CN104700456A - Method and system for pixel screening and filtering of SeaWinds scatterometer wind vector inversion - Google Patents

Abstract

The invention provides a method and a system for pixel screening and filtering of a SeaWinds scatterometer wind vector inversion. The method for pixel screening of the SeaWinds scatterometer wind vector inversion comprises the steps as follows: determining each fuzzy solution and a total gray value for each fuzzy solution of pixels in the SeaWinds scatterometer wind vector inversion; finding out a target fuzzy solution that is closest to numerical wind field data; judging whether the total gray value of the target fuzzy solution is greater than a preset gray value threshold; if it is, making a mark to the pixels. The method for filtering the SeaWinds scatterometer wind vector inversion comprises the following steps: using the provided screening method for screening the pixels in the SeaWinds scatterometer wind vector inversion; selecting a filtering window; acquiring a number of the pixels having the marks in the filtering window; judging whether the number of the pixels is greater than zero; if it is, calculating a wind vector of the filtering window based on a preset calculation manner corresponding to the number of the pixels; if it is not, acquiring the wind vector of the filtering window based on L2B data. Accordingly, the precision of the wind vector inversion can be increased.

Description

The picture dot screening of SeaWinds scatterometer wind vector retrieval, filtering method and system

Technical field

The present invention relates to ocean remote sensing field, particularly relate to a kind of picture dot screening of SeaWinds scatterometer wind vector retrieval, filtering method and system.

Background technology

Sea wind is the important motivity factor acting on sea surface, drive ocean from surface wave the marine motor to the various yardsticks of deep layer ocean current system, little, form ocean wave, greatly then promote ocean current development.Sea wind is by regulating the exchange of heat, steam and chemical substance between seawater and air, adjust Air-sea flux effect, maintain the whole world and Using A Regional Climate Model, this coupling has material impact to the whole world and regional climate, even can cause the change of global environment, as EI Nino phenomenon.And sea wind can bring the numerical weather forecast system in region or the whole world into, improve the ability of weather forecast.Therefore, the sea wind data of acquisition high precision, high-spatial and temporal resolution have very important scientific research value and realistic meaning for ocean dynamics, meteorology, climatological research and mankind's Appropriate application wind energy resources.

The conventional vision systems of Ocean Wind-field is mainly through boats and ships, marine marker and along bank station etc.For the ocean of area 70% covering the whole world, the Wind Data that conventional means obtains is very limited and spend huge.Satellite remote sensing technology is that the measurement of Ocean Wind-field provides brand-new means, and it has broad covered area, observation continuously, by advantages such as weather effect are little, the businessization that entered is run.The satellite borne sensor of Global ocean wind field information can be provided to comprise scatterometer, microwave radiometer, synthetic-aperture radar (SAR), satellite altimeter, but only have scatterometer can provide the complete wind vector information comprising wind speed size and wind direction on a large scale simultaneously.

Scatterometer WIND FIELDS is divided into three each and every one steps: the first step, set up physical geography module function (geophysical model function, GMF), second step, 2-4 fuzzy solution is obtained by physical geography module function inverting, 3rd step, adopts circle median filter method to remove fuzzy solution, obtains unique true solution.The precondition that circle median filter algorithm effectively runs is that the picture dot number of wind vector application condition in filter window large (wind direction inversion error is greater than 45 degree) can not exceed half.Because the inversion error entirety of SeaWinds scatterometer observation during high wind speed is larger, thus when using circle median filter algorithm, the picture dot number that in filter window, inversion error is larger, often beyond half, makes circle median filter poor effect, makes final inversion error higher.

Summary of the invention

One object of the present invention is the picture dot screening technique and the system that provide a kind of SeaWinds scatterometer wind vector retrieval, can contribute to the wind vector retrieval precision under raising high wind speed situation.

This purpose of the present invention is achieved through the following technical solutions:

A picture dot screening technique for SeaWinds scatterometer wind vector retrieval, comprises the steps:

Determine each fuzzy solution of the picture dot in SeaWinds scatterometer wind vector retrieval and total gray-scale value of each described fuzzy solution;

Each described fuzzy solution and numerical value wind field data are contrasted, finds out objective fuzzy solution immediate with numerical value wind field data;

Judge whether total gray-scale value of described objective fuzzy solution is greater than default gray threshold;

If so, a mark is done to described picture dot.

A picture dot screening system for SeaWinds scatterometer wind vector retrieval, comprising:

Processing module, for total gray-scale value of each fuzzy solution and each described fuzzy solution of determining the picture dot in SeaWinds scatterometer wind vector retrieval;

Contrast module, for each described fuzzy solution and numerical value wind field data being contrasted, finds out objective fuzzy solution immediate with numerical value wind field data;

First judge module, for judging whether total gray-scale value of described objective fuzzy solution is greater than default gray threshold;

Mark module, for when the judged result of described first judge module is for being, makes a mark to described picture dot.

According to the scheme of the invention described above, it first determines each fuzzy solution of picture dot and corresponding total gray-scale value thereof, objective fuzzy solution is obtained again by each fuzzy solution and numerical value wind field data are carried out contrast, judge whether total gray-scale value of this objective fuzzy solution is greater than default gray threshold, if, object meta makes a mark, each picture dot in SeaWinds scatterometer wind vector retrieval can screen in this manner, the markd picture dot finally obtained is the less picture dot of the inversion error that filters out, circle median filter can be carried out based on these picture dots, because screening process effectively can reject the larger picture dot of wind vector retrieval error, thus improve the precision of circle median filter algorithm largely, namely improve final wind vector retrieval precision.

Another object of the present invention is to filtering method and system that a kind of SeaWinds scatterometer wind vector retrieval is provided, the wind vector retrieval precision under raising high wind speed situation can be contributed to.

This purpose of the present invention is achieved through the following technical solutions:

A filtering method for SeaWinds scatterometer wind vector retrieval, comprises the steps:

The picture dot screening technique of SeaWinds scatterometer wind vector retrieval as above is adopted to screen the picture dot in SeaWinds scatterometer wind vector retrieval;

Choose filter window, obtain the markd picture dot number in this filter window;

Judge whether described picture dot number is greater than zero;

If so, then corresponding according to described picture dot number default account form calculates the wind vector of described filter window;

If not, then according to the wind vector of the described filter window of the acquisition of L2B data.

A filtering system for SeaWinds scatterometer wind vector retrieval, comprises the picture dot screening system of SeaWinds scatterometer wind vector retrieval as above, also comprises:

Choosing module, for choosing filter window, obtaining the markd picture dot number in this filter window;

Second judge module, for judging whether described picture dot number is greater than zero;

Filtration module, for when the judged result of described second judge module is for being, the default account form corresponding according to described picture dot number calculates the wind vector of described filter window, when the judged result of described second judge module is no, according to the wind vector of the described filter window of the acquisition of L2B data.

According to the scheme of the invention described above, it first adopts the picture dot screening technique of SeaWinds scatterometer wind vector retrieval as above to screen the picture dot in SeaWinds scatterometer wind vector retrieval under high wind speed situation, choose filter window again, obtain the markd picture dot number in this filter window, judge whether described picture dot number is greater than zero, if, then corresponding according to described picture dot number default account form calculates the wind vector of described filter window, if not, then according to the wind vector of the described filter window of the acquisition of L2B data, the picture dot that the wind vector retrieval error can effectively rejected in filter window due to screening process is larger, thus improve the precision of circle median filter algorithm largely, simultaneously, the picture dot occurrence probability very large due to inversion error under high wind speed situation is larger, in whole filter window, total gray-scale value is greater than the number also instability of the picture dot of the gray threshold of setting, may be 0, may be 1, also may be 2, the account form corresponding according to different picture dot numbers calculates the wind vector of filter window, also improves final wind vector retrieval precision.

Accompanying drawing explanation

Fig. 1 is the model table being coordinate with Radar backscattering coefficients and relative bearing of SeaWinds scatterometer physical geography module function;

Fig. 2 is the collection of illustrative plates schematic diagram of same picture dot observed reading in wind vector solution space;

Fig. 3 is the schematic flow sheet of the picture dot screening technique embodiment of SeaWinds scatterometer wind vector retrieval of the present invention;

Fig. 4 is fuzzy solution region intersection point distribution schematic diagram;

Fig. 5 is the schematic flow sheet of mode wherein in an embodiment of the total gray-scale value determining fuzzy solution;

Fig. 6 is collection of illustrative plates algorithm wind vector retrieval error and total gray-scale relation schematic diagram;

Fig. 7 is the schematic flow sheet of the filtering method embodiment of SeaWinds scatterometer wind vector retrieval of the present invention;

Fig. 8 is the structural representation of the picture dot screening system embodiment of SeaWinds scatterometer wind vector retrieval of the present invention;

Fig. 9 is processing module in Fig. 8 structural representation wherein in an embodiment;

Figure 10 is the schematic flow sheet of the filtering system embodiment of SeaWinds scatterometer wind vector retrieval of the present invention.

Embodiment

For making object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is described in further detail.Should be appreciated that embodiment described herein only in order to explain the present invention, do not limit protection scope of the present invention.

For the ease of understanding the solution of the present invention, first traditional scatterometer Wind-field Retrieval mode is described below.

1) general step of traditional wind field backscattering meter Wind-field Retrieval mode

Tradition wind field backscattering meter Wind-field Retrieval mode can be divided into three steps: set up physical geography module function, wind vector solution search (inversion algorithm) and wind vector fuzzy solution are removed.Be transmitted into now from first satellite SEASAT-A being specifically designed to ocean remote sensing in 1978, foreign scholar proposes the multiple pattern function for the remote sensing of scatterometer Ocean Wind-field and inversion algorithm.Due to the biharmonic characteristic of physical geography module function and the error of existence in observing, the result that wind vector is searched for is unique, but there is 1-4 minimal value, is called fuzzy solution.Got rid of owing to being difficult to obtain enough information from single picture dot, thus need an independently fuzzy solution removal program, got rid of to obtain unique true solution.

2) ultimate principle of scatterometer work

Scatterometer is a kind of special active microwave sensor, is mainly used for obtaining multiple measurement data such as sea radar cross section.By launching the microwave of certain wavelength to earth's surface, those measurement data having different geometry observed parameter in same surface units are combined, obtain this unit ocean surface wind speed and wind direction through certain process.The measurement of scatterometer to sea surface wind vector is a kind of indirect relation.

Sea microwave scattering mainly contains two kinds of mechanism: when incident angle is less than 20 °, microwave backscattering mainly mirror-reflection; When incident angle is greater than 20 °, the mainly Bragg diffraction of microwave backscattering.And the incident angle of scatterometer is generally greater than 20 °, as inside and outside SeaWinds scatterometer, the incident angle of two wave beams is respectively 46 ° and 54 °, so during scatterometer work, sea microwave scattering is based on Bragg diffraction.Be under the condition of microwave in incident wavelength, what can cause Bragg diffraction can only be superimposed on centimeter wave on wave or capillary wave, and Bragg diffraction and the angle between centimeter wave crest and radar line of sight have very large correlativity.

Wind transmits momentum during from air to sea, sea can be made to become coarse, and form the centimeter wave of small scale, centimeter wave quantity is directly proportional relevant to the friction wind speed on sea, and centimeter wave crest direction is responsive to wind direction, so by anti-wind vector and the wind direction releasing sea of the observation of micron wave.Scatterometer is exactly the back scattering energy size measuring sea, and the sea radar cross section NRCS of standard of appraisal (NormalizedRadar Cross Section).

3) physical geography module function of ocean surface wind retrieving

After utilizing the scatterometer sea that (incident angle and polarization mode) records under different Parameter Conditions, footpath Radar Cross Section carries out Wind-field Retrieval, needs to know that sea normalized radar backscatter cross section amasss the relation between several variablees such as (NRCS) and wind speed, relative bearing (angle between wind direction and radar observation direction), incident angle and polarization mode.The object of pattern function is exactly the relation of accurate description normalized radar backscatter cross section long-pending (NRCS) and sea environmental parameter and instrument parameter.Pattern function has crucial effect for the design of scatterometer and inversion algorithm, and its precision can have influence on the major issue such as the precision of Wind-field Retrieval and the feasibility of scatterometer design widely.

The satellite SEASAT being specifically designed to ocean remote sensing from first is transmitted into now, and people constantly propose pattern function for the remote sensing of scatterometer Ocean Wind-field and inversion method, the non-linear complicacy determining Wind-field Retrieval method of pattern function.But having the measurement data of a large amount of Ku wave bands to prove, marine radar back scattering effects on surface wind speed is very responsive.

General physical geography module function can be expressed as form:

σ ⁰＝F(ω,χ；....；θ,p,f) (1)

In formula (1), ω and χ is illustrated respectively in indifferent equilibrium wind speed on sea on 19.5 meters of height and apparent wind to (angle of scatterometer beams incident position angle and actual wind direction); " ... " represent other secondary non-sea wind envirment factor of impact, as air stability, wave long wave state, Hai Wen, sea water permittivity etc.; θ, p and f are scatterometer parameters, represent relevant beams incident angle, polarization mode (horizontal or vertical) and radar wave frequency respectively.

Usually, SeaWinds scatterometer adopts new model function QSCAT1.The incident angle that this pattern function is corresponding only has 46 ° (horizontal polarizations) and 54 ° (vertical polarization).In order to obtain unified scatter meter model so that research, Wentz, F.J. after 2000, on the basis that SASS2 is corrected, in SASS2 model 46 ° and 54 ° of incident angle parts are replaced with the data of QSCAT1 model, the model be merged into is called new model function QSCAT1 model.The expression-form that this model data is not fixed, for be conducive to improving travelling speed and be conducive to model and algorithl and separate, the form of employing form.This form is represented by incident angle, speed and this three-dimensional of phase the other side wind direction.In FIG, incident angle, from 40 ° to 60 °, is spaced apart 1 degree, and is horizontal polarization mode from 40 ° to 50 °, is vertical polarization mode from 51 ° to 60 °; Speed, from 1m/s to 50m/s, is spaced apart 1m/s; And apparent wind is to from 0 ° to 180 °, be spaced apart 2 °.In addition, under identical wind speed and direction condition, the value that 46 ° of wave beams record is relatively little.Shown in Fig. 1 be incident angle is 46 °, and the model table being coordinate with Radar backscattering coefficients and relative bearing, unit is natural unit, and different wind speed is represented by different curves, and the wind speed of 0 ° to 360 ° is about 180 ° of symmetries.With the relation that ocean surface wind speed is non-linear increasing, wherein, m/s represents metre per second (m/s).

4) traditional wind vector solves mode

Geophysics algorithm utilizes pattern function exactly, measures σ with the radar raster-displaying after atmospheric attenuation corrects ⁰calculate the wind speed and direction at certain altitude place, sea.But, due to the characteristic of physical geography module function, same observed reading σ ⁰, can corresponding a series of possible wind speed and direction value, as shown in Figure 2, in Fig. 2, the thin curve of every bar represents an observed reading σ ⁰corresponding possible wind speed and direction value.By Fig. 2, can find out obviously, at least need two curves (that is, two observed readings) just may determine wind speed and direction solution (that is, the intersection point of two curves) in theory.In scatterometer observation practice, scatterometer carries out the repeated measures of multi-angle to same place sea, obtains multiple observed reading σ of different directions ⁰, then have many curves corresponding to wind vector solution space, last wind vector solution is just near these intersections of complex curve.And the task of wind vector retrieval algorithm is, by these observed readings (curve in figure), search out corresponding wind vector solution by certain computing rule.

Before satellite scatterometer is launched, three candidate algorithm are just had to define.Since satellite launch, these algorithms by comparing with independent measurement data, have become very ripe in a series of data analysis process.Chong-yung Chi and Fuk L.Li has carried out comparative study to different algorithms, find MLE (maximum likelihood method) and L1 two kinds of algorithms better.And maximum likelihood algorithm is concerning the rationale having it more deep Wind-field Retrieval, so be used to treatment S eaWinds scatterometer data at present.

Ideally, namely suppose there is not model error and measuring error, in same resolution cell, all normalized radar backscatter cross section amasss possible wind vector solution curve corresponding to measured value and should meet at any or some in wind speed, wind direction two-dimensional space, and the wind vector that these points are corresponding is wind vector fuzzy solution.Owing to there is interfering noise and measuring error, n bar curve does not meet at a bit or some, but around this point or some, form the relatively intensive intersection point collection of several distribution (boxed area in Fig. 2), as can be seen from Figure 2, in figure, intersections of complex curve has four than the region of comparatively dense, correspond to four possible solutions of wind vector (due to the biharmonic characteristic of physical geography module function, make the intersection point of the curve corresponding to every two observed readings have 2-4, this means by observed reading σ ⁰inverting possible wind vector solution out has 2-4), these four solutions are referred to as fuzzy solution, and final wind vector true solution is one in these four solutions.The principle of traditional approach is determined the position (namely determining each fuzzy solution) of each fuzzy solution, and the general type of objective function is as follows:

$J(U,\mathrm{\φ})=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\left|\frac{e_i}{{\mathrm{\δ}}_i}\right|}^p+q\·\ln {\mathrm{\δ}}_i^p---\left(2\right)$

Wherein, e _ifor the difference between NRCS observed reading and theoretical value, δ _ibe the standard deviation of i-th observed reading, p, q are constant parameter, relevant with concrete inversion algorithm.

5) fuzzy solution minimizing technology

For obtaining last true solution, needing the information by around picture dot, removing unnecessary puppet solution by the method for filtering.Early stage scientist finds by the puppet solution in the initialized wind field of the first fuzzy solution very similar to the noise in digital picture, and the median filtering algorithm in digital image processing techniques can effectively can remove noise while maintenance image border, according to this point, they introduce and revise median filtering algorithm, make it be applicable to the wind vector ambiguity removal of circle distribution.Facts have proved, circle median filter can effectively remove wind vector fuzzy solution, becomes the main algorithm of scatterometer ambiguity removal.

The introduction (Xie Xuetong etc., 2005) of following is circle median filter:

I) definition counted in circle

For circle distributing vector, in circle, number makes following formula minimum:

$S\left(m\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\left|\right|V_m-V_i\left|\right|,1\&le;m<N---\left(3\right)$

Wherein, N is the sum of picture dot in filter window, and m represents m picture dot in filter window.

II) concrete steps of circle median filter algorithm

Circle median filter algorithm is the window (filter window) by opening a certain size in wind field two-dimensional space, calculate number in the circle of this window data, then the wind vector with the immediate alternative current window center of number in circle is found out in several fuzzy solutions that imago unit is corresponding in the window, shift to the next position, repeat this operation until to whole wind field end of operation.This process iterates, until wind field no longer changes or iterations reaches default maximum iteration time.

The prerequisite that scatterometer median filtering algorithm normally works is that initial wind field must meet true value and accounts for more than 50%, and pseudo-solution stochastic distribution, so initialization will be carried out to wind field by certain principle before filtering, the wind field after initialization is made to meet the precondition of median filtering algorithm.For SASS and NSCAT scatterometer, first fuzzy solution accounts for the ratio of true value more than 50%, and in whole space stochastic distribution, this meets the starting condition of median filtering algorithm, so they utilize the first fuzzy solution initialization wind field just, circle median filter method is then utilized to carry out wind vector ambiguity removal.But SeaWinds scatterometer is due to its Instrument Design feature, the first fuzzy solution of the wind field outside track is made not to be wind vector true value in very large possibility, this easily causes spatially pseudo-solution integrated distribution, therefore initialization must be carried out according to other engineering philosophy, to give full play to the effect of median filtering algorithm in scatterometer ambiguity removal.The initial method of median filter that NASA adopts Freilich and Dunbar to propose in 1994, namely utilizes numerical value wind field data (NWP, numerical weather production) to carry out initialization wind field.First logarithm value wind field data carry out interpolation by the grid of SeaWinds scatterometer, then, concentrate select an immediate initial wind vector as this picture dot with the wind direction numerical value wind field data from the fuzzy solution of each picture dot.

For SeaWinds scatterometer, its circle median filter algorithm steps can be summarized as follows:

A) utilize numerical value wind field data to the picture dot in filter window carry out initialization (that is, in Choose filtering window each picture dot fuzzy solution in the wind direction value of numerical value wind field data that fuzzy solution immediate as initial wind arrow value);

B) to the picture dot in the filter window after initialization, formula (3) is utilized to obtain in its circle the wind arrow value corresponding to counting, as the wind vector true solution of filter window center picture dot;

C) cycling above-mentioned a), b), until the wind arrow value of each picture dot no longer changes in filter window.

Known as previously mentioned, SeaWinds scatterometer is due to its Instrument Design feature, the first fuzzy solution of the wind field outside track is made not to be wind vector true value in very large possibility, this easily causes spatially pseudo-ly separates integrated distribution, the requirement therefore needing to utilize numerical value wind field data to carry out initialization wind field and normally work to reach filtering algorithm.Although under most situation, circle median filter algorithm can remove the puppet solution because observational error causes effectively, but, when the observational error of picture dot most of in whole filter window is all larger (because observation condition difference causes), most of picture dot and the immediate fuzzy solution of numerical value wind field data all can depart from comparatively far away with true solution, now, the wind field obtained after circle median filter algorithm process, its wind direction error will be larger, that is, fuzzy solution identification error occurs.In order to improve the precision of circle median filter algorithm, to improve final wind vector retrieval precision further, proposing the solution of the present invention, below the present invention program being described in detail.

In the following description, first the embodiment for the picture dot screening technique of SeaWinds scatterometer wind vector retrieval of the present invention, the filtering method of SeaWinds scatterometer wind vector retrieval is described, then is described each embodiment of the picture dot screening system of SeaWinds scatterometer wind vector retrieval of the present invention of the present invention, the filtering system of SeaWinds scatterometer wind vector retrieval.

Shown in Figure 3, be the schematic flow sheet of the picture dot screening technique embodiment of SeaWinds scatterometer wind vector retrieval of the present invention.As shown in Figure 3, the picture dot screening technique of the SeaWinds scatterometer wind vector retrieval of the present embodiment comprises the steps:

Step S101: determine each fuzzy solution of the picture dot in SeaWinds scatterometer wind vector retrieval and total gray-scale value of each described fuzzy solution;

As previously mentioned, can be determined the position of each fuzzy solution by the extreme value of searching for objective function, namely determine each fuzzy solution of picture dot, wherein objective function is formula (2), does not repeat them here;

As shown in Figure 2, the each fuzzy solution region (below this region being referred to as intersection point close quarters) that a corresponding intersection point is relatively intensive respectively, in the present embodiment fuzzy solution total gray-scale value reaction be the dense degree of intersection point in corresponding intersection point close quarters;

The center that fuzzy solution is positioned at intersection point close quarters certain feature locations maybe near this place, the distribution of intersection point close quarters collection and morphological feature (i.e. TuPu method) have expressed size and the Producing reason of interfering noise and measuring error, that is, the information of observational error is wherein contained; From foregoing teachings, in wind vector solution space, (Fig. 2) every bar curve all represents an observed reading, and the distribution of their intersection point reflects the information of observational error; Before the distribution characteristics (TuPu method) of these intersection points of quantitative description, be necessary first qualitatively to learn about its distribution pattern roughly, as shown in Figure 4.

Three figure in Fig. 4 are that the region intercepting one fixed width near fuzzy solution from whole solution space (0-360 °) obtains, as can be seen from Figure 4, in each intersection point close quarters intersection point distribution be different, obviously the standard deviation only adopting fuzzy solution region intersection point cannot describe the distribution of intersection point completely, for this reason, be the distribution describing intersection point with total gray-scale value of fuzzy solution in the present embodiment;

Wherein in an embodiment, as shown in Figure 5, determine that the mode of the gray-scale value of fuzzy solution can comprise the steps:

Step S201: each intersection point close quarters of observed reading in wind vector solution space obtaining described picture dot;

As shown in Figure 2, the wind speed that the thin curve of every bar is possible under representing an observed reading, wind direction combine; For a given wind direction value, every bar curve (represents an observed reading σ ⁰) have a corresponding air speed value, n bar curve then has n corresponding air speed value, and the thick line above in Fig. 2 is the mean value of these air speed value (n is individual), below that thick line be then the mean square deviation of these air speed value; Mean square deviation minimal value place (namely, the lowest point of thick line below) differ minimum place for air speed value, also intersections of complex curve thick can be thought, therefore, determine that the mode of intersection point close quarters can be specifically: first according to the mean square deviation determination minimum point of air speed value, again to be that certain angle is expanded on middle alignment both sides by minimum point and perpendicular to the line of transverse axis, as 5 degree or 10 degree, namely form an intersection point close quarters;

Step S202: respectively pixel division is carried out to each described intersection point close quarters, and determine the intersection point number falling into each pixel;

When carrying out pixel and dividing, the line number of pixel and columns can set according to actual needs, such as, each intersection point close quarters are divided into 20 × 10 pixels, after completing pixel division, add up the intersection point number in each pixel respectively;

Step S203: determine the total number of intersection point in the area of the ellipse that each described intersection point close quarters is corresponding and each described intersection point close quarters according to described intersection point number respectively, wherein, described ellipse is for including the intersection point in described intersection point close quarters;

In order to the distribution of intersection point can be described completely, the intersection point included by an ellipse in intersection point close quarters in the present embodiment, the area of the ellipse that each described intersection point close quarters is corresponding can be determined according to described intersection point number, particularly, first oval major axis and minor axis can be determined by following formula (4), formula (5), and based on the major axis obtained, minor axis, and calculate oval area in conjunction with ellipse area computing formula;

$a={\left(\frac{2[{\mathrm{\&mu;}}_{20}+{\mathrm{\&mu;}}_{02}+\sqrt{{({\mathrm{\&mu;}}_{20}-{\mathrm{\&mu;}}_{02})}^2+4{\mathrm{\&mu;}}_{11}^2}}{{\mathrm{\&mu;}}_{00}}\right)}^{\frac{1}{2}}---\left(4\right)$

$b={\left(\frac{2[{\mathrm{\&mu;}}_{20}+{\mathrm{\&mu;}}_{02}-\sqrt{{({\mathrm{\&mu;}}_{20}-{\mathrm{\&mu;}}_{02})}^2+4{\mathrm{\&mu;}}_{11}^2}}{{\mathrm{\&mu;}}_{00}}\right)}^{\frac{1}{2}}---\left(5\right)$

Wherein, μ ₂₀, μ ₀₂, μ ₁₁pass through p, q=0,1,2 determine, x, y are respectively horizontal ordinate, the ordinate of pixel center point, f (x, y) for falling into the intersection point number of pixel, for horizontal stroke, the ordinate of the weighting center of gravity of pixel, a, b represent major axis and the minor axis of described ellipse respectively, μ _pqfor two-dimensional geometry center square;

The total number of described intersection point can be determined according to following formula (6), wherein, μ ₀₀represent the total number of described intersection point;

μ ₀₀＝∑∑f(x,y) (6)

Wherein, μ ₀₀represent the total number of described intersection point, from formula (6), it is p, q two-dimensional geometry center square when being zero;

Step S204: the total gray-scale value determining each described fuzzy solution according to the ratio of the total number of described intersection point and described area, particularly, can be determined by following formula (7);

$I=\frac{{\mathrm{\&mu;}}_{00}}{\mathrm{\&pi;ab}}---\left(7\right)$

Wherein, I represents total gray-scale value, and π represents circular constant;

Step S102: each described fuzzy solution and numerical value wind field data are contrasted, finds out objective fuzzy solution immediate with numerical value wind field data;

Wherein, with numerical value wind field data closest to generally referring to that the wind direction value in fuzzy solution corresponding wind direction value and numerical value wind field data is closest, using such fuzzy solution as objective fuzzy solution;

Step S103: judge whether total gray-scale value of described objective fuzzy solution is greater than default gray threshold, if so, enters step S103;

Gray threshold in the present embodiment can set according to actual needs, wherein in an embodiment, be that gray threshold is set as 3, be of value to raising screening effect, judge whether total gray-scale value of described objective fuzzy solution is greater than the principle of default gray threshold as follows:

From the definition (7) of total gray scale, denominator is ellipse area, molecule is the total number of intersection point in intersection point close quarters, and therefore, this index of total gray scale reflects the dense degree of intersection point distribution, understand from angle qualitatively, observational error is larger, and the curve representated by each observed reading disperses more open, and intersection point is also more sparse, total gray scale is correspondingly less, and corresponding inversion error is likely larger; In order to obtain the relation between total gray scale of intersection point close quarters and inversion error, adopt the L2A data (using buoy wind arrow value as actual value) with buoy registration, verify after its inverting, result of calculation is listed in Fig. 6.Wherein, horizontal ordinate is total gray scale, and ordinate is wind vector retrieval error (6-a figure ordinate is wind direction inversion error, and 6-b figure ordinate is wind speed retrieval error).

As can be seen from Figure 6, the point that inversion error is larger all concentrates on the smaller region of total gray scale substantially, be greater than after 3 in total gray scale, the distribution of wind vector retrieval error is substantially just relatively steady, is limited in (wind direction error is approximately limited within 20 ° as shown in Figure 6) within certain scope.Therefore, total gray scale is greater than to the picture dot of some threshold values, can concludes that its inversion error is smaller, namely the type that fuzzy solution identification is correct is belonged to, this means for this class picture dot, by means of adjacent pixel, only inherently can not can judge its error degree from single picture dot;

Step S104 a: mark is done to described picture dot;

Total gray-scale value of objective fuzzy solution is greater than gray threshold and makes a mark, then mark whether can be had to judge whether it is the picture dot that inversion error is less according to picture dot.

Accordingly, according to the scheme of above-mentioned the present embodiment, it first determines each fuzzy solution of picture dot and corresponding total gray-scale value thereof, objective fuzzy solution is obtained again by each fuzzy solution and numerical value wind field data are carried out contrast, judge whether total gray-scale value of this objective fuzzy solution is greater than default gray threshold, if, object meta makes a mark, each picture dot in SeaWinds scatterometer wind vector retrieval can screen in this manner, the markd picture dot finally obtained is the less picture dot of the inversion error that filters out, circle median filter can be carried out based on these picture dots, because screening process effectively can reject the larger picture dot of wind vector retrieval error, thus improve the precision of circle median filter algorithm largely, namely, improve final wind vector retrieval precision.

According to the picture dot screening technique of the SeaWinds scatterometer wind vector retrieval of the invention described above, the present invention also provides a kind of filtering method of SeaWinds scatterometer wind vector retrieval.

Shown in Figure 7, be the schematic flow sheet of the filtering method embodiment of SeaWinds scatterometer wind vector retrieval of the present invention.As shown in Figure 3, the filtering method of the SeaWinds scatterometer wind vector retrieval of the present embodiment comprises the steps:

Step S301: the picture dot in SeaWinds scatterometer wind vector retrieval is screened;

The picture dot screening technique of the SeaWinds scatterometer wind vector retrieval in above-described embodiment can be adopted to screen the picture dot in SeaWinds scatterometer wind vector retrieval, do not repeat them here;

Step S302: choose filter window, obtains the markd picture dot number in this filter window;

The size of filter window can set according to actual conditions, and such as 7 × 7;

Step S303: judge whether described picture dot number is greater than zero, if so, then enters step S304, if not, then enters step S305;

Step S304: the default account form corresponding according to described picture dot number calculates the wind vector of described filter window;

The picture dot occurrence probability very large due to inversion error under high wind speed situation is larger, in whole filter window, total gray-scale value is greater than the number also instability of the picture dot of the gray threshold of setting, may be 0, may be 1, also may be 2, the account form corresponding according to different picture dot numbers calculates the wind vector of filter window, also improves final wind vector retrieval precision;

In one embodiment, the step that the described default account form corresponding according to described picture dot number calculates the wind vector of described filter window can comprise:

If described picture dot number is greater than two, then circle median filter is carried out to the markd picture dot in filter window, obtain the wind vector of described filter window;

If described picture dot number equals two, then corresponding to the markd picture dot in filter window wind vector be averaged value process, obtain the wind vector of described filter window;

If described picture dot number equals one, be then the wind vector of described filter window with the wind vector that the markd picture dot in filter window is corresponding.

Step S305: according to the wind vector of the described filter window of the acquisition of L2B data;

Wherein, L2B is wind vector standardized product, is that the businessization that current precision is the highest runs wind vector product, according to, wherein, L2B data are the survey wind data of L2B wind vector standardized product.

Circulation performs above-mentioned steps S302 ~ S304, until no longer change.

Accordingly, according to the scheme of above-described embodiment, it first adopts the picture dot screening technique of SeaWinds scatterometer wind vector retrieval as above to screen the picture dot in SeaWinds scatterometer wind vector retrieval under high wind speed situation, choose filter window again, obtain the markd picture dot number in this filter window, judge whether described picture dot number is greater than zero, if, then corresponding according to described picture dot number default account form calculates the wind vector of described filter window, if not, then according to the wind vector of the described filter window of the acquisition of L2B data, the picture dot that the wind vector retrieval error can effectively rejected in filter window due to screening process is larger, thus improve the precision of circle median filter algorithm largely, simultaneously, the picture dot occurrence probability very large due to inversion error under high wind speed situation is larger, in whole filter window, total gray-scale value is greater than the number also instability of the picture dot of the gray threshold of setting, may be 0, may be 1, also may be 2, the account form corresponding according to different picture dot numbers calculates the wind vector of filter window, also improves final wind vector retrieval precision.

In order to verify the validity of scheme in the present embodiment, after processing adopting the filtering method of SeaWinds scatterometer wind vector retrieval, list in table 1 with the comparing result of L2B product.Can be seen by table 1: (1) is (row 1-76) in whole observation scope, the inversion accuracy of two kinds of algorithms relatively, the wind direction inversion accuracy of new algorithm slightly improves, meanwhile, its fuzzy solution identification error (wind direction inversion error is greater than 45 degree) number has obvious decline (192-153); (2) when observation scope narrows down to region gradually, the inversion error of two kinds of algorithms is all in rising, but new algorithm increases gradually relative to the advantage of L2B product again, immediately below region (row 37-40) time, the wind direction inversion accuracy of new algorithm improves about 5 degree than L2B product, and the ratio that fuzzy solution identification error number accounts for total picture dot number drops to 7.04% by 11.73%.

The comparison of table 1. grayscale restraint filtering algorithm and L2B product

According to the picture dot screening technique of the SeaWinds scatterometer wind vector retrieval of the invention described above, the present invention also provides a kind of picture dot screening system of SeaWinds scatterometer wind vector retrieval, and just the embodiment of the picture dot screening system of SeaWinds scatterometer wind vector retrieval of the present invention is described in detail below.The structural representation of the embodiment of the picture dot screening system of SeaWinds scatterometer wind vector retrieval of the present invention has been shown in Fig. 8.For convenience of explanation, part related to the present invention is merely illustrated in fig. 8.

As shown in Figure 8, the picture dot screening system of the SeaWinds scatterometer wind vector retrieval in the present embodiment, comprises processing module 401, contrast module 402, first judge module 403, mark module 404, wherein:

Processing module 401, for total gray-scale value of each fuzzy solution and each described fuzzy solution of determining the picture dot in SeaWinds scatterometer wind vector retrieval;

Contrast module 402, for each described fuzzy solution and numerical value wind field data being contrasted, finds out objective fuzzy solution immediate with numerical value wind field data;

First judge module 403, for judging whether total gray-scale value of described objective fuzzy solution is greater than default gray threshold;

Mark module 404, for when the judged result of the first judge module 403 is for being, makes a mark to described picture dot.

Wherein in an embodiment, as shown in Figure 9, processing module 401 can comprise:

Acquiring unit 501, for obtaining each intersection point close quarters of observed reading in wind vector solution space of described picture dot;

Division unit 502, for carrying out pixel division to each described intersection point close quarters respectively, and determines the intersection point number falling into each pixel;

Processing unit 503, for determining the total number of intersection point in the area of the ellipse that each described intersection point close quarters is corresponding and each described intersection point close quarters according to described intersection point number respectively, wherein, described ellipse, for including the intersection point in described intersection point close quarters, determines total gray-scale value of each described fuzzy solution according to the ratio of the total number of described intersection point and described area.

Wherein in an embodiment, processing module 401 can basis determine the major axis of described ellipse, according to determine the minor axis of described ellipse, and determine described area, according to μ according to described major axis and described minor axis ₀₀=∑ ∑ f (x, y) determines the total number of described intersection point;

Wherein, μ ₂₀, μ ₀₂, μ ₁₁pass through p, q=0,1,2 determine, x, y are respectively horizontal ordinate, the ordinate of pixel center point, f (x, y) for falling into the intersection point number of pixel, for horizontal stroke, the ordinate of the weighting center of gravity of pixel, a, b represent major axis and the minor axis of described ellipse respectively, μ ₀₀represent the total number of described intersection point.

Wherein in an embodiment, the gray threshold in above-described embodiment can be 3.

The picture dot screening system of SeaWinds scatterometer wind vector retrieval of the present invention and the picture dot screening technique one_to_one corresponding of SeaWinds scatterometer wind vector retrieval of the present invention, the technical characteristic of setting forth in the embodiment of the picture dot screening technique of above-mentioned SeaWinds scatterometer wind vector retrieval and beneficial effect thereof are all applicable to, in the embodiment of the picture dot screening system of SeaWinds scatterometer wind vector retrieval, hereby state.

According to the filtering method of the SeaWinds scatterometer wind vector retrieval of the invention described above, the present invention also provides a kind of filtering system of SeaWinds scatterometer wind vector retrieval, and just the embodiment of the filtering system of SeaWinds scatterometer wind vector retrieval of the present invention is described in detail below.The structural representation of the embodiment of the filtering system of SeaWinds scatterometer wind vector retrieval of the present invention has been shown in Figure 10.For convenience of explanation, part related to the present invention is merely illustrated in Fig. 10.

As shown in Figure 10, the filtering system of the SeaWinds scatterometer wind vector retrieval in the present embodiment, comprise the picture dot screening system 601 of the SeaWinds scatterometer wind vector retrieval in any one embodiment above-mentioned, also comprise and choose module 602, second judge module 603, filtration module 604, wherein:

Choosing module 602, for choosing filter window, obtaining the markd picture dot number in this filter window;

Second judge module 603, for judging whether described picture dot number is greater than zero;

Filtration module 604, for when the judged result of the second judge module 603 is for being, the default account form corresponding according to described picture dot number calculates the wind vector of described filter window, when the judged result of the second judge module 603 is no, according to the wind vector of the described filter window of the acquisition of L2B data.

In one embodiment, described filtration module 604 performs the process that the described default account form corresponding according to described picture dot number calculate the wind vector of described filter window and can be further used for:

The step that the described account form preset corresponding according to described picture dot number calculates the wind vector of described filter window comprises:

If described picture dot number is greater than two, then circle median filter is carried out to the markd picture dot in filter window, obtain the wind vector of described filter window;

If described picture dot number equals two, then corresponding to the markd picture dot in filter window wind vector be averaged value process, obtain the wind vector of described filter window;

If described picture dot number equals one, be then the wind vector of described filter window with the wind vector that the markd picture dot in filter window is corresponding.

The filtering system of SeaWinds scatterometer wind vector retrieval of the present invention and the filtering method one_to_one corresponding of SeaWinds scatterometer wind vector retrieval of the present invention, the technical characteristic of setting forth in the embodiment of the filtering method of above-mentioned SeaWinds scatterometer wind vector retrieval and beneficial effect thereof are all applicable to, in the embodiment of the filtering system of SeaWinds scatterometer wind vector retrieval, hereby state.

The above embodiment only have expressed several embodiment of the present invention, and it describes comparatively concrete and detailed, but therefore can not be interpreted as the restriction to the scope of the claims of the present invention.It should be pointed out that for the person of ordinary skill of the art, without departing from the inventive concept of the premise, can also make some distortion and improvement, these all belong to protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be as the criterion with claims.

Claims (10)

1. a picture dot screening technique for SeaWinds scatterometer wind vector retrieval, is characterized in that, comprise the steps:

Determine each fuzzy solution of the picture dot in SeaWinds scatterometer wind vector retrieval and total gray-scale value of each described fuzzy solution;

Each described fuzzy solution and numerical value wind field data are contrasted, finds out objective fuzzy solution immediate with numerical value wind field data;

Judge whether total gray-scale value of described objective fuzzy solution is greater than default gray threshold;

If so, a mark is done to described picture dot.

2. the picture dot screening technique of SeaWinds scatterometer wind vector retrieval according to claim 1, is characterized in that, determines that the mode of total gray-scale value of fuzzy solution comprises the steps:

Obtain each intersection point close quarters of observed reading in wind vector solution space of described picture dot;

Respectively pixel division is carried out to each described intersection point close quarters, and determine the intersection point number falling into each pixel;

Determine the total number of intersection point in the area of the ellipse that each described intersection point close quarters is corresponding and each described intersection point close quarters according to described intersection point number respectively, wherein, described ellipse is for including the intersection point in described intersection point close quarters;

Total gray-scale value of each described fuzzy solution is determined according to the ratio of the total number of described intersection point and described area.

3. the picture dot screening technique of SeaWinds scatterometer wind vector retrieval according to claim 2, it is characterized in that, the described step determining the area of ellipse corresponding to described intersection point close quarters and the total number of intersection point of described intersection point close quarters according to described intersection point number comprises:

According to $a={\left(\frac{2[{\mathrm{\&mu;}}_{20}+{\mathrm{\&mu;}}_{02}+\sqrt{{({\mathrm{\&mu;}}_{20}-{\mathrm{\&mu;}}_{02})}^2+4{\mathrm{\&mu;}}_{11}^2}}{{\mathrm{\&mu;}}_{00}}\right)}^{\frac{1}{2}},b={\left(\frac{2[{\mathrm{\&mu;}}_{20}+{\mathrm{\&mu;}}_{02}-\sqrt{{({\mathrm{\&mu;}}_{20}-{\mathrm{\&mu;}}_{02})}^2+4{\mathrm{\&mu;}}_{11}^2}}{{\mathrm{\&mu;}}_{00}}\right)}^{\frac{1}{2}}$ Determine major axis and the minor axis of described ellipse, and determine described area according to described major axis and described minor axis;

According to μ ₀₀=Σ Σ f (x, y) determines the total number of described intersection point;

Wherein, μ ₂₀, μ ₀₂, μ ₁₁pass through p, q=0,1,2 determine, x, y are respectively horizontal ordinate, the ordinate of pixel center point, f (x, y) for falling into the intersection point number of pixel, for horizontal stroke, the ordinate of the weighting center of gravity of pixel, a, b represent major axis and the minor axis of described ellipse respectively, μ ₀₀represent the total number of described intersection point.

4. a filtering method for SeaWinds scatterometer wind vector retrieval, is characterized in that, comprises the steps:

The method as described in one of claims 1 to 3 is adopted to screen the picture dot in SeaWinds scatterometer wind vector retrieval;

Choose filter window, obtain the markd picture dot number in this filter window;

Judge whether described picture dot number is greater than zero;

If so, then corresponding according to described picture dot number default account form calculates the wind vector of described filter window;

If not, then according to the wind vector of the described filter window of the acquisition of L2B data.

5. the filtering method of SeaWinds scatterometer wind vector retrieval according to claim 4, is characterized in that, the step that the described default account form corresponding according to described picture dot number calculates the wind vector of described filter window comprises:

If described picture dot number is greater than two, then circle median filter is carried out to the markd picture dot in filter window, obtain the wind vector of described filter window;

If described picture dot number equals two, then corresponding to the markd picture dot in filter window wind vector be averaged value process, obtain the wind vector of described filter window;

If described picture dot number equals one, be then the wind vector of described filter window with the wind vector that the markd picture dot in filter window is corresponding.

6. a picture dot screening system for SeaWinds scatterometer wind vector retrieval, is characterized in that, comprising:

Processing module, for total gray-scale value of each fuzzy solution and each described fuzzy solution of determining the picture dot in SeaWinds scatterometer wind vector retrieval;

Contrast module, for each described fuzzy solution and numerical value wind field data being contrasted, finds out objective fuzzy solution immediate with numerical value wind field data;

First judge module, for judging whether total gray-scale value of described objective fuzzy solution is greater than default gray threshold;

Mark module, for when the judged result of described first judge module is for being, makes a mark to described picture dot.

7. the picture dot screening system of SeaWinds scatterometer wind vector retrieval according to claim 6, it is characterized in that, described processing module comprises:

Acquiring unit, for obtaining each intersection point close quarters of observed reading in wind vector solution space of described picture dot;

Division unit, for carrying out pixel division to each described intersection point close quarters respectively, and determines the intersection point number falling into each pixel;

Processing unit, for determining the total number of intersection point in the area of the ellipse that each described intersection point close quarters is corresponding and each described intersection point close quarters according to described intersection point number respectively, wherein, described ellipse, for including the intersection point in described intersection point close quarters, determines total gray-scale value of each described fuzzy solution according to the ratio of the total number of described intersection point and described area.

8. the picture dot screening system of SeaWinds scatterometer wind vector retrieval according to claim 7, is characterized in that:

Described processing unit according to determine the major axis of described ellipse, according to determine the minor axis of described ellipse, and determine described area, according to μ according to described major axis and described minor axis ₀₀=Σ Σ f (x, y) determines the total number of described intersection point;

Wherein, μ ₂₀, μ ₀₂, μ ₁₁pass through p, q=0,1,2 determine, x, y are respectively horizontal ordinate, the ordinate of pixel center point, f (x, y) for falling into the intersection point number of pixel, for horizontal stroke, the ordinate of the weighting center of gravity of pixel, a, b represent major axis and the minor axis of described ellipse respectively, μ ₀₀represent the total number of described intersection point.

9. a filtering system for SeaWinds scatterometer wind vector retrieval, is characterized in that, comprises the picture dot screening system of the SeaWinds scatterometer wind vector retrieval as described in one of claim 6 to 8, also comprises:

Choosing module, for choosing filter window, obtaining the markd picture dot number in this filter window;

Second judge module, for judging whether described picture dot number is greater than zero;

Filtration module, for when the judged result of described second judge module is for being, the default account form corresponding according to described picture dot number calculates the wind vector of described filter window, when the judged result of described second judge module is no, according to the wind vector of the described filter window of the acquisition of L2B data.

10. the filtering system of SeaWinds scatterometer wind vector retrieval according to claim 9, it is characterized in that, the process that the described default account form corresponding according to described picture dot number of described filtration module execution calculates the wind vector of described filter window is further used for:

The step that the described account form preset corresponding according to described picture dot number calculates the wind vector of described filter window comprises:

If described picture dot number is greater than two, then circle median filter is carried out to the markd picture dot in filter window, obtain the wind vector of described filter window;

If described picture dot number equals two, then corresponding to the markd picture dot in filter window wind vector be averaged value process, obtain the wind vector of described filter window;

If described picture dot number equals one, be then the wind vector of described filter window with the wind vector that the markd picture dot in filter window is corresponding.