A kind of anti-atmospheric aerosol vegetation index computing method
Technical field
The present invention relates to a kind of anti-atmospheric aerosol vegetation index computing method, particularly relate to a kind of method calculating anti-atmospheric aerosol vegetation index based on neighborhood pixel.
Background technology
Vegetation-cover index (NDVI) index based on remote sensing plays an important role in the change of monitoring Global surface vegetation.Solar radiation arrives earth's surface, and the process entering satellite sensor through reflection is subject to atmospheric disturbance.Wherein, gasoloid is main interference factors.Therefore, gasoloid corrects becomes the key link being obtained earth's surface real reflectance by remote sensing observations.The error of atmospheric correction directly affects subsequent applications and comprises Images Classification, the inverting of ground mulching change and other Land Surface Parameters; Such as NDVI is affected.Atmospheric components cause observing NDVI and the true NDVI in earth's surface produce in error and (reduce 0.04-0.20 unit) based on aerosol scattering, secondly being steam (reducing 0.04-0.08 unit), is finally Ruili scattering (reducing 0.02-0.04 unit).Along with Urbanization Process In China is accelerated, the impact of gasoloid on climate change more and more comes into one's own, also for the development of remote optical sensing brings challenges.Atmospheric aerosol reduces vegetation index value by the contrast reduced between ruddiness and near infrared signal.PintyandVerstraete points out 0.15-0.2 the unit (M.M.Verstraete, B.P. (1992) .GEMI:anon-linearindextomonitorglobalvegetationfromsatel lites.Vegetation.) larger than observation NDVI value of the NDVI value after correcting.
At present, there are two class methods to be used for eliminating gasoloid to the impact of vegetation index: the first, obtain earth's surface real reflectance by atmospheric correction.These class methods mainly comprise radiative transfer model method (RTM), the dark goal approach (IDOS) of improvement and experience linear method (ELM).(Zhou,J.,Wang,J.,Li,J.,&Hu,D.(2011).AtmosphericcorrectionofPROBA/CHRISdatainanurbanenvironment.InternationalJournalofRemoteSensing,32(9),2591-2604.doi:10.1080/01431161003698443)。The atmospheric correction method of current widespread use comprises 6S, MODTRAN and LOWTRAN.They are all based on radiative transfer model, and AOD is wherein one of important input parameter.Theocean method is mainly comprised at present by the method for remote sensing inverting aerosol concentration, brightness method, contrastreduction method and DenselyDarkVegetation (DDV) method (YoramJ.Kaufman, A.E.W., LorraineA.Remer, Bo-CaiGao, Rong-RongLi, andLukeFlynn. (1997) .TheMODIS2.1-μm of Channel-CorrelationwithVisibleReflectanceforUseinRemoteS ensingofAerosol.IEEETransactionsonGeoscienceandRemoteSen sing, 35 (5), 1286-1298.).The second, define new Chinese People's Anti-Japanese Military and Political College's gas vegetation index.Kaufman and Tanr é utilizes blue wave band calibrating gas colloidal sol on the impact of red spectral band and proposes new vegetation index (ARVI) (Tanr é, Y.J.K.a.D. (1992) .Atmosphericallyresistantvegetationindex (ARVI) forEOS-MODIS.IEEETransactionsonGeoscienceandRemoteSensin g, 30 (2), 261-270.).Afterwards, not only anti-Soil Background but also the vegetation index of anti-atmospheric effect were as SARVI, MNDVI and SARVI2 (Huete, A.R., Liu, H.Q, Batchily, K, andVanLeeuwen, W. (1997) .AcomparisonofvegetationindicesoveraglobalsetofTMimagesf orEOS-MODIS.RemoteSensingofEnvironment, 59,440-451.).The people such as Karnieli propose a kind of anti-vegetation index AFVI newly, and this index uses short-wave infrared (1.6 μm or 2.1 μm) to replace red spectral band.
Existing the having some limitations property of method obtaining antiaero-sol vegetation index, is mainly manifested in the following aspects: 1, DDV method is difficult to be applied to the impact not having dark pixel at present; 2, the heterogeneity of gasoloid space distribution increases the complicacy of 6S model; 3, the 3rd wave band is used to replace the wave band affected by gasoloid to have input parameter spatial resolution and product space resolution is inconsistent and between wave band, spatial resolution is inconsistent phenomenon.
Summary of the invention
The object of this invention is to provide a kind of anti-atmospheric aerosol vegetation index computing method, need obtain and input the heterogeneous model error that brings of complicated air profile parameters, parameter space distribution and image to be corrected does not exist dark pixel and cannot carry out the technical matters of atmospheric correction to solve.
For realizing above goal of the invention, the invention provides a kind of anti-atmospheric aerosol vegetation index computing method, comprising the steps;
Step 1: pre-service is carried out to satellite image, obtains apparent reflectance;
Step 2: atmospheric correction is carried out to satellite image;
Step 3: remove satellite image medium cloud pixel;
Step 4: satellite image sample area size is set; And satellite image is divided according to sample area size; Described sample area is arranged according to homogeneous intensity;
Step 5: traversal sample area, and judge whether that traversal completes;
If when traversal does not complete, then perform step 501;
If when having traveled through, then perform step 502;
Step 501: the slope value traveling through line between each neighborhood pixel and center pixel apparent reflectance in calculating sampling region; And judge whether that traversal completes;
If when having traveled through, then perform step 50101 to step 50102;
If when traversal does not complete, then perform step 501;
Step 50101: all slope value in step 501 being positive number are averaged;
Step 50102: by the mean value in step 50101, recovers the vegetation-cover index of corresponding center pixel; And perform step 5;
Step 502: the vegetation-cover index according to recovering corresponding center pixel carries out image co-registration to satellite image.
It is further, as follows to the computing formula of the slope value of line between each neighborhood pixel and center pixel apparent reflectance in described step 501,
Wherein, k
ij' represent slope value, N
jand R
jrepresent the jth neighborhood pixel apparent reflectance at ruddiness and near-infrared band respectively, N
iand R
ibe respectively the apparent reflectance of center pixel i at ruddiness and near-infrared band.
Further, the computing formula of averaging to all slope value in step 501 being positive number in described step 50101 is as follows,
Wherein, k
i' represent mean value.
Further, the computing formula being recovered the vegetation-cover index of corresponding center pixel in described step 50102 by the mean value in step 50101 is as follows,
Wherein NDVI
i' represent the vegetation-cover index of sample area center pixel.
Further, described sample area is 5 pixel *, 5 pixels according to the size that homogeneous intensity carries out arranging.
Further, also comprise in described step 50101 and first statistical method filtration is carried out to all slope value being positive number.
Further, the method that described all slope value to being positive number carry out statistical method filtration comprises, the value remove minimum value, removing maximal value and remove beyond average two standard deviations.
Compared with prior art, the invention has the beneficial effects as follows:
1. use and satellite image is carried out sample area division, and the mean value of the slope value of line between neighborhood pixel each in sample area and center pixel apparent reflectance is calculated, go out with this mean value calculation the technical scheme that vegetation-cover index merges image, obtain without the need to the air profile parameters obtained and input is complicated, avoid parameter space to distribute the heterogeneous model error that brings and avoid image to be corrected to there is not dark pixel and the technique effect of atmospheric correction cannot be carried out;
2. use the technical scheme of slope value being carried out to statistical method process, obtain the pixel of avoiding occurring singular value and increase the technique effect that slope mean value causes error.
Accompanying drawing explanation
Fig. 1 is the process flow diagram in background technology of the present invention;
Fig. 2 is method in the present invention corrects vegetation-cover index and the apparent vegetation-cover index concentration change obtained comparison diagram to gasoloid.
Embodiment
Below in conjunction with the drawings and specific embodiments, the invention will be further described.
Embodiment 1:
As shown in Figure 1, anti-atmospheric aerosol vegetation index computing method of the present invention, comprise the steps;
Step 1: pre-service is carried out to satellite image, obtains apparent reflectance;
Step 2: atmospheric correction is carried out to satellite image;
Step 3: remove satellite image medium cloud pixel;
Step 4: satellite image sample area size is set; And satellite image is divided according to sample area size; Described sample area is arranged according to homogeneous intensity;
Step 5: traversal sample area, and judge whether that traversal completes;
If when traversal does not complete, then perform step 501;
If when having traveled through, then perform step 502;
Step 501: the slope value traveling through line between each neighborhood pixel and center pixel apparent reflectance in calculating sampling region; And judge whether that traversal completes;
If when having traveled through, then perform step 50101 to step 50102;
If when traversal does not complete, then perform step 501;
Step 50101: all slope value in step 501 being positive number are averaged;
Step 50102: by the mean value in step 50101, recovers the vegetation-cover index of corresponding center pixel; And perform step 5;
Step 502: the vegetation-cover index according to recovering corresponding center pixel carries out image co-registration to satellite image;
Specifically, when needs are analyzed satellite image, first radiation calibration is carried out to obtain apparent reflectance to image, then use atmospheric correction software; The Flash module of such as complete Remote Sensing Image Processing (ENVI), corrects steam in image and ozone, and is set to not process to aerosol model and gasoloid inverting, and initial visibility is set to 100km.The cloud pixel in image is removed according to data BQA wave band quality document.Sample area size is set, the earth's surface stronger to homogenieity; Such as: large stretches of forests, meadow or farmland, in order to ensure model stability, sample area can be set and be the bigger the better; For the earth's surface that homogenieity is more weak, although or inner homogeneous discretize, fragmentation earth's surface, now adopt region should not arrange excessive.Be generally the stability ensureing model, sample area be set to 5*5 (unit is pixel) best.After sample area sets, satellite image is divided into several sample area.The slope value k of line between each neighborhood pixel and center pixel apparent reflectance is calculated in each sample area
i', computing formula is as follows:
Wherein, k
ij' represent slope value, N
jand R
jrepresent the jth neighborhood pixel apparent reflectance at ruddiness and near-infrared band respectively, N
iand R
ibe respectively the apparent reflectance of center pixel i at ruddiness and near-infrared band.
After the slope value of line has calculated between all neighborhood pixels in a sample area and center pixel apparent reflectance, choose the slope k that all slope value are greater than zero
j', calculate its mean value, computing formula is as follows:
Wherein k
i' represent the slope k that all slope value are greater than zero
j' mean value.
Due at slope calculations k
j' process in, may singular value be there is, error be produced to the calculating of follow-up mean value, so when calculating mean value, slope value is greater than to the slope k of zero
j' carry out the process of statistical method; Such as: remove minimum value method at large stretches of forests or large stretch of arable land regional choice; The method of the value beyond average two standard deviations is removed at plaque rupture regional choice.
The mean value k of a sample area
i' calculated after, can according to k
i' vegetation-cover index in this employing region is calculated, computing formula is as follows:
Wherein NDVI
i' be the vegetation-cover index in this employing region.
Finally according to the vegetation-cover index of all sample area, image co-registration is carried out to satellite image.
Embodiment 2:
Below utilize simulated data to checking of the present invention.
As Fig. 2, this figure is the method used in the present invention, corrects to satellite image the comparison diagram that the NDVI that obtains and apparent NDVI changes with aerosol concentration.The experiment parameter of concrete utilization is as follows: utilize ESPA to provide August in 2014 19 day the morning 11 time China Landsat8 earth's surface, Efficiency in Buildings in Tianjin Area real reflectance product, use the apparent reflectance that 6S atmospheric correction models earth's surface real reflectance image is corresponding when AOD=0.05, AOD=0.3, AOD=0.5 and AOD=1.0.
As seen in Figure 2, gasoloid has a significant impact apparent reflectance, causes apparent NDVI cannot truly reflect vegetation cover situation.The present invention well eliminates the impact of gasoloid on NDVI, achieves the effect of earth's surface vegetation variation being carried out to Fast Evaluation under complicated aerosol concentration.
Embodiment 3:
Form 1
Form 2
As shown in form 1, form 2, the precision test result that form 1 is Yichun, Suihua Area uses the inventive method to correct gasoloid, form 2 is precision test results that Beijing area uses the inventive method and corrects gasoloid.
In figure, " region " represents the region of interest selected from remote sensing image; " AE (absoluteerror) " represents absolute error; " size " represents the number of pixel in region of interest; " vegetationcoverage " represents vegetation coverage grade; " extentofcorrection " represents the degree (being tried to achieve by form secondary series/first row) that gasoloid corrects.
The experiment parameter of concrete utilization is as follows: when utilizing 2014 on August 19, the morning 11 China Efficiency in Buildings in Tianjin Area with 2014 on August 9, the morning 10 time Heilongjiang Province of China Yichun and corresponding time of providing of the Lansat8OLI image of Suihua City intersection and ESPA and regional earth's surface real reflectance product.
Because Chinese Efficiency in Buildings in Tianjin Area is located in Jing-jin-ji region economic circle, aerosol concentration is high compared with other areas; Heilongjiang Province's Yichun and the domestic river valley of Suihua City gather, and have large stretch of forest land, meadow and farmland.First utilize Flash atmospheric correction module in ENVI software to carry out steam and ozone correction to image, obtain the ruddiness after correcting and near-infrared band.Method in the rear the present invention of utilization carries out gasoloid correction on this basis, and correction result and earth's surface real reflectance product carry out contrast verification arithmetic accuracy.
The result shows, the method in the present invention has higher precision on the homogeneous districts such as the forest on molecular scattering correction basis, farmland or meadow.
In addition to the implementation, the present invention can also have other embodiments, and all employings are equal to the technical scheme of replacement or equivalent transformation formation, all drop in protection scope of the present invention.