CN107274361A - Landsat TM remote sensing image datas remove cloud method and system - Google Patents
Landsat TM remote sensing image datas remove cloud method and system Download PDFInfo
- Publication number
- CN107274361A CN107274361A CN201710384330.9A CN201710384330A CN107274361A CN 107274361 A CN107274361 A CN 107274361A CN 201710384330 A CN201710384330 A CN 201710384330A CN 107274361 A CN107274361 A CN 107274361A
- Authority
- CN
- China
- Prior art keywords
- cloud
- msub
- landsat
- thickc
- mrow
- 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
- 241001269238 Data Species 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000001228 spectrum Methods 0.000 claims abstract description 21
- 238000009826 distribution Methods 0.000 claims abstract description 10
- 230000009977 dual effect Effects 0.000 claims abstract description 9
- 230000003595 spectral effect Effects 0.000 claims description 62
- 230000011218 segmentation Effects 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 description 6
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 5
- 238000001429 visible spectrum Methods 0.000 description 4
- 241000406668 Loxodonta cyclotis Species 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 239000003595 mist Substances 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 230000002123 temporal effect Effects 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000002329 infrared spectrum Methods 0.000 description 2
- 241000894007 species Species 0.000 description 2
- 235000007926 Craterellus fallax Nutrition 0.000 description 1
- 240000007175 Datura inoxia Species 0.000 description 1
- 206010012373 Depressed level of consciousness Diseases 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
Classifications
-
- G06T5/70—
-
- G06T5/77—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
Abstract
Cloud method is removed the present invention relates to a kind of Landsat TM remote sensing image datas, including：Input Landsat TM remote sensing image datas；Multi-scale division is carried out to above-mentioned Landsat TM remote sensing image datas；To the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to spissatus spectrum characteristic, operating characteristic " ThickC " is set up；The object that the operating characteristic " ThickC " of threshold condition will be met is classified to " Thick Cloud " by " unclassified "；To the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to the spectrum characteristic and characteristic distributions of thin cloud, opening relationships feature " ThinC "；In remaining " unclassified " object, the object " unclassified " for the condition that meets is classified to " Thin Cloud " using dual threshold classification；By it is above-mentioned be classified to " Thick Cloud " and " Thin Cloud " object homogeneous classification to " Cloud ", complete data remove cloud.Cloud system is removed the invention further relates to a kind of Landsat TM remote sensing image datas.The present invention can efficiently reduce or remove the influence of cloud, reduce wrong point, leak sub-category amount, improve classification effectiveness and precision.
Description
Technical field
Cloud method and system are removed the present invention relates to a kind of Landsat TM remote sensing image datas.
Background technology
It is a kind of phenomenon very common during Remote Sensing Data Processing that obnubilation, which covers, is especially utilizing remotely-sensed data to carry out soil
When cover monitoring, renewal, due to blocking for cloud layer, the efficiency that data are utilized is had a strong impact on.On the other hand, cloud and mist is South China
Target in air at most, least stable, understands at State Satellite Meterological Center at retrieval, due to the influence of cloud and mist, obtains
South China region NOAA/AVHRR remotely-sensed datas average effectiveness level less than 7%, in any time of sensors for data
The situation that cloud and mist is blocked all is likely encountered, because the presence of cloud, is needed when carrying out using satellite remote sensing images and drawing etc. and apply
Consume the substantial amounts of energy of related work person and remove cloud.
Due to weather, it is difficult to obtain cloudless remote sensing image completely, most of remote sensing image when obtaining meeting or
More or less influenceed by cloud, this is brought to being changed the research work such as monitoring, land cover classification using remote sensing image
Huge trouble, therefore except cloud also becomes the big problem that numerous Landsat TM apply worker to face.
Numerous except in cloud, if data have large-scale thin cloud, using homomorphic filtering method preferably, this is
Because homomorphic filtering method combines frequency filter with grey scale change, separation cloud and background atural object are finally removed from image
The influence of cloud, but this method is due to being related to wave filter and the selection by frequency, during filtering sometimes
Some useful informations are lost, and for this computationally intensive remotely-sensed datas of Landsat TM, are operated very inconvenient, and
And for there is spissatus image cannot be in this way.When having large stretch of spissatus in the data of processing, conventional hand
Section is to use time averaging method, but this algorithm is only applicable to characters of ground object and changes over time less region, for
The remote sensing image of this middle temporal resolutions of Landsat TM is tended not to this simple alternate algorithm.
In summary, in the case of limited spectral resolution ratio, cloud covering noise is difficult to be gone with multispectral method
Remove, for the low remote sensing platform of temporal resolution, except cloud is one of key factor for causing remotely-sensed data utilization rate low, and
And have a strong impact on subsequently using for remotely-sensed data, such as image recognition, variation monitoring, supervised classification etc. application problem.
The content of the invention
In view of this, it is necessary to provide a kind of Landsat TM remote sensing image datas except cloud method and system.
The present invention provides a kind of Landsat TM remote sensing image datas and removes cloud method, and this method comprises the following steps：A. it is defeated
Enter Landsat TM remote sensing image datas；B. using Object--oriented method is based on, to above-mentioned Landsat TM remote sensing image numbers
According to progress multi-scale division；C. to the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to spissatus spectrum
Feature, sets up operating characteristic " ThickC "；D. threshold classification method is utilized, the operating characteristic " ThickC " of threshold condition will be met
Object is classified to " Thick Cloud " by " unclassified "；E. to the Landsat TM remote sensing after above-mentioned multi-scale division
Image data, according to the spectrum characteristic and characteristic distributions of thin cloud, opening relationships feature " ThinC "；F. remaining
In " unclassified " object, the object " unclassified " for the condition that meets is classified to using dual threshold classification
“Thin Cloud”；G. " Thick Cloud " and " Thin Cloud " object homogeneous classification is extremely are classified to by above-mentioned
" Cloud ", completes data and removes cloud.
Wherein, the Landsat TM remote sensing image datas include 7 spectral coverages：B1Bluish-green spectral coverage, B2Green spectral coverage, B3Red spectrum
Section, B4Near-infrared spectral coverage, B5Nearly short-wave infrared spectral coverage, B6Thermal infrared spectral coverage, B7Nearly short-wave infrared spectral coverage.
Described step b is specifically included：According to the Landsat TM remote sensing image datas of input, imaged object layer is set up；
Selection participates in the weight for each wave band data that multi-scale division is calculated；The scale parameter of multi-scale division is set；Many chis are set
Spend the form parameter of segmentation；Degree of the compacting parameter of multi-scale division is set；The knot of multi-scale division is calculated on imaged object layer
Really, it is " unclassified " object produced after segmentation to be assigned into class.
Described step c is specifically included：For each object ω={ P1,P2,P3,…,Pn, P is included by the object
Pixel, the pixel quantity that n is included for the object, the specific formula for calculation of its operating characteristic " ThickC " is as follows：Wherein, Wherein, ThickC represents the operating characteristic of the object
The numerical value of " ThickC "；
N represents the pixel quantity that the object is included；Represent the B of a pixel2Brightness value；Represent a pixel
B4Brightness value；Represent the B of a pixel7Brightness value；C represents a scaling constant, is set according to user's request；To shadow
As each object { ω in object layer1,ω2,ω3,…,ωn{ operating characteristic " ThickC " is all calculated respectively.
Described step e is specifically included：" ThinC " if represent centered on an object object, Edge Distance this
The beeline at center object edge will be used as the calculating center object relationship characteristic no more than other objects of d pixel
The element of " ThinC " value, it is described as follows：
Wherein, d represents pre-determined distance value；d(ωi,ωj) represent object ωiWith object ωjThe distance between；Table
Show the set for all objects for participating in calculated relationship feature " ThinC "；{ω∣d(ωi,ωj)≤d, i=1,2,3 ..., n, j=
1,2,3 ..., n } represent to meet Rule of judgment d (ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,3 ..., n all objects
Set；
According to operating characteristic and its statistic algorithm calculated relationship feature " ThinC "： Wherein,
ThinC represents the relationship characteristic value of the object；ThickC represents to participate in the operating characteristic of the object of calculated relationship feature " ThinC "
" ThickC " value；N represents to participate in the quantity of the object of calculated relationship feature " ThinC ".
The present invention provides a kind of Landsat TM remote sensing image datas and removes cloud system, and the system, which includes the system, includes input
Module, segmentation module, operating characteristic set up module, sort module, relationship characteristic and set up module, wherein：The input module is used
In input Landsat TM remote sensing image datas；The segmentation module is used for using Object--oriented method is based on, to above-mentioned
Landsat TM remote sensing image datas carry out multi-scale division；The operating characteristic, which sets up module, to be used for above-mentioned multi-scale division
Landsat TM remote sensing image datas afterwards, according to spissatus spectrum characteristic, set up operating characteristic " ThickC "；The classification mould
Block is used to utilize threshold classification method, will meet the object of operating characteristic " ThickC " of threshold condition by " unclassified " point
Class is to " Thick Cloud "；The relationship characteristic, which sets up module, to be used for the Landsat TM remote sensing after above-mentioned multi-scale division
Image data, according to the spectrum characteristic and characteristic distributions of thin cloud, opening relationships feature " ThinC "；The sort module is additionally operable to
In remaining " unclassified " object, the object " unclassified " for the condition that meets is divided using dual threshold classification
Class is to " Thin Cloud "；The sort module is additionally operable to be classified to " Thick Cloud " and " Thin Cloud's " by above-mentioned
Object homogeneous classification completes data and removes cloud to " Cloud ".
Wherein, described Landsat TM remote sensing image datas include 7 spectral coverages：B1Bluish-green spectral coverage, B2Green spectral coverage, B3It is red
Spectral coverage, B4Near-infrared spectral coverage, B5Nearly short-wave infrared spectral coverage, B6Thermal infrared spectral coverage, B7Nearly short-wave infrared spectral coverage.
Described segmentation module specifically for：According to the Landsat TM remote sensing image datas of input, imaged object is set up
Layer；Selection participates in the weight for each wave band data that multi-scale division is calculated；The scale parameter of multi-scale division is set；Set many
The form parameter of multi-scale segmentation；Degree of the compacting parameter of multi-scale division is set；Multi-scale division is calculated on imaged object layer
As a result, it is " unclassified " object produced after segmentation to be assigned into class.
Described operating characteristic set up module specifically for：
For each object ω={ P1,P2,P3,…,Pn, the pixel that P is included for the object, n is wrapped by the object
The pixel quantity contained, the specific formula for calculation of its operating characteristic " ThickC " is as follows：
Wherein,
Wherein, ThickC represents the numerical value of the operating characteristic " ThickC " of the object；N represents the pixel that the object is included
Quantity；Represent the B of a pixel2Brightness value；Represent the B of a pixel4Brightness value；Represent the B of a pixel7
Brightness value；C represents a scaling constant, is set according to user's request；
To each object { ω in imaged object layer1,ω2,ω3,…,ωnOperating characteristic is all calculated respectively
“ThickC”。
Described relationship characteristic set up module specifically for：
" ThinC " is if represent object, the most short distance at this center object edge of Edge Distance centered on an object
Will be as the element for calculating center object relationship characteristic " ThinC " value from other objects no more than d pixel, it is described such as
Under：
Wherein, d represents pre-determined distance value；d(ωi,ωj) represent object ωiWith object ωjThe distance between；Table
Show the set for all objects for participating in calculated relationship feature " ThinC "；{ω∣d(ωi,ωj)≤d, i=1,2,3 ..., n, j=
1,2,3 ..., n } represent to meet Rule of judgment d (ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,3 ..., n all objects
Set；
According to operating characteristic and its statistic algorithm calculated relationship feature " ThinC "：
Wherein, ThinC represents the relationship characteristic value of the object；ThickC represents to participate in calculated relationship feature " ThinC "
Operating characteristic " ThickC " value of object；N represents to participate in the quantity of the object of calculated relationship feature " ThinC ".
Spissatus identification technology and Bao Yun identification technologies have been effectively incorporated into one by the present invention according to its NATURAL DISTRIBUTION rule
Rise, it is relatively low for temporal resolution, it is difficult to which that for the Landsat TM data of time averaging method, the principle of the invention is very simple
Single, relatively low to the quality requirement of data, seldom, computational efficiency is higher, strong robustness for operational parameter, obtained cloud recognition result compared with
To be reliable, the influence of cloud can be efficiently reduced or removed, classification difficulty is alleviated for classification worker, reduce wrong point, leakage
Sub-category amount, improves classification effectiveness and precision.
Brief description of the drawings
Fig. 1 is the flow chart that Landsat TM remote sensing image datas of the present invention remove cloud method；
Fig. 2 is the hardware architecture diagram that Landsat TM remote sensing image datas of the present invention remove cloud system.
Embodiment
Below in conjunction with the accompanying drawings and specific embodiment the present invention is further detailed explanation.
As shown in fig.1, being the job stream that Landsat TM remote sensing image datas of the present invention remove cloud method preferred embodiment
Cheng Tu.
Step S1, inputs Landsat TM remote sensing image datas.Wherein：
The Landsat TM remote sensing image datas include 7 spectral coverages, are respectively：3 visible spectrums, 1 near-infrared
Spectral coverage, 2 nearly short-wave infrared spectral coverages, 1 thermal infrared spectral coverage, 3 visible spectrums include：It is bluish-green spectral coverage, green spectral coverage, red
Spectral coverage；And be named as 7 spectral coverages successively：B1(bluish-green spectral coverage), B2(green spectral coverage), B3(red spectral coverage), B4(near-infrared spectra
Section), B5(nearly short-wave infrared spectral coverage), B6(thermal infrared spectral coverage), B7(nearly short-wave infrared spectral coverage).
Step S2, using based on Object--oriented method, is carried out multiple dimensioned to above-mentioned Landsat TM remote sensing image datas
Segmentation.Specifically：
Participating in the wave band of multi-scale division includes B1、B2、B3、B4、B5、B7, in the present embodiment, segmentation yardstick is set to
50, form factor and degree of the compacting factor are voluntarily adjusted according to user's request, are by the object produced after segmentation tax class
“unclassified”。
Further, the flow of multi-scale division is as follows：
(1) according to the Landsat TM remote sensing image datas of input, imaged object layer is set up；
(2) selection participates in the weight for each wave band data that multi-scale division is calculated, and the wave band of segmentation is participated in herein and is included
B1、B2、B3、B4、B5、B7, weight is disposed as 1 in the present embodiment, can also be changed according to user's request；
(3) scale parameter of multi-scale division is set, in the present embodiment according to Landsat TM remote sensing image datas
Classification characteristicses are set to 50, can also be changed according to user's request；
(4) form parameter of multi-scale division is set, in the present embodiment according to Landsat TM remote sensing image datas
Classification characteristicses are set to 0.1, can also be changed according to user's request；
(5) degree of the compacting parameter of multi-scale division is set, in the present embodiment according to Landsat TM remote sensing image datas
Classification characteristicses be set to 0.3, can also be changed according to user's request；
(6) result of multi-scale division is calculated on imaged object layer, is by the object produced after segmentation tax class
“unclassified”。
Step S3, to the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to spissatus (Thick
Cloud spectrum characteristic), sets up operating characteristic " ThickC ".Namely：
It is according to spectrum characteristic spissatus in the Landsat TM remote sensing image datas, i.e., spissatus in wave band B2、B4、B7On
Will be much higher than other ground class with higher brightness value, and often, in order to stretch this species diversity, the present embodiment builds one
Plant the thicker cloud of new operating characteristic " ThickC " automatic distinguishing and other atural objects.Specifically include：
(1) for each object ω={ P1,P2,P3,…,Pn, the pixel that P is included for the object, n is the object
Comprising pixel quantity, the specific formula for calculation of its operating characteristic " ThickC " is as follows：
Wherein,
Wherein, ThickC represents the numerical value of the operating characteristic " ThickC " of the object；N represents the pixel that the object is included
Quantity；Represent the B of a pixel2Brightness value；Represent the B of a pixel4Brightness value；Represent the B of a pixel7
Brightness value；C represents a scaling constant, is set according to user's request.
(2) to each object { ω in imaged object layer1,ω2,ω3,…,ωnOperating characteristic is all calculated respectively
“ThickC”。
Step S4, using threshold classification method, will meet threshold condition operating characteristic " ThickC " object by
" unclassified " is classified to " Thick Cloud ".Namely：
The object of certain operating characteristic " ThickC " threshold k will be met by " unclassified " point using threshold classification method
Class is to " Thick Cloud ", specific formula is as follows：
Wherein, K is the threshold value with reference to the setting of the factors such as the quality of image by user；Unclassified (ThickC) represents to divide
Generic attribute is the ThickC value sets of Unclassified all objects；{ ThickC ∣ ThickC >=K } represents to meet ThickC
The set of all objects of >=K conditions；Thick Cloud (ThickC) presentation class attribute is all right of Thick Cloud
The ThickC value sets of elephant.
Step S5, to the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to thin cloud (Thin
Cloud spectrum characteristic) and characteristic distributions, opening relationships feature " ThinC ".Comprise the following steps that：
Except being identified according to spectrum characteristic spissatus in the Landsat TM remote sensing image datas containing spissatus pair
As outer, contain substantial amounts of thin cloud toward contact in the Landsat TM remote sensing image datas, these thin clouds are often all distributed in thickness
The edge zone of cloud is distributed in relatively spissatus region, can using the special distribution character of cloud layer in remotely-sensed data
To construct Bao Yun recognition methods.Based on this principle, the present embodiment is except utilizing the Landsat TM remote sensing image datas
In thin cloud spectrum characteristic outside, also use related in imaged object layer spatial location of thin cloud object and spissatus object and close
System, constructs a kind of new cloud discrimination index and comes the relatively thin cloud of automatic distinguishing and other atural objects, the discrimination index is exactly that relation is special
" ThinC " is levied, it is as follows that it implements step：
(1) with reference to factors such as the qualities of image, setting meets the pre-determined distance d values of user's accuracy requirement as far as possible.Pre-determined distance d
It is a relative quantity, its value is weighed with the quantity of pixel in imaged object layer, and calculated relationship feature " ThinC " is participated in for limiting
Object condition, " ThinC " if represent centered on an object object, this center object edge of Edge Distance is most
Short distance will be as the element for calculating center object relationship characteristic " ThinC " value no more than other objects of d pixel, and it is retouched
State as follows：
Wherein, d represents pre-determined distance value；d(ωi,ωj) represent object ωiWith object ωjThe distance between；Table
Show the set for all objects for participating in calculated relationship feature " ThinC "；{ω∣d(ωi,ωj)≤d, i=1,2,3 ..., n, j=
1,2,3 ..., n } represent to meet Rule of judgment d (ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,3 ..., n all objects
Set.
In simple terms, pre-determined distance d values be to provide center object and around it participate in calculated relationship characteristic value its
The syntople of its object, this syntople is that the object of sensu lato syntople, i.e., two can be with imaged object layer
Contact, can not also be contacted.D values are small, then the number of objects for participating in calculated relationship characteristic value is just few；D values are big, then participate in calculating and close
Be characteristic value number of objects it is just many, the sizes of d values depending on thin cloud and it is spissatus between position relationship, by user's foundation precision
Demand is set.
(2) according to operating characteristic and its statistic algorithm calculated relationship feature " ThinC ".When participation calculated relationship feature
After the object of " ThinC " is all determined, it is possible to the calculating of relationship characteristic " ThinC " is carried out according to formula below：
Wherein, ThinC represents the relationship characteristic value of the object；ThickC represents to participate in calculated relationship feature " ThinC "
Operating characteristic " ThickC " value of object；N represents to participate in the quantity of the object of calculated relationship feature " ThinC ".
Step S6, in remaining " unclassified " object, using dual threshold classification by the object for the condition that meets
" unclassified " is classified to that " Thin Cloud ", 2 characteristic of divisions of use are respectively operating characteristic " ThickC " and pass
It is feature " ThinC ".Specifically：
The object " unclassified " for the condition that meets is classified to " Thin Cloud ", use using dual threshold classification
2 characteristic of divisions be respectively operating characteristic " ThickC " and relationship characteristic " ThinC ", specific formula is as follows：
Wherein, K1For by threshold value of the user with reference to the setting of the factors such as the quality of image；K2For by user with reference to quality of image etc. because
The threshold value of element setting；Unclassified (ThickC, ThinC) presentation class attribute is Unclassified all objects
(ThickC, ThinC) value set；{ ThickC, ThinC ∣ ThickC >=K1, and ThinC >=K2Represent to meet ThickC >=K1,
And ThinC >=K2The set of all objects of condition；Thin Cloud (ThickC, ThinC) presentation class attribute is Thin
(ThickC, ThinC) value set of Cloud all objects.
Step S7, " Thick Cloud " and " Thin Cloud " object homogeneous classification is extremely are classified to by above-mentioned
" Cloud ", completes data and removes cloud.Specifically：
It is " Thick Cloud " and " Thin Cloud " object is unified to assign classification " Cloud ", so far by category attribute
Just the cloud identifying processing to the Landsat TM remote sensing image datas is completed, the arrow with cloud category attribute has been finally given
Measure data.
As shown in fig.2, being the hardware architecture diagram that Landsat TM remote sensing image datas of the present invention remove cloud system 10.This is
System includes：Input module 101, segmentation module 102, operating characteristic set up module 103, sort module 104, relationship characteristic and set up mould
Block 105.
The input module 101 is used to input Landsat TM remote sensing image datas.Wherein：
The Landsat TM remote sensing image datas include 7 spectral coverages, are respectively：3 visible spectrums, 1 near-infrared
Spectral coverage, 2 nearly short-wave infrared spectral coverages, 1 thermal infrared spectral coverage, 3 visible spectrums include：It is bluish-green spectral coverage, green spectral coverage, red
Spectral coverage；And be named as 7 spectral coverages successively：B1(bluish-green spectral coverage), B2(green spectral coverage), B3(red spectral coverage), B4(near-infrared spectra
Section), B5(nearly short-wave infrared spectral coverage), B6(thermal infrared spectral coverage), B7(nearly short-wave infrared spectral coverage).
The segmentation module 102 is used for using Object--oriented method is based on, to above-mentioned Landsat TM remote sensing image numbers
According to progress multi-scale division.Specifically：
Participating in the wave band of multi-scale division includes B1、B2、B3、B4、B5、B7, in the present embodiment, segmentation yardstick is set to
50, form factor and degree of the compacting factor are voluntarily adjusted according to user's request, are by the object produced after segmentation tax class
“unclassified”。
Further, it is described segmentation module 102 specifically for：
(1) according to the Landsat TM remote sensing image datas of input, imaged object layer is set up；
(2) selection participates in the weight for each wave band data that multi-scale division is calculated, and the wave band of segmentation is participated in herein and is included
B1、B2、B3、B4、B5、B7, weight is disposed as 1 in the present embodiment, can also be changed according to user's request；
(3) scale parameter of multi-scale division is set, in the present embodiment according to Landsat TM remote sensing image datas
Classification characteristicses are set to 50, can also be changed according to user's request；
(4) form parameter of multi-scale division is set, in the present embodiment according to Landsat TM remote sensing image datas
Classification characteristicses are set to 0.1, can also be changed according to user's request；
(5) degree of the compacting parameter of multi-scale division is set, in the present embodiment according to Landsat TM remote sensing image datas
Classification characteristicses be set to 0.3, can also be changed according to user's request；
(6) result of multi-scale division is calculated on imaged object layer, is by the object produced after segmentation tax class
“unclassified”。
The operating characteristic, which sets up module 103, to be used for the Landsat TM remote sensing image numbers after above-mentioned multi-scale division
According to according to the spectrum characteristic of spissatus (Thick Cloud), setting up operating characteristic " ThickC ".Namely：
The operating characteristic sets up module 103 according to Spectral Properties spissatus in the Landsat TM remote sensing image datas
Point, i.e., it is spissatus in wave band B2、B4、B7It is upper that there is higher brightness value, and often will be much higher than other ground class, in order to draw
This species diversity is stretched, the present embodiment builds a kind of thicker cloud of new operating characteristic " ThickC " automatic distinguishing and other atural objects.Tool
Body includes：
(1) for each object ω={ P1,P2,P3,…,Pn, the pixel that P is included for the object, n is the object
Comprising pixel quantity, the specific formula for calculation of its operating characteristic " ThickC " is as follows：
Wherein,
Wherein, ThickC represents the numerical value of the operating characteristic " ThickC " of the object；N represents the pixel that the object is included
Quantity；Represent the B of a pixel2Brightness value；Represent the B of a pixel4Brightness value；Represent the B of a pixel7
Brightness value；C represents a scaling constant, is set according to user's request.
(2) to each object { ω in imaged object layer1,ω2,ω3,…,ωnOperating characteristic is all calculated respectively
“ThickC”。
The sort module 104 is used to utilize threshold classification method, will meet the operating characteristic " ThickC " of threshold condition
Object is classified to " Thick Cloud " by " unclassified ".Namely：
The object of certain operating characteristic " ThickC " threshold k will be met by " unclassified " point using threshold classification method
Class is to " Thick Cloud ", specific formula is as follows：
Wherein, K is the threshold value with reference to the setting of the factors such as the quality of image by user；Unclassified (ThickC) represents to divide
Generic attribute is the ThickC value sets of Unclassified all objects；{ ThickC ∣ ThickC >=K } represents to meet ThickC
The set of all objects of >=K conditions；Thick Cloud (ThickC) presentation class attribute is all right of Thick Cloud
The ThickC value sets of elephant.
The relationship characteristic, which sets up module 105, to be used for the Landsat TM remote sensing image numbers after above-mentioned multi-scale division
According to according to the spectrum characteristic and characteristic distributions of thin cloud (Thin Cloud), opening relationships feature " ThinC ".It is specific as follows：
Except being identified according to spectrum characteristic spissatus in the Landsat TM remote sensing image datas containing spissatus pair
As outer, contain substantial amounts of thin cloud toward contact in the Landsat TM remote sensing image datas, these thin clouds are often all distributed in thickness
The edge zone of cloud is distributed in relatively spissatus region, can using the special distribution character of cloud layer in remotely-sensed data
To construct Bao Yun recognition methods.Based on this principle, the present embodiment is except utilizing the Landsat TM remote sensing image datas
In thin cloud spectrum characteristic outside, also use related in imaged object layer spatial location of thin cloud object and spissatus object and close
System, constructs a kind of new cloud discrimination index and comes the relatively thin cloud of automatic distinguishing and other atural objects, the discrimination index is exactly that relation is special
" ThinC " is levied, its is specific as follows：
(1) with reference to factors such as the qualities of image, setting meets the pre-determined distance d values of user's accuracy requirement as far as possible.Pre-determined distance d
It is a relative quantity, its value is weighed with the quantity of pixel in imaged object layer, and calculated relationship feature " ThinC " is participated in for limiting
Object condition, " ThinC " if represent centered on an object object, this center object edge of Edge Distance is most
Short distance will be as the element for calculating center object relationship characteristic " ThinC " value no more than other objects of d pixel, and it is retouched
State as follows：
Wherein, d represents pre-determined distance value；d(ωi,ωj) represent object ωiWith object ωjThe distance between；Table
Show the set for all objects for participating in calculated relationship feature " ThinC "；{ω∣d(ωi,ωj)≤d, i=1,2,3 ..., n, j=
1,2,3 ..., n } represent to meet Rule of judgment d (ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,3 ..., n all objects
Set.
In simple terms, pre-determined distance d values be to provide center object and around it participate in calculated relationship characteristic value its
The syntople of its object, this syntople is that the object of sensu lato syntople, i.e., two can be with imaged object layer
Contact, can not also be contacted.D values are small, then the number of objects for participating in calculated relationship characteristic value is just few；D values are big, then participate in calculating and close
Be characteristic value number of objects it is just many, the sizes of d values depending on thin cloud and it is spissatus between position relationship, by user's foundation precision
Demand is set.
(2) according to operating characteristic and its statistic algorithm calculated relationship feature " ThinC ".When participation calculated relationship feature
After the object of " ThinC " is all determined, it is possible to the calculating of relationship characteristic " ThinC " is carried out according to formula below：
Wherein, ThinC represents the relationship characteristic value of the object；ThickC represents to participate in calculated relationship feature " ThinC "
Operating characteristic " ThickC " value of object；N represents to participate in the quantity of the object of calculated relationship feature " ThinC ".
The sort module 104 is additionally operable in remaining " unclassified " object, will using dual threshold classification
The object " unclassified " for meeting condition is classified to that " Thin Cloud ", 2 characteristic of divisions of use are respectively that computing is special
Levy " ThickC " and relationship characteristic " ThinC ".Specifically：
The object " unclassified " for the condition that meets is classified to " Thin Cloud ", use using dual threshold classification
2 characteristic of divisions be respectively operating characteristic " ThickC " and relationship characteristic " ThinC ", specific formula is as follows：
Wherein, K1For by threshold value of the user with reference to the setting of the factors such as the quality of image；K2For by user with reference to quality of image etc. because
The threshold value of element setting；Unclassified (ThickC, ThinC) presentation class attribute is Unclassified all objects
(ThickC, ThinC) value set；{ ThickC, ThinC ∣ ThickC >=K1, and ThinC >=K1Represent to meet ThickC >=K1,
And ThinC >=K2The set of all objects of condition；ThinCloud (ThickC, ThinC) presentation class attribute is Thin
(ThickC, ThinC) value set of Cloud all objects.
The sort module 104 is additionally operable to be classified to " Thick Cloud " and " Thin Cloud " object system by above-mentioned
One is classified to " Cloud ", completes data and removes cloud.Specifically：
It is " Thick Cloud " and " Thin Cloud " object is unified to assign classification " Cloud ", so far by category attribute
Just the cloud identifying processing to the Landsat TM remote sensing image datas is completed, the arrow with cloud category attribute has been finally given
Measure data.
It should be noted that the present invention be directed to the design of Landsat TM remote sensing image datas, and obtained more
Preferable experimental result, in theory for, band class information and the Landsat TM of the remote sensing image data simply entered more connect
Closely, it can be carried out using similar method or system except cloud processing, all within the scope of the present invention.
An important image classification method used in the present invention is exactly the sorting technique of facing elephant.Object-oriented
Minimum unit handled by sorting technique is the object of multiple pixels composition containing more semantic informations, and is no longer towards picture
Single pixel handled by the sorting technique of member.For object is compared to pixel, not only contain spectral information, also contain such as
The semantic information of the more horn of plenty such as geological information, texture information, topology information, relative position information.The classification side of object-oriented
Method has two important advantages：1) object-based classification processing can respectively be carried out in different scale layers, so that effectively
Overcome the taxonomic defficiency based on single pixel, single level；2) the utilized characteristic of division of object-based classification is multi-semantic meaning
, there are characteristics of objects, class correlated characteristic, scene characteristic etc., can effectively overcome the limitation only brought using spectral classification
Problem.
Although the present invention is described with reference to current better embodiment, those skilled in the art should be able to manage
Solution, above-mentioned better embodiment is only used for illustrating the present invention, any in the present invention not for limiting protection scope of the present invention
Spirit and spirit within, any modification, equivalence replacement, improvements for being done etc., should be included in the present invention right guarantor
Within the scope of shield.
Claims (10)
1. a kind of Landsat TM remote sensing image datas remove cloud method, it is characterised in that this method comprises the following steps：
A. Landsat TM remote sensing image datas are inputted；
B. using Object--oriented method is based on, multi-scale division is carried out to above-mentioned Landsat TM remote sensing image datas；
C. to the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to spissatus spectrum characteristic, computing is set up
Feature " ThickC "；
D. threshold classification method is utilized, the object of operating characteristic " ThickC " of threshold condition will be met by " unclassified " point
Class is to " Thick Cloud "；
E. it is special according to the spectrum characteristic of thin cloud and distribution to the Landsat TM remote sensing image datas after above-mentioned multi-scale division
Point, opening relationships feature " ThinC "；
F. in remaining " unclassified " object, using dual threshold classification by the object for the condition that meets
" unclassified " is classified to " Thin Cloud "；
G. " Thick Cloud " and " Thin Cloud " object homogeneous classification to " Cloud ", completion data are classified to by above-mentioned
Except cloud.
2. the method as described in claim 1, it is characterised in that the Landsat TM remote sensing image datas include 7 spectral coverages：
B1Bluish-green spectral coverage, B2Green spectral coverage, B3Red spectral coverage, B4Near-infrared spectral coverage, B5Nearly short-wave infrared spectral coverage, B6Thermal infrared spectral coverage, B7It is near short
The infrared spectral coverage of ripple.
3. method as claimed in claim 2, it is characterised in that described step b is specifically included：
According to the Landsat TM remote sensing image datas of input, imaged object layer is set up；
Selection participates in the weight for each wave band data that multi-scale division is calculated；
The scale parameter of multi-scale division is set；
The form parameter of multi-scale division is set；
Degree of the compacting parameter of multi-scale division is set；
The result of multi-scale division is calculated on imaged object layer, it is " unclassified " that the object produced after segmentation is assigned into class.
4. method as claimed in claim 3, it is characterised in that described step c is specifically included：
For each object ω={ P1,P2,P3,…,Pn, the pixel that P is included for the object, n is included for the object
Pixel quantity, the specific formula for calculation of its operating characteristic " ThickC " is as follows：
<mrow>
<mi>T</mi>
<mi>h</mi>
<mi>i</mi>
<mi>c</mi>
<mi>k</mi>
<mi>C</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msub>
<mi>&sigma;</mi>
<mi>i</mi>
</msub>
</mrow>
<mi>n</mi>
</mfrac>
</mrow>
Wherein,
<mrow>
<mi>&sigma;</mi>
<mo>=</mo>
<msub>
<mi>&sigma;</mi>
<msub>
<mi>B</mi>
<mn>2</mn>
</msub>
</msub>
<mo>*</mo>
<msub>
<mi>&sigma;</mi>
<msub>
<mi>B</mi>
<mn>4</mn>
</msub>
</msub>
<mo>*</mo>
<msub>
<mi>&sigma;</mi>
<msub>
<mi>B</mi>
<mn>7</mn>
</msub>
</msub>
<mo>*</mo>
<mi>C</mi>
</mrow>
Wherein, ThickC represents the numerical value of the operating characteristic " ThickC " of the object；N represents the pixel number that the object is included
Amount；Represent the B of a pixel2Brightness value；Represent the B of a pixel4Brightness value；Represent the B of a pixel7It is bright
Angle value；C represents a scaling constant, is set according to user's request；
To each object { ω in imaged object layer1,ω2,ω3,…,ωnOperating characteristic " ThickC " is all calculated respectively.
5. method as claimed in claim 4, it is characterised in that described step e is specifically included：
" ThinC ", if representing the object centered on an object, the beeline at this center object edge of Edge Distance is not
Other objects more than d pixel will be as the element for calculating center object relationship characteristic " ThinC " value, and it is described as follows：
Wherein, d represents pre-determined distance value；d(ωi,ωj) represent object ωiWith object ωjThe distance between；Represent to participate in
The set of all objects of calculated relationship feature " ThinC "；{ω∣d(ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,
3 ..., n } represent to meet Rule of judgment d (ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,3 ..., the collection of n all objects
Close；
According to operating characteristic and its statistic algorithm calculated relationship feature " ThinC "：
<mrow>
<mi>T</mi>
<mi>h</mi>
<mi>i</mi>
<mi>n</mi>
<mi>C</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msub>
<mi>ThickC</mi>
<mi>i</mi>
</msub>
</mrow>
<mi>n</mi>
</mfrac>
</mrow>
Wherein, ThinC represents the relationship characteristic value of the object；ThickC represents to participate in the object of calculated relationship feature " ThinC "
Operating characteristic " ThickC " value；N represents to participate in the quantity of the object of calculated relationship feature " ThinC ".
6. a kind of Landsat TM remote sensing image datas remove cloud system, it is characterised in that the system includes input module, segmentation mould
Block, operating characteristic set up module, sort module, relationship characteristic and set up module, wherein：
The input module is used to input Landsat TM remote sensing image datas；
The segmentation module is used to, using Object--oriented method is based on, carry out above-mentioned Landsat TM remote sensing image datas many
Multi-scale segmentation；
The operating characteristic, which sets up module, to be used for the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to thickness
The spectrum characteristic of cloud, sets up operating characteristic " ThickC "；
The sort module be used for utilize threshold classification method, will meet threshold condition operating characteristic " ThickC " object by
" unclassified " is classified to " Thick Cloud "；
The relationship characteristic, which sets up module, to be used for the Landsat TM remote sensing image datas after above-mentioned multi-scale division, according to thin
The spectrum characteristic and characteristic distributions of cloud, opening relationships feature " ThinC "；
The sort module is additionally operable in remaining " unclassified " object, and condition will be met using dual threshold classification
Object " unclassified " be classified to " Thin Cloud "；
The sort module is additionally operable to be classified to " Thick Cloud " and " Thin Cloud " object homogeneous classification is extremely by above-mentioned
" Cloud ", completes data and removes cloud.
7. system as claimed in claim 6, it is characterised in that described Landsat TM remote sensing image datas include 7 spectrums
Section：B1Bluish-green spectral coverage, B2Green spectral coverage, B3Red spectral coverage, B4Near-infrared spectral coverage, B5Nearly short-wave infrared spectral coverage, B6Thermal infrared spectral coverage, B7Closely
Short-wave infrared spectral coverage.
8. system as claimed in claim 7, it is characterised in that described segmentation module specifically for：
According to the Landsat TM remote sensing image datas of input, imaged object layer is set up；
Selection participates in the weight for each wave band data that multi-scale division is calculated；
The scale parameter of multi-scale division is set；
The form parameter of multi-scale division is set；
Degree of the compacting parameter of multi-scale division is set；
The result of multi-scale division is calculated on imaged object layer, it is " unclassified " that the object produced after segmentation is assigned into class.
9. system as claimed in claim 8, it is characterised in that described operating characteristic set up module specifically for：
For each object ω={ P1,P2,P3,…,Pn, the pixel that P is included for the object, n is included for the object
Pixel quantity, the specific formula for calculation of its operating characteristic " ThickC " is as follows：
<mrow>
<mi>T</mi>
<mi>h</mi>
<mi>i</mi>
<mi>c</mi>
<mi>k</mi>
<mi>C</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msub>
<mi>&sigma;</mi>
<mi>i</mi>
</msub>
</mrow>
<mi>n</mi>
</mfrac>
</mrow>
Wherein,
<mrow>
<mi>&sigma;</mi>
<mo>=</mo>
<msub>
<mi>&sigma;</mi>
<msub>
<mi>B</mi>
<mn>2</mn>
</msub>
</msub>
<mo>*</mo>
<msub>
<mi>&sigma;</mi>
<msub>
<mi>B</mi>
<mn>4</mn>
</msub>
</msub>
<mo>*</mo>
<msub>
<mi>&sigma;</mi>
<msub>
<mi>B</mi>
<mn>7</mn>
</msub>
</msub>
<mo>*</mo>
<mi>C</mi>
</mrow>
Wherein, ThickC represents the numerical value of the operating characteristic " ThickC " of the object；N represents the pixel number that the object is included
Amount；Represent the B of a pixel2Brightness value；Represent the B of a pixel4Brightness value；Represent the B of a pixel7It is bright
Angle value；C represents a scaling constant, is set according to user's request；
To each object { ω in imaged object layer1,ω2,ω3,…,ωnOperating characteristic " ThickC " is all calculated respectively.
10. system as claimed in claim 9, it is characterised in that described relationship characteristic set up module specifically for：
" ThinC ", if representing the object centered on an object, the beeline at this center object edge of Edge Distance is not
Other objects more than d pixel will be as the element for calculating center object relationship characteristic " ThinC " value, and it is described as follows：
Wherein, d represents pre-determined distance value；d(ωi,ωj) represent object ωiWith object ωjThe distance between；Represent to participate in
The set of all objects of calculated relationship feature " ThinC "；{ω∣d(ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,
3 ..., n } represent to meet Rule of judgment d (ωi,ωj)≤d, i=1,2,3 ..., n, j=1,2,3 ..., the collection of n all objects
Close；
According to operating characteristic and its statistic algorithm calculated relationship feature " ThinC "：
<mrow>
<mi>T</mi>
<mi>h</mi>
<mi>i</mi>
<mi>n</mi>
<mi>C</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msub>
<mi>ThickC</mi>
<mi>i</mi>
</msub>
</mrow>
<mi>n</mi>
</mfrac>
</mrow>
Wherein, ThinC represents the relationship characteristic value of the object；ThickC represents to participate in the object of calculated relationship feature " ThinC "
Operating characteristic " ThickC " value；N represents to participate in the quantity of the object of calculated relationship feature " ThinC ".
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710384330.9A CN107274361B (en) | 2017-05-26 | 2017-05-26 | Landsat TM remote sensing image data removes cloud method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710384330.9A CN107274361B (en) | 2017-05-26 | 2017-05-26 | Landsat TM remote sensing image data removes cloud method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107274361A true CN107274361A (en) | 2017-10-20 |
| CN107274361B CN107274361B (en) | 2019-11-19 |
Family
ID=60065324
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710384330.9A Active CN107274361B (en) | 2017-05-26 | 2017-05-26 | Landsat TM remote sensing image data removes cloud method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107274361B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108765329A (en) * | 2018-05-21 | 2018-11-06 | 北京师范大学 | A kind of spissatus minimizing technology and system of remote sensing image |
| CN109410164A (en) * | 2018-11-14 | 2019-03-01 | 西北工业大学 | The satellite PAN and multi-spectral image interfusion method of multiple dimensioned convolutional neural networks |
| CN111476723A (en) * | 2020-03-17 | 2020-07-31 | 哈尔滨师范大学 | Method for recovering lost pixels of remote sensing image with failed L andsat-7 scanning line corrector |
| CN114220020A (en) * | 2021-12-08 | 2022-03-22 | 深圳先进技术研究院 | Teak forest identification method, system and terminal |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105404753A (en) * | 2015-12-08 | 2016-03-16 | 中国科学院东北地理与农业生态研究所 | Marsh wetland mapping method based on object-oriented random forest classification method and medium-resolution remote sensing image |
| CN105957079A (en) * | 2016-04-28 | 2016-09-21 | 淮阴师范学院 | Lake water area information extraction method based on Landsat OLI multispectral image |
| CN105956557A (en) * | 2016-05-04 | 2016-09-21 | 长江水利委员会长江科学院 | Object-oriented timing sequence remote sensing image cloud coverage area automatic detection method |
-
2017
- 2017-05-26 CN CN201710384330.9A patent/CN107274361B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105404753A (en) * | 2015-12-08 | 2016-03-16 | 中国科学院东北地理与农业生态研究所 | Marsh wetland mapping method based on object-oriented random forest classification method and medium-resolution remote sensing image |
| CN105957079A (en) * | 2016-04-28 | 2016-09-21 | 淮阴师范学院 | Lake water area information extraction method based on Landsat OLI multispectral image |
| CN105956557A (en) * | 2016-05-04 | 2016-09-21 | 长江水利委员会长江科学院 | Object-oriented timing sequence remote sensing image cloud coverage area automatic detection method |
Non-Patent Citations (3)
| Title |
|---|
| 李炳燮等: "Landsat TM遥感影像中厚云和阴影去除", 《遥感学报》 * |
| 沈金祥等: "遥感影像云及云影多特征协同检测方法", 《地球信息科学学报》 * |
| 申文明等: "《基于天地一体化的工业固体废物监管技术研究》", 31 December 2013, 中国环境科学出版社 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108765329A (en) * | 2018-05-21 | 2018-11-06 | 北京师范大学 | A kind of spissatus minimizing technology and system of remote sensing image |
| CN108765329B (en) * | 2018-05-21 | 2020-12-04 | 北京师范大学 | Thick cloud removing method and system for remote sensing image |
| CN109410164A (en) * | 2018-11-14 | 2019-03-01 | 西北工业大学 | The satellite PAN and multi-spectral image interfusion method of multiple dimensioned convolutional neural networks |
| CN109410164B (en) * | 2018-11-14 | 2019-10-22 | 西北工业大学 | The satellite PAN and multi-spectral image interfusion method of multiple dimensioned convolutional neural networks |
| CN111476723A (en) * | 2020-03-17 | 2020-07-31 | 哈尔滨师范大学 | Method for recovering lost pixels of remote sensing image with failed L andsat-7 scanning line corrector |
| CN114220020A (en) * | 2021-12-08 | 2022-03-22 | 深圳先进技术研究院 | Teak forest identification method, system and terminal |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107274361B (en) | 2019-11-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107392925B (en) | Remote sensing image ground object classification method based on super-pixel coding and convolutional neural network | |
| CN110929607B (en) | Remote sensing identification method and system for urban building construction progress | |
| CN108846352B (en) | Vegetation classification and identification method | |
| Cihlar et al. | Classification by progressive generalization: A new automated methodology for remote sensing multichannel data | |
| CN101840581B (en) | Method for extracting profile of building from satellite remote sensing image | |
| CN107358260B (en) | Multispectral image classification method based on surface wave CNN | |
| CN107274361A (en) | Landsat TM remote sensing image datas remove cloud method and system | |
| CN106294705B (en) | A kind of batch remote sensing image preprocess method | |
| CN103035013B (en) | A kind of precise motion shadow detection method based on multi-feature fusion | |
| CN106650812B (en) | A kind of urban water-body extracting method of satellite remote-sensing image | |
| CN105976310A (en) | Partition-based VCA end-member extraction method | |
| CN109948593A (en) | Based on the MCNN people counting method for combining global density feature | |
| CN115713694B (en) | Land mapping information management method | |
| CN115439759B (en) | Method and device for extracting vegetation in remote sensing image, electronic equipment and medium | |
| CN107292328A (en) | The remote sensing image shadow Detection extracting method and system of multiple dimensioned multiple features fusion | |
| JP2019537151A (en) | Image processing apparatus, image processing method, and image processing program | |
| CN116030352B (en) | Long-time-sequence land utilization classification method integrating multi-scale segmentation and super-pixel segmentation | |
| CN107992856A (en) | High score remote sensing building effects detection method under City scenarios | |
| CN108229426B (en) | Remote sensing image change vector change detection method based on difference descriptor | |
| CN115082582A (en) | True color simulation method, system, equipment and medium for satellite remote sensing data | |
| CN111882573B (en) | Cultivated land block extraction method and system based on high-resolution image data | |
| Xiao et al. | Individual tree detection and crown delineation with 3D information from multi-view satellite images | |
| CN113378912A (en) | Forest area illegal reclamation land block detection method based on deep learning target detection | |
| CN112597936A (en) | Building rubbish separation method based on object-oriented hierarchical segmentation and related products | |
| Shingare et al. | Fusion classification of multispectral and panchromatic image using improved decision tree algorithm |
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 |