 0,
Therefore using the cloud pixel in blue, green, red and near infrared band identification sea area overhead, further, since being floated in inland river lake
The influence for swimming plant and silt, can be such that reflectivity is increased in near-infrared, therefore using blue, green and red light on water body
Empty pixel carries out cloud identification.I.e. ground surface type is that the apparent reflectance of water body meets formula (2) then the pixel is identified as cloud picture
Otherwise member is identified as clear sky pixel；Ground surface type is that the apparent reflectance in sea area meets formula (3), then the pixel is differentiated
For cloud pixel, it is otherwise identified as clear sky pixel.
0.08 CCI≤0.95 ∪ (2) 0.13 ∪ Green > of Blue >, 0.10 ∪ Red >
0.06 CCI≤0.95 ∪ (3) (0.12 ∪ Green > 0.08 of Blue >) 0.06 ∩ NIR > of ∩ Red >
2, wetland ground surface type:
For wetland based on aquatile, soil layer is wet, and vegetation type is abundant, and the ecosystem is relatively stable.Due to its back
Scape is moist soil or shoaling layer, therefore reflectivity is lower, and seasonal variations are unobvious for mixed pixel reflectivity changes, institute
Not consider the influence of mixed pixel.When ground surface type is wetland, apparent reflectance meets formula (4), then is identified as
Cloud:
0.13 CCI≤0.95 ∪ (4) 0.13 ∪ Green > of Blue >, 0.15 ∪ Red >
3, exposed soil ground surface type:
For exposed soil region, the present embodiment is with reddish brown sandy loam, palm fibre to dark-brown sand, taupe brown loam and brown
For typical case's exposed soil type such as sandy loam, constantly increased in the reflectivity of visible light wave range, differing texture soil, and 1.0 μm
Belong to high-reflection region between~2.7 μm, it is smaller with the Reflectivity difference of cloud；And dark blue wave band reflectivity is minimum, in cirrus
Wave band reflectivity is high, and cloud then has high reflectance in the two wave bands, and CCI has preferable identification effect for exposed soil and cloud
Fruit, therefore using CCI and the blue wave band for having very big difference with cloud is combined equally, judge whether exposed soil overhead pixel is cloud picture
Member.Specific discrimination formula is formula (5).
0.95 ∪ Blue > 0.25 (5) of CCI <
4, artificial ground surface type:
In visible-range, the SPECTRAL DIVERSITY of artificial earth's surface and cloud is maximum, so selection visible light wave range is to man-made land
Table overhead pixel carry out cloud differentiation, meanwhile, using the low characteristic of the bright temperature of cloud, constructive formula (6) such as meets formula (6), then by
Cloud pixel subject to judgement.Quasi- cloud pixel 0.0025 is being after the amendment of bright temperature, being greater than using cirrus wave band Cirrus reflectivity
Cloud detected unidentified cirrus out.
(298 ∩ NDVI < 0.3 of (0.25 ∪ Green > 0.2 of Blue >) ∩ BT <) ∩ BT < Tcor_arti∪ Cirrus >
0.0025 (6)
About change to attributes earth's surface:
For arable land, forest, meadow and shrub, Various Seasonal, diverse geographic location will cause vegetation growth shape
The difference of state, this species diversity have and can cause reflectivity changes, such as: frigid zone forest is mostly coniferous forest, and temperate forests are mostly to fall leaves
Woods, and the torrid zone is mostly broad-leaf forest and the four seasons are evergreen；Summer vegetation blade is abundant, shows as vegetation characteristics, and winter is existing because falling leaves
As majority shows as exposed soil feature.Therefore when judging cloud pixel using threshold method, it need to consider change in time and space and mixed pixel pair
The influence of reflectivity.The present embodiment considers the vegetation of Various Seasonal, different latitude the influence of mixed pixel, obtains in addition to arable land
Corresponding cloud detection threshold value out.The present embodiment by latitude be divided into tropical (0 °~23.5 °), temperate zone (23.5 °~66.5 °) and
Frigid zone (66.5 °~90 °) three classes.
1, land surface type:
The actual conditions in arable land are complex, and since the planted crop in different plot is different, plant growing cycle is unlike forest
There is apparent globality rule with meadow etc., therefore in waveband selection, first to reflect in visible light wave range totally removal arable land
The lower region of rate (vegetation growing area), extracts cloud and exposed soil, which utilizes blue and green light and red spectral band, reflectivity
Threshold value is respectively 0.2,0.25 and 0.2.By the Spectral Characteristic of exposed soil and cloud it is found that in near-infrared and short-wave infrared reflectivity
Trend is opposite: the reflectivity of exposed soil is lower than short infrared wave band near infrared band, and cloud is then near infrared band than short
Wave infrared band is high, although vegetation near infrared reflectivity as cloud is relatively high, in the test of visible light before
The influence of vegetation is eliminated, therefore can preferably distinguish highlighting in remaining part using normalization building index NDBI
Bare area and cloud, NDBI calculation formula are formula (7), and land surface cloud pixel discrimination formula is formula (8), meet formula (8)
Pixel is judged as cloud pixel.When differentiating, antiradar reflectivity earth surface area is removed first with visible light and vegetation index, is remained
Remaining spissatus and bright exposed soil is influenced with NDBI removal bare area later, while recycling CCI to supplement thin cirrus and identifying.
NDBI=(SWIR1-NIR)/(SWIR1+NIR) (7)
(0.15 ∩ NDVI < 0.3 of (0.2 ∪ Green > of Blue >, 0.25 ∪ Red > 0.2) ∩ NDBI <) ∪ CCI≤
0.95 (8)
SWIR1 is the reflectivity of 6 wave band of Band of the land Landsat-8 OLI imager.SWIR1 corresponds to wave band
1.560μm–1.651μm。
2, forest, meadow and shrub ground surface type:
Different with ploughing, forest, meadow and the shrub growing way in relative Repeat are relatively unified.Although three's reflectivity
Curve is slightly different, but vegetation class has marked difference in the reflectivity of visible light wave range and cloud, therefore selects blue and green light, feux rouges
Three wave bands carry out the cloud identification in meadow and forest overhead；It is carried out on shrub using blue and green light and short-wave infrared (Band 7)
Empty cloud detection.
The reflectance value of mixed pixel is not usually to be made of the reflectivity of certain atural object, but deposit according in the pixel
It is obtained in the percentage of different atural object components, mixed pixel calculation formula is as follows:
Wherein, P is the reflectance value of mixed pixel, ciFor each component eiAccounting, n is uncertain factor；I is some picture
The number of each atural object component in member, N are the sum of atural object component in some pixel.In forest, meadow and shrub threshold calculations
When, the Spectral Characteristic based on exposed soil and vegetation assigns Various Seasonal and the relatively reasonable vegetation of latitude and exposed soil accounting, then
The threshold value of each season, latitude and wave band is calculated according to formula (9).
For forest, evergreen forest, deciduous forest and coniferous forest are regarded as the main life in the torrid zone, temperate zone and frigid zone respectively
Long vegetation, and using above-mentioned three kinds of typical vegetation spectral profiles as reference, the cloud detection threshold value of forest is set.The torrid zone and frigid zone
Due to the factor of locating latitude, long-term temperature is relatively stable, four seasons fuzzy, wherein Tropical forests because of long-term hot humid,
Mostly evergreen forest, plant leaf blade are roomy luxuriant, it is assumed that the exposed soil and vegetation proportion of pixel are respectively 0 and 1；Frigid zone is long-term
Low temperature, forest are distributed coniferous forest more, and the coniferous forest four seasons are evergreen, and blade is needle-shaped, therefore assume that exposed soil is with vegetation proportion
0.2 and 0.8；Temperate zone makes a clear distinction between the four seasons, and forest plant is in the majority in deciduous forest, according to spring (3~May), summer (6~August), autumn
Four season forest defoliations of (9~November) and winter (12~2 months) assign different exposed soils, vegetation respectively in various degree
Mixed proportion obtains the cloud detection threshold value of Various Seasonal.
Since sensibility of the meadow type to latitude is relatively weak, do not classify to the type of grass, sets meadow
It is respectively 0.2,0.23 and 0.3 in the reflectivity threshold value of blue and green light and feux rouges.The meadow of Temperate Region in China is according to spring, summer, autumn and winter
Four seasons assign different grass and exposed soil ratio threshold value calculated；The torrid zone does not consider meadow autumn and winter due to climatic factor
Decay, so setting grass with exposed soil ratio be 1:0；The substantially long-term low temperature in frigid zone, vegetation is sparse, but the spring of summer and temperate zone
Ji Wendu is similar, therefore frigid zone summer threshold value is arranged to, remaining cold time and temperate zone winter threshold consistent with temperate zone spring threshold value
Value is consistent.
Many kinds of in view of shrub, leaf color, the flower color growth cycle of plant are non-constant, and uniformity is poor.
In view of the shrub of some kinds is because of the influence of its blade or flowers and fruits, in reflection to red light rate compared with other types height, therefore use blue
Light, green light and with cloud reflectivity difference relatively large short-wave infrared (SWIR 2) wave band progress shrub overhead cloud detection.
SWIR 2 is the reflectivity of 7 wave band of Band of the land Landsat-8 OLI imager.SWIR2 correspond to wave band be 2.100 μm-
2.300μm.It is respectively 0.16,0.18 and 0.25 that shrub reflectivity threshold value, which is set, in three above wave band.Press spring and summer in Temperate Region in China
Four seasons of autumn and winter assign different threshold values；Torrid areas shrub is dense and green for a long time, therefore sets according to temperate zone summer threshold value；
The warmer area of the characteristic of frigid zone shrub is more cold-resistant, and since frigid zone temperature change amplitude is smaller, the growth of frigid zone shrub
Balanced condition, therefore temperate zone spring threshold value is directly chosen as frigid zone shrub area cloud detection threshold value.
Table 1 lists exposed soil/vegetation mixing ratio that Various Seasonal is arranged in forest, meadow and shrub earth's surface and each
Wave band threshold value.
1 forest of table, meadow and the setting of shrub earth's surface threshold value
In table 1, forest one is classified as main growth vegetation spectrum character curve setting reflectance value, and meadow one is classified as hypothesis
Careless reflectance value, shrub one, which is classified as, assumes Typical Shrub reflectance value, and soil is to assume exposed soil reflectance value, TBlue、TGreen、
TRedAnd TSWIR2The reflectivity threshold of blue and green light, feux rouges and short-wave infrared SWIR 2 after respectively being calculated by mixed pixel
Value, Ns、Nf、NgAnd NbThe respectively pro rate of pixel shared by exposed soil, forest, meadow and shrub.Forest cover is more dense,
And it is that meadow and shrub are presented on the image, vegetation sparse degree difference is very big, or even the state of bare area can be presented, therefore add
The exclusion for adding bare area and cloud to there is the dark blue wave band of bigger difference to carry out bare area when sparse vegetation earth's surface occurs.To sum up, forest,
The discriminate of meadow and shrub is respectively formula (10), formula (11) and formula (12)
Blue > TBlue∪ Green > TGreen∪ Red > TRed∪CCI≤0.95 (10)
((Blue > TBlue∪ Green > TGreen∪ Red > TRed)∩Coastal≥0.25)∪CCI≤0.95 (11)
((Blue > TBlue∪ Green > TGreen∪ SWIR2 > TSWIR2)∩Coastal≥0.25)∪CCI≤0.95 (12)
After the completion of detecting cloud pixel using the above method, need to carry out the knot based on ground surface type to testing result
Fruit amendment.Amendment includes the removal of scrappy pixel and the artificial earth's surface overhead modified result based on bright temperature.
1, it is removed about scrappy pixel:
Due to the possible atural object edge bias of the reflectivity of earth's surface complexity itself and ground surface type library, can generate
Some tiny scrappy wrong identifications cause cloud detection result inaccurate.Therefore, after the completion of above-mentioned cloud detection step, increase
Scrappy pixel removal.Minimizing technology are as follows: all pixels for being identified as cloud of traversal, in the eight neighborhood of the pixel, cloud pixel
When number is less than or equal to 2, then the pixel is determined as scrappy pixel and removed.
2, the artificial earth's surface overhead modified result based on bright temperature:
Highlighted urban area and cloud is difficult to separate on reflectivity, because they have similar spectral properties (high anti-
Penetrate rate), and because of the difference of artificial material, the wave spectrum between each artificial earth's surface of high reflectance can also have different degrees of difference.Cause
This, only relies on and extracts cloud pixel by means of reflectivity information, the artificial earth's surface in part is easily identified as cloud pixel.And temperature information is cloud inspection
The important parameter surveyed.Also in the highlight regions that white is presented on image, the bright Wen Zehui of cloud is lower than artificial earth's surface to be obtained
It is more.But there is also biggish instability for bright temperature, for example in middle latitude or high latitude, earth's surface itself has lower bright
Warm (such as winter), it is difficult to meet the condition of global range with constant bright temperature threshold value.Assuming that the man-made land in certain image range
The bright temperature value of table be it is relatively uniform, on the basis of slightly wide in range artificial earth's surface Cloud Over testing result is arranged in threshold value, utilize
Dynamically bright temperature threshold value therefrom removes misrecognition into the city pixel of cloud.
In the corrigendum of this result, the quasi- cloud detection result in the city overhead completed before is as input data, it is therefore an objective to
The city pixel that misjudgement is cloud is excluded in the result that artificial earth's surface overhead has been detected.This test is only in " complete clear sky city
City " pixel carries out when being greater than 1% with total pixel number ratio, to ensure to have enough clear sky pixels to be used to count.
It is that the pixel for being determined as cloud is removed in all city pixels as complete clear sky pixel.In addition to this, also
The pixel of feux rouges and near infrared reflectivity value less than 0.1 is removed, this is to respectively, remove vegetation and Yun Yin in artificial earth's surface
The influence of shadow causes bright temperature to count relatively low result.The bright temperature value of clear sky pixel by from minimum value to maximum value with 0.1K be step
Long to count the pixel number fallen in each section and calculate frequency, the formula of the cloud pixel of debug identification is as follows:
Wherein, Tcor_artiThreshold value, T are corrected for the bright temperature of dynamicmaxAnd TminBright temperature respectively in the artificial earth's surface pixel of clear sky
Maximum value and minimum value, NtotalFor pixel total number, NjFor the pixel number in j-th of section.
The technical effect of the scheme of this embodiment of the invention is verified and is illustrated below:
It uniformly chooses 40 images in the world to verify the precision of cloud detection of the invention, to each earth's surface
The sample areas that type randomly selects 6 500*500 pixel sizes in corresponding image carries out the regional work of visual interpretation cloud.
For vegetation class earth's surface, the diversity changed when selection for latitude can be shown from figure.Visual interpretation is completed later
Result as true value, be compared with the testing result using cloud detection method of optic of the invention.Using cloud accounting (CP,
CloudProportion), cloud accuracy (CRC, Correct Rate ofCloud), clear sky accuracy (CRS, Correct
Rate ofclear-sky), overall accuracy (TCR, Total CorrectRate) four evaluation indexes carry out quantitative assessment sheet
The precision of inventive method.Specific judgement schematics are formula (14)~formula (17):
Wherein, TNC is total pixel number of cloud, and NT is total pixel number；NC and NS respectively indicates method of the invention
It as a result is the pixel number of cloud and clear sky, NC with visual interpretation (true value) resultVAnd NSVIt respectively indicates in visual interpretation result
The number of cloud pixel and non-cloud pixel.
1, cloud accounting is verified
Fig. 5 is the cloud accounting statistical result regression analysis figure of cloud detection method of optic and visual interpretation method of the invention.By scheming
5 it can be seen that each ground surface type is when compared with true value, and the cloud accounting of detection method of the invention is underestimated now there are most
As it is smaller to underestimate deviation, but general trend is consistent with visual interpretation result, is compared close to reference line, overall root mean square misses
Poor RMSE is 3.95%, and the two possesses higher correlation, reaches degree of precision.
Table 2 is the RMSE calculated result of each ground surface type cloud accounting.
Each ground surface type cloud accounting RMSE result of table 2
Found out by Fig. 5 and table 2, shrub and meadow deviate reference line it is more, it is bright compared with other ground surface types to underestimate phenomenon
Aobvious, the RMSE of the two is respectively 7.71%, 6.77%, shrub deviation it is maximum；Followed by bare area and arable land, RMSE are respectively
5.02% and 4.55%；The smallest ground surface type of error is ocean, RMSE 1.25% in cloud accounting statistics.
The statistics shows various regions table type especially water body class and wetland cloud accounting statistics closest to true value.
2, cloud evaluation number is verified
It chooses tri- indexes of CRC, CRS and TCR and carries out more detailed precision evaluation,
Fig. 6 indicates the precision figure of six sample areas of each ground surface type.
The precision distribution of the overhead cloud detection of each ground surface type can be substantially found out from Fig. 6.
From the distribution of the precision of CRC index, it can be seen that all types of cloud detection accuracy have higher 0.75~1.0
Cloud accuracy of identification, therefore misdetection rate is in the range of 0~0.25.Wherein, water body class cloud pixel identification overall accuracy compared with
Height, misdetection rate is minimum, and wherein the six of ocean sample areas precision is distributed registration height on the diagram, precision is close and difference most
It is small；Secondly, the cloud accuracy of identification of generally wetland is higher, misdetection rate is lower.
Precision distribution for CRS index, clear sky pixel accuracy is relatively high, substantially between 0.95 to 1, equally
The false determination ratio on ground, cloud is very low, between 0 to 0.05.Wherein, it is higher and the most to be shown as whole clear sky pixel accuracy for forest
Stablize (accuracy value of six samples is close)；In addition, meadow, ocean also show the correct precision of relatively high and stable clear sky and
Lower cloud false determination ratio illustrates that the above type also has very strong stability in the case where clear sky accuracy is high.
Six sample areas of comprehensive each ground surface type, obtain the overall accuracy of each index of each earth's surface, cloud pixel is correct
Discrimination is 0.80 or more, wherein the highest ground surface type of cloud pixel discrimination be ocean (0.94), wetland (0.93) and
Water body (0.93)；Clear sky pixel accuracy is relatively high, reaches 0.9 or more, illustrates that this method is very quasi- for the judgement of clear sky
Really, and effect stability.
Fig. 7 is the statistical results chart of each ground surface type totality accuracy of cloud detection method of optic of the invention.
Referring to Fig. 7, the differentiation situation of the statistical result combination cloud pixel and clear sky pixel obtains overall accuracy
(TCR), the overall accuracy of cloud detection method of optic of the invention is between 0.92~1.0, since ocean and forest are more secretly
The spectral information of table and cloud discrimination in visible-range is big, easily reaches degree of precision, therefore the highest two kinds of ground of precision
Table type is ocean and forest, and overall accuracy respectively reaches 0.98% and 0.97%.
Summarize the cloud detection of all ground surface types as a result, for statistical analysis to all pixels, to cloud detection side of the present invention
The overall precision of method is calculated and is evaluated.
The cloud detection method of optic overall accuracy evaluation table of the present invention of table 3
As shown in Table 3, cloud detection method of optic of the invention shows that higher clear sky differentiates accuracy, and overall cloud
Recognition correct rate can reach 0.8903, overall accuracy 0.9685.
To sum up, cloud detection of the invention can reach preferable precision, and the party is simple, be easy it can be readily appreciated that realizing,
And it can totally reach degree of precision, provide the think of for supporting this new using ground surface type prior data bank for cloud detection
Road.
Embodiment 2:
The embodiment of the present invention 2 discloses a kind of satellite image cloud detection system, comprising:
Satellite image obtains module, for obtaining satellite image to be detected；
Ground surface type division module obtains different earth's surface classes for dividing to the satellite image by ground surface type
The satellite image of type；
Cloud detection module, for being carried out in the satellite image of each ground surface type according to threshold value corresponding to ground surface type
Cloud detection obtains the cloud atlas picture under each ground surface type.
Optionally, the ground surface type division module includes:
Submodule is divided, for the ground surface type to be divided into wetland, water body, people according to history ground surface type data
Make earth's surface, bare area, sea area, arable land, forest, meadow and shrub；The wherein wetland, the water body, the artificial earth's surface, institute
It states bare area and the sea area belongs to constant attribute earth's surface；The arable land, the forest, the meadow and the shrub belong to change
Change attribute earth's surface.
Optionally, the cloud detection module includes:
Constant attribute earth's surface cloud detection submodule chooses dark blue-cirrus wave for being directed to each constant attribute earth's surface
Section index and ground surface type and cloud sector do not spend maximum wave band and carry out cloud detection；Dark blue-cirrus the band index is
Wherein CCI is dark blue-cirrus band index, and Coastal is dark blue wave band reflectivity, and Cirrus is cirrus wave band
Reflectivity；
Change to attributes earth's surface cloud detection submodule, for being directed to each change to attributes earth's surface, according to each ground surface type institute
The latitude at place and season determine the threshold value of each wave band and carry out cloud detection in conjunction with the dark blue-cirrus band index.
Optionally, the constant attribute earth's surface cloud detection submodule includes:
Water body earth's surface cloud detection unit will meet formula (0.12 ∪ Green > of Blue > for being directed to the water body
0.08) pixel of 0.06 CCI≤0.95 ∪ 0.06 ∩ NIR > of ∩ Red > is determined as cloud pixel；
Sea area earth's surface cloud detection unit will meet 0.13 ∪ Green > of formula Blue > for being directed to the sea area
The pixel of 0.08 CCI≤0.95 ∪ 0.10 ∪ Red > is determined as cloud pixel；
Wetland earth's surface cloud detection unit will meet 0.13 ∪ Green > of formula Blue > for being directed to the wetland
The pixel of 0.13 CCI≤0.95 ∪ 0.15 ∪ Red > is determined as cloud pixel；
Bare area earth's surface cloud detection unit will meet 0.95 ∪ Blue > 0.25 of formula CCI < for being directed to the bare area
Pixel be determined as cloud pixel；
Artificial earth's surface cloud detection unit will meet formula ((0.25 ∪ of Blue > for being directed to the artificial earth's surface
Green > 0.2) 298 ∩ NDVI < 0.3 of ∩ BT <) ∩ BT < Tcor_artiThe pixel of ∪ Cirrus > 0.0025 is determined as cloud picture
Member；
Wherein Blue indicates that the reflectivity of blue wave band, Green indicate that the reflectivity of green light band, Red indicate feux rouges wave
The reflectivity of section, NIR indicate that the reflectivity of near infrared band, BT indicate bright temperature value, and NDVI indicates normalized differential vegetation index,
Tcor_artiIndicate that the bright temperature of dynamic corrects threshold value.
Optionally, the change to attributes earth's surface cloud detection submodule includes: green plant earth's surface cloud detection unit and land surface
Cloud detection unit；
The green satellite planted earth's surface cloud detection unit and be used to be directed to the forest, the meadow and the shrub type
Image carries out cloud detection；
The land surface cloud detection unit is used to carry out cloud detection for the satellite image of the farmland types；
The green plant earth's surface cloud detection unit includes:
Mixed pixel reflectivity threshold calculations subelement, for utilizing formulaIt calculates not
Mixed pixel reflectivity threshold value under same wave band, different latitude and Various Seasonal；Wherein P is the reflectivity threshold value of mixed pixel,
ciFor each component eiAccounting, n is uncertain factor；
Green growing area domain cloud pixel determines subelement, for that will meet reflectivity greater than the mixed pixel reflectivity threshold value
It is determined as cloud pixel with dark blue-pixel of the cirrus band index less than 0.95；
The land surface cloud detection unit includes:
Low reflectivity regions remove subelement, for being lower than default threshold using visible light and vegetation index removal reflectivity
The region of value obtains cloud and bright exposed soil region；
Bright exposed soil region removes subelement, for using described in normalization building index and normalized differential vegetation index removal
Bright exposed soil region obtains cloud sector domain；
Arable land region cloud pixel determines subelement, for by the pixel in the cloud sector domain and the land surface type
Dark blue described in satellite image-pixel of the cirrus band index less than 0.95 is determined as cloud pixel.
The specific embodiment provided according to the present invention, the invention discloses following technical effects: disclosed in this invention to defend
Corresponding threshold value is arranged as cloud detection method of optic and system, for different ground surface types in star chart, to make satellite mapping of the invention
As cloud detection method of optic and system consider influence of the ground surface type to threshold value, the standard of the whole cloud detection result of raising in cloud detection
Exactness reduces the difference degree of different zones cloud detection accuracy simultaneously.
For the system disclosed in the embodiment, since it is corresponded to the methods disclosed in the examples, so the ratio of description
Relatively simple, reference may be made to the description of the method.
Used herein a specific example illustrates the principle and implementation of the invention, above embodiments
Illustrate to be merely used to help understand method and its core concept of the invention；At the same time, for those skilled in the art,
According to the thought of the present invention, there will be changes in the specific implementation manner and application range.In conclusion this specification
Content should not be construed as limiting the invention.
Claims (10)
1. a kind of satellite image cloud detection method of optic characterized by comprising
Obtain satellite image to be detected；
The satellite image is divided by ground surface type, obtains the satellite image of different ground surface types；
Cloud detection is carried out according to threshold value corresponding to ground surface type in the satellite image of each ground surface type, obtains each ground surface type
Under cloud atlas picture.
2. satellite image cloud detection method of optic according to claim 1, which is characterized in that described to press ground to the satellite image
Table type is divided, and is obtained the satellite image of different ground surface types, is specifically included:
The ground surface type is divided into wetland according to history ground surface type data, water body, artificial earth's surface, bare area, sea area, is ploughed
Ground, forest, meadow and shrub；Wherein the wetland, the water body, the artificial earth's surface, the bare area and the sea area belong to
Constant attribute earth's surface；The arable land, the forest, the meadow and the shrub belong to change to attributes earth's surface.
3. satellite image cloud detection method of optic according to claim 2, which is characterized in that the satellite in each ground surface type
Cloud detection is carried out according to threshold value corresponding to ground surface type in image, the cloud atlas picture under each ground surface type is obtained, specifically includes:
For each constant attribute earth's surface, choose dark blue-cirrus band index and ground surface type do not spent with cloud sector it is maximum
Wave band carries out cloud detection；Dark blue-cirrus the band index is
Wherein CCI is dark blue-cirrus band index, and Coastal is dark blue wave band reflectivity, and Cirrus is the reflection of cirrus wave band
Rate；
For each change to attributes earth's surface, the threshold value and knot of each wave band are determined according to latitude locating for each ground surface type and season
It closes the dark blue-cirrus band index and carries out cloud detection.
4. satellite image cloud detection method of optic according to claim 3, which is characterized in that described to be directed to each constant attribute
Earth's surface, chooses dark blue-cirrus band index and ground surface type and cloud sector does not spend maximum wave band and carries out cloud detection, specific to wrap
It includes:
For the sea area, 0.06 ∪ of formula (0.12 ∪ Green > 0.08 of Blue >) 0.06 ∩ NIR > of ∩ Red > will be met
The pixel of CCI≤0.95 is determined as cloud pixel；
For the water body, the picture of 0.08 CCI≤0.95 ∪ 0.13 ∪ Green > of formula Blue >, 0.10 ∪ Red > will be met
Member is determined as cloud pixel；
For the wetland, the picture of 0.13 CCI≤0.95 ∪ 0.13 ∪ Green > of formula Blue >, 0.15 ∪ Red > will be met
Member is determined as cloud pixel；
For the bare area, the pixel for meeting 0.95 ∪ Blue > 0.25 of formula CCI < is determined as cloud pixel；
For the artificial earth's surface, formula (298 ∩ NDVI < 0.3 of (0.25 ∪ Green > 0.2 of Blue >) ∩ BT <) will be met
∩ BT < Tcor_artiThe pixel of ∪ Cirrus > 0.0025 is determined as cloud pixel；
Wherein Blue indicates that the reflectivity of blue wave band, Green indicate that the reflectivity of green light band, Red indicate red spectral band
Reflectivity, NIR indicate that the reflectivity of near infrared band, BT indicate bright temperature value, and NDVI indicates normalized differential vegetation index, Tcor_artiTable
Show the bright temperature correction threshold value of dynamic.
5. satellite image cloud detection method of optic according to claim 3, which is characterized in that described to be directed to each change to attributes
Earth's surface determines the threshold value of each wave band according to latitude locating for each ground surface type and season and refers in conjunction with the dark blue-cirrus wave band
Number carries out cloud detection, specifically includes:
For the forest, the meadow and the shrub:
Utilize formulaCalculate the mixed pixel under different-waveband, different latitude and Various Seasonal
Reflectivity threshold value；Wherein P is the reflectivity threshold value of mixed pixel, ciFor each component eiAccounting, n is uncertain factor；
It is true less than 0.95 pixel greater than the mixed pixel reflectivity threshold value and dark blue-cirrus band index that reflectivity will be met
It is set to cloud pixel；
For the arable land:
It is lower than the region of preset threshold using visible light and vegetation index removal reflectivity, obtains cloud and bright exposed soil region；
The bright exposed soil region, which is removed, using normalization building index and normalized differential vegetation index obtains cloud sector domain；
The pixel in the cloud sector domain and the dark blue-pixel of the cirrus band index less than 0.95 are determined as cloud pixel.
6. a kind of satellite image cloud detection system characterized by comprising
Satellite image obtains module, for obtaining satellite image to be detected；
Ground surface type division module obtains different ground surface types for dividing to the satellite image by ground surface type
Satellite image；
Cloud detection module, for carrying out cloud inspection according to threshold value corresponding to ground surface type in the satellite image of each ground surface type
It surveys, obtains the cloud atlas picture under each ground surface type.
7. satellite image cloud detection system according to claim 6, which is characterized in that the ground surface type division module packet
It includes:
Submodule is divided, for the ground surface type to be divided into wetland, water body, man-made land according to history ground surface type data
Table, bare area, sea area, arable land, forest, meadow and shrub；The wherein wetland, the water body, the artificial earth's surface, the bare area
Belong to constant attribute earth's surface with the sea area；The arable land, the forest, the meadow and the shrub are with belonging to change to attributes
Table.
8. satellite image cloud detection system according to claim 7, which is characterized in that the cloud detection module includes:
Constant attribute earth's surface cloud detection submodule chooses dark blue-cirrus band index for being directed to each constant attribute earth's surface
And ground surface type and cloud sector do not spend maximum wave band and carry out cloud detection；Dark blue-cirrus the band index is
Wherein CCI is dark blue-cirrus band index, and Coastal is dark blue wave band reflectivity, and Cirrus is the reflection of cirrus wave band
Rate；
Change to attributes earth's surface cloud detection submodule, for being directed to each change to attributes earth's surface, according to locating for each ground surface type
Latitude and season determine the threshold value of each wave band and carry out cloud detection in conjunction with the dark blue-cirrus band index.
9. satellite image cloud detection system according to claim 8, which is characterized in that the constant attribute earth's surface cloud detection
Submodule includes:
Sea area earth's surface cloud detection unit will meet formula (0.12 ∪ Green > 0.08 of Blue >) ∩ for being directed to the sea area
The pixel of 0.06 CCI≤0.95 ∪ 0.06 ∩ NIR > of Red > is determined as cloud pixel；
Water body earth's surface cloud detection unit will meet 0.13 ∪ Green > of formula Blue >, 0.10 ∪ for being directed to the water body
The pixel of 0.08 CCI≤0.95 ∪ Red > is determined as cloud pixel；
Wetland earth's surface cloud detection unit will meet 0.13 ∪ Green > of formula Blue >, 0.15 ∪ for being directed to the wetland
The pixel of 0.13 CCI≤0.95 ∪ Red > is determined as cloud pixel；
Bare area earth's surface cloud detection unit will meet the pixel of 0.95 ∪ Blue > 0.25 of formula CCI < for being directed to the bare area
It is determined as cloud pixel；
Artificial earth's surface cloud detection unit will meet formula ((0.25 ∪ Green > of Blue > for being directed to the artificial earth's surface
0.2) 298 ∩ NDVI < 0.3 of ∩ BT <) ∩ BT < Tcor_artiThe pixel of ∪ Cirrus > 0.0025 is determined as cloud pixel；
Wherein Blue indicates that the reflectivity of blue wave band, Green indicate that the reflectivity of green light band, Red indicate red spectral band
Reflectivity, NIR indicate that the reflectivity of near infrared band, BT indicate bright temperature value, and NDVI indicates normalized differential vegetation index, Tcor_artiTable
Show the bright temperature correction threshold value of dynamic.
10. satellite image cloud detection system according to claim 8, which is characterized in that the change to attributes earth's surface cloud inspection
Surveying submodule includes: green plant earth's surface cloud detection unit and land surface cloud detection unit；
It is described it is green plant earth's surface cloud detection unit be used for for the forest, the meadow and the shrub type satellite image into
It racks detection；
The land surface cloud detection unit is used to carry out cloud detection for the satellite image of the farmland types；
The green plant earth's surface cloud detection unit includes:
Mixed pixel reflectivity threshold calculations subelement, for utilizing formulaCalculate different waves
Mixed pixel reflectivity threshold value under section, different latitude and Various Seasonal；Wherein P is the reflectivity threshold value of mixed pixel, ciFor
Each component eiAccounting, n is uncertain factor；
Green growing area domain cloud pixel determines subelement, for that will meet reflectivity greater than the mixed pixel reflectivity threshold value and depth
Indigo plant-pixel of the cirrus band index less than 0.95 is determined as cloud pixel；
The land surface cloud detection unit includes:
Low reflectivity regions remove subelement, for being lower than the area of preset threshold using visible light and vegetation index removal reflectivity
Domain obtains cloud and bright exposed soil region；
Bright exposed soil region removes subelement, for removing the bright exposed soil using normalization building index and normalized differential vegetation index
Region obtains cloud sector domain；
Arable land region cloud pixel determines subelement, for by the satellite mapping of the pixel in the cloud sector domain and the land surface type
Dark blue-pixel of the cirrus band index less than 0.95 as described in is determined as cloud pixel.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910602938.3A CN110308097A (en) | 2019-07-05 | 2019-07-05 | A kind of satellite image cloud detection method of optic and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910602938.3A CN110308097A (en) | 2019-07-05 | 2019-07-05 | A kind of satellite image cloud detection method of optic and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110308097A true CN110308097A (en) | 2019-10-08 |
Family
ID=68078949
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910602938.3A Withdrawn CN110308097A (en) | 2019-07-05 | 2019-07-05 | A kind of satellite image cloud detection method of optic and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110308097A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110969609A (en) * | 2019-11-29 | 2020-04-07 | 华中科技大学 | Microwave remote sensing bright temperature image cloud detection method |
| CN111007024A (en) * | 2019-12-25 | 2020-04-14 | 武汉大学 | Cloud reflectance rapid determination method suitable for oxygen A band |
| CN111158020A (en) * | 2020-01-06 | 2020-05-15 | 中国科学院微小卫星创新研究院 | Satellite-borne real-time cloud judgment system and method for satellite |
| CN111175231A (en) * | 2019-12-31 | 2020-05-19 | 国家卫星气象中心(国家空间天气监测预警中心) | Inversion method and device of canopy vegetation index and server |
| CN111402182A (en) * | 2020-03-18 | 2020-07-10 | 中国资源卫星应用中心 | Land-coverage-information-based midsplit image synthesis method |
| CN111783555A (en) * | 2020-06-08 | 2020-10-16 | 山东科技大学 | Deep learning satellite data cloud detection algorithm using unified samples |
| CN114049566A (en) * | 2021-11-08 | 2022-02-15 | 北京师范大学 | Method and device for detecting cloud and cloud shadow of land satellite image in step-by-step refinement manner |
| CN117408949A (en) * | 2023-09-20 | 2024-01-16 | 宁波大学 | Cloud and cloud shadow detection method and device for seasonal dynamic threshold |
| WO2024021225A1 (en) * | 2022-07-29 | 2024-02-01 | 知天(珠海横琴)气象科技有限公司 | High-resolution true-color visible light model generation method, high-resolution true-color visible light model inversion method, and system |
-
2019
- 2019-07-05 CN CN201910602938.3A patent/CN110308097A/en not_active Withdrawn
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110969609B (en) * | 2019-11-29 | 2022-04-29 | 华中科技大学 | Microwave remote sensing bright temperature image cloud detection method |
| CN110969609A (en) * | 2019-11-29 | 2020-04-07 | 华中科技大学 | Microwave remote sensing bright temperature image cloud detection method |
| CN111007024A (en) * | 2019-12-25 | 2020-04-14 | 武汉大学 | Cloud reflectance rapid determination method suitable for oxygen A band |
| CN111175231B (en) * | 2019-12-31 | 2022-06-17 | 国家卫星气象中心(国家空间天气监测预警中心) | Inversion method and device of canopy vegetation index and server |
| CN111175231A (en) * | 2019-12-31 | 2020-05-19 | 国家卫星气象中心(国家空间天气监测预警中心) | Inversion method and device of canopy vegetation index and server |
| CN111158020B (en) * | 2020-01-06 | 2022-03-15 | 中国科学院微小卫星创新研究院 | Satellite-borne real-time cloud judgment system and method for satellite |
| CN111158020A (en) * | 2020-01-06 | 2020-05-15 | 中国科学院微小卫星创新研究院 | Satellite-borne real-time cloud judgment system and method for satellite |
| CN111402182A (en) * | 2020-03-18 | 2020-07-10 | 中国资源卫星应用中心 | Land-coverage-information-based midsplit image synthesis method |
| CN111402182B (en) * | 2020-03-18 | 2023-04-28 | 中国资源卫星应用中心 | Middle-divided image synthesis method based on land coverage information |
| CN111783555A (en) * | 2020-06-08 | 2020-10-16 | 山东科技大学 | Deep learning satellite data cloud detection algorithm using unified samples |
| CN114049566A (en) * | 2021-11-08 | 2022-02-15 | 北京师范大学 | Method and device for detecting cloud and cloud shadow of land satellite image in step-by-step refinement manner |
| WO2024021225A1 (en) * | 2022-07-29 | 2024-02-01 | 知天(珠海横琴)气象科技有限公司 | High-resolution true-color visible light model generation method, high-resolution true-color visible light model inversion method, and system |
| CN117408949A (en) * | 2023-09-20 | 2024-01-16 | 宁波大学 | Cloud and cloud shadow detection method and device for seasonal dynamic threshold |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110308097A (en) | A kind of satellite image cloud detection method of optic and system | |
| CN106650689B (en) | A kind of coastal cities time series Land-use extracting method | |
| Lambin et al. | The surface temperature-vegetation index space for land cover and land-cover change analysis | |
| Liu et al. | Generation of new cloud masks from MODIS land surface reflectance products | |
| Schmidt et al. | Spectral discrimination of vegetation types in a coastal wetland | |
| Artigas et al. | Spectral discrimination of marsh vegetation types in the New Jersey Meadowlands, USA | |
| CN108319923A (en) | A kind of cloud Shadow recognition method and system | |
| CN108458978A (en) | Based on the seeds multispectral remote sensing recognition methods that sensitive band and band combination are optimal | |
| Liu et al. | Mapping of boreal vegetation of a temperate mountain in China by multitemporal Landsat TM imagery | |
| Castaldi et al. | Assessing the capability of Sentinel-2 time-series to estimate soil organic carbon and clay content at local scale in croplands | |
| Pang et al. | Identifying spectral features of characteristics of Sphagnum to assess the remote sensing potential of peatlands: a case study in China. | |
| Alavipanah et al. | Criteria of selecting satellite data for studying land resources | |
| Idrees et al. | Urban land use land cover mapping in tropical savannah using Landsat-8 derived normalized difference vegetation index (NDVI) threshold | |
| Sun et al. | A new cloud detection method supported by GlobeLand30 data set | |
| Bolle et al. | Mediterranean landsurface processes assessed from space | |
| Püssa | Forest edges on medium resolution landsat thematic mapper satellite images | |
| Pieri et al. | Assessment of three algorithms for the operational estimation of [CHL] from MODIS data in the Western Mediterranean Sea | |
| Hailegiorgis | Remote Sensing analysis of summer time Evapotranspiration using SEBS algorithm | |
| Ramachandra | Comparative assessment of techniques for bioresource monitoring using GIS and remote sensing | |
| Oliveira et al. | Multispectral indices for wildfire management | |
| Dutkiewicz | Evaluating hyperspectral imagery for mapping the surface symptoms of dryland salinity | |
| Lanne | Monitoring indigenous tropical montane forests in the Taita Hills using airborne digital camera imagery | |
| Nandargi et al. | Temporal and spatial analysis of rainfall and associated Normalized Difference Vegetation Index (NDVI) over the Pune district, India | |
| Kachmar et al. | Improved forest cover classification in an industrialized mountain area in Japan | |
| Treat et al. | Classification of a high latitude bog using multispectral drone 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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20191008 |
|
| WW01 | Invention patent application withdrawn after publication |