CN106096545A - Method based on the hail cloud recognition estimating excavation - Google Patents

Abstract

Technology based on the hail cloud recognition estimating excavation belongs to field of statistics, especially belongs to art of image analysis.The existing prediction to hail cloud depends on experience often and differentiates, such method exists a lot of not enough.In order to overcome the deficiencies in the prior art, the present invention is open: a kind of method predicting hail, it is characterized in that, described method comprises the following steps: the first step, utilize cloud atlas to obtain dBZ, second step, RGB image is converted into gray level image, 3rd step, the reflectance of gray level image is divided into 15 magnitudes, the 4th step, calculates the proportion of each magnitude, 5th step, calculates average x of each magnitude₁, variance x₂, degree of bias x₃, kurtosis x₄, angle second moment x₅, entropy x₆, the moment of inertia x₇, dependency x₈, the 6th step, utilize formula: y=879.83840 x₁+186.90265·x₂‑1335·x₃+770.04074·x₄+2145·x₅+59.31601·x₆‑13.01776·x₇‑6022·x₈y₀If=158.23206 thus have: y ＞ y₀, then X ∈ G₁, otherwise X ∈ G₂, G₁Conjunction, G is converged for hail shooting₂For converging conjunction without hail.The present invention can pre-hail detection cloud.

Description

Method based on the hail cloud recognition estimating excavation

Technical field

The invention belongs to field of statistics, especially belong to art of image analysis.

Background technology

Hail, as a kind of strong convective weather, is characterized in that space scale is little, life cycle is short, sudden strong, development evolvement Rapidly, its forecast difficulty is well-known.Through " 15 ", the development of Eleventh Five-Year Plan and construction, Xinjiang meteorology hail suppression technology Level has had the biggest lifting, but compared with the demand that national economy and social development is growing, however it remains necessarily not enough, main Accuracy rate and the level of becoming more meticulous of hail weather to be showed themselves in that prediction have much room for improvement, and particularly close on local hail weather (0～3 hour) and in short-term (3～12 hours) prediction ability urgently strengthen." 13 " period is Xinjiang social stability and economy The critical period of development, in the urgent need to providing the monitoring and forecast more become more meticulous to Xinjiang hail weather.And to hail cloud recognition Service basic is mainly derived from weather radar, the radar image of real-time monitored Convective Cloud, and image covers cloud cluster life development The information of change, by processing the radar image of this cloud cluster in real time, whether hail shooting is for anti-for look-ahead to this cloud cluster Hail mitigation is significant.

The existing prediction to hail cloud depends on experience often and differentiates, such method exists a lot of not enough.

Summary of the invention

In order to overcome the deficiencies in the prior art, the present invention is open:

1, method based on the hail cloud recognition estimating excavation, it is characterized in that, described method comprises the following steps:

The first step, gathers radar RGB cloud atlas, utilizes cloud atlas to obtain dBZ,

Second step, is converted into gray level image the RGB image of first step gained,

3rd step, is divided into 15 magnitudes the reflectance of the gray level image of second step gained,

4th step, calculates each magnitude proportion in the picture,

5th step, calculates average x of each magnitude₁, variance x₂, degree of bias x₃, kurtosis x₄, angle second moment x₅, entropy x₆, inertia Square x₇, dependency x₈,

6th step, average x of the 5th step gained₁, variance x₂, degree of bias x₃, kurtosis x₄, angle second moment x₅, entropy x₆, the moment of inertia x₇, dependency x₈:

Y=879.83840 x₁+186.90265·x₂-1335·x₃+770.04074·x₄+2145·x₅+ 59.31601·x₆-13.01776·x₇-6022·x₈

If therefore have: y ＞ y₀Then X ∈ G₁, otherwise X ∈ G₂。G₁Conjunction, G is converged for hail shooting₂For converging conjunction without hail.

2, it is characterized in that: described y₀=158.23206

3, it is characterized in that: also include:

Utilize formula:

$y_0=\frac{n_1{\stackrel{\‾}{y}}^{\left(1\right)}+n_2{\stackrel{\‾}{y}}^{\left(2\right)}}{n_1+n_2}$

Obtain y₀。

Hail shooting converges conjunction G₁In meansigma methods,Hail shooting converges conjunction G₂In meansigma methods, n₁Conjunction G is converged for hail shooting₁In Element number, n₂For converging conjunction G without hail₂In element number.

Reflectance separates for unit according to 5, dBZ value s₁Corresponding gray value is calculated as s₁-2.5 arrive s₁The gray scale of+2.5 is put down Average.

4, it is characterized in that: described reflectance, magnitude, rgb value, relation that gray value is corresponding be:

Reflectance is :-5, and corresponding magnitude is: 1, and corresponding rgb value is :-201,201,201, and corresponding gray value is: 201；

Reflectance is: 0, and corresponding magnitude is: 2, and corresponding rgb value is :-118,118,118, and corresponding gray value is: 118；

Reflectance is: 5, and corresponding magnitude is: 3, and corresponding rgb value is :-255,170,170, and corresponding gray value is: 196；

Reflectance is: 10, and corresponding magnitude is: 4, and corresponding rgb value is :-238,140,140, and corresponding gray value is: 170；

Reflectance is: 15, and corresponding magnitude is: 5, and corresponding rgb value is :-201,112,112, and corresponding gray value is: 139；

Reflectance is: 20, and corresponding magnitude is: 6, and corresponding rgb value is: 0,251,144, and corresponding gray value is: 164；

Reflectance is: 25, and corresponding magnitude is: 7, and corresponding rgb value is: 0,187,0, and corresponding gray value is: 110；

Reflectance is: 30, and corresponding magnitude is: 8, and corresponding rgb value is :-255,255,112, and corresponding gray value is: 239；

Reflectance is: 35, and corresponding magnitude is: 9, and corresponding rgb value is: 208,208,96, and corresponding gray value is: 195；

Reflectance is: 40, and corresponding magnitude is: 10, and corresponding rgb value is: 255,96,96, and corresponding gray value is: 144；

Reflectance is: 45, and corresponding magnitude is: 11, and corresponding rgb value is: 218,0,0, and corresponding gray value is: 65；

Reflectance is: 50, and corresponding magnitude is 12, and corresponding rgb value is: 174,0,0, and corresponding gray value is: 52；

Reflectance is: 55, and corresponding magnitude is: 13, and corresponding rgb value is: 0,0,255, and corresponding gray value is: 29；

Reflectance is: 60, and corresponding magnitude is: 14, and corresponding rgb value is :-160,255,255, corresponding gray value For: 227；

Reflectance is: 65, and corresponding magnitude is: 15, and corresponding rgb value is: 231,0,255, and corresponding gray value is: 98。

5, it is characterized in that: by the conversion relation formula of RGB Yu gray value

Gray=R × 0.2999+G × 0.587+B × 0.114

Obtain corresponding gray value.The method can improve the accuracy rate of hail cloud forecast, reduces rate of false alarm.Effective raising Identification hail cloud atlas, as accuracy rate, provides reference for Forecast of Hail.

Detailed description of the invention:

1 intensity statistics

Radar reflectivity image is based on RGB color space, based on the legend of base reflectivity figure, utilizes The color production theory of RGB color space, reads echo reflection figure.Will reflectance be-5dBZ 65dBZ, be divided into 15 kinds accordingly Magnitude.Wherein it is as shown in the table, by the conversion relation formula (1) of RGB Yu gray value for different rgb values corresponding to magnitude, it is known that corresponding Gray value.

Gray=R × 0.2999+G × 0.587+B × 0.114 (1)

Table 1 is used for reflectance and rgb value, gray value corresponding relation are described.

Table 1:

What the different colours of radar reflectivity image represented is different reflex strengths, i.e. can be by statistics different colours Proportion shared in the picture, reflects the feature of echo strength in image, and the different proportion shared by intensity can be by probability Represent, calculate as shown in formula (2).If a width size is the image of M × N, then containing M × N number of pixel, reflectivity intensity is divided into 15 Individual magnitude, this varying strength proportion in the picture is:

p_i=s_i/ M × N i=1,2 ..., 15 (2)

Wherein, s_iFor the color of the i-th magnitude number of pixels on image.

2, the modelling of eight rank hail cloud recognition

According to meteorological theory, the extreme weather of strong convection, it is only possible to occur in reflex strength more than 40dBZ's Region.Therefore only choosing magnitude is 10 15, eight first-order statistics being calculated hail cloud and non-hail cloud in accordance with the following methods are surveyed Angle value.

Average:

$E=\underset{i=1}{\overset{17}{\Σ}}i\×p_i$

Variance:

$D=\underset{i=1}{\overset{17}{\Σ}}{(i-mean)}^2\·p_i$

The degree of bias:

$skewness=\frac{1}{{\left(\mathrm{var}iance\right)}^{3/2}}\·\underset{i=1}{\overset{17}{\Σ}}{(i-mean)}^3\·p_i$

Kurtosis:

$kurtosis=\frac{1}{{\left(\mathrm{var}iance\right)}^2}\·\underset{i=1}{\overset{17}{\Σ}}{(i-mean)}^4\·p_i$

Dependency:

$u_i=\underset{i=1}{\overset{k}{\Σ}}\underset{j=1}{\overset{k}{\Σ}}i\·G(i,j)$

$u_j=\underset{i=1}{\overset{k}{\Σ}}\underset{j=1}{\overset{k}{\Σ}}j\·G(i,j)$

$s_i^2=\underset{i=1}{\overset{k}{\Σ}}\underset{j=1}{\overset{k}{\Σ}}G(i,j){(i-u_i)}^2$

$s_j^2=\underset{i=1}{\overset{k}{\Σ}}\underset{j=1}{\overset{k}{\Σ}}G(i,j){(j-u_j)}^2$

The moment of inertia:

Angle direction second moment:

$ASM=\underset{i}{\Σ}{\[p_{\mathrm{\Δ}}\left(i\right)\]}^2$

Entropy:

$ENT=-\underset{i}{\Σ}{p}_{\mathrm{\Δ}}\left(i\right)\lg p_{\mathrm{\Δ}}\left(i\right)$

Determine differentiation critical point, take y₀ForWithWeighted mean, it may be assumed that

$y_0=\frac{n_1{\stackrel{\‾}{y}}^{\left(1\right)}+n_2{\stackrel{\‾}{y}}^{\left(2\right)}}{n_1+n_2}$

Hail shooting converges conjunction G₁In meansigma methods,Hail shooting converges conjunction G₂In meansigma methods, n₁Conjunction G is converged for hail shooting₁In Element number, n₂For converging conjunction G without hail₂In element number.

Discriminant function is:

Y=879.83840 x₁+186.90265·x₂-1335·x₃+770.04074·x₄+2145·x₅+ 59.31601·x₆-13.01776·x₇-6022·x₈ (3)

Differentiate that critical point is: y₀=158.23206, andIf therefore have: y ＞ y₀Then X ∈ G₁, otherwise X ∈ G₂。G₁ Conjunction, G is converged for hail shooting₂For converging conjunction without hail.

3, model testing

Utilizing method of the present invention, inventor has carried out following checking, to prove that prediction hail is had relatively by the present invention High accuracy rate.

Data decimation is Shihezi, area, Shawan and the data being different from training sample of Aksu Prefecture between 2009, make For sample to be tested table 2.Owing to the classification of sample to be tested is it is known that then utilize this discriminant function to differentiate, and will differentiate result with Legitimate reading compares, and i.e. may know that the accuracy rate that discriminant function differentiates for albedo image.Result is as shown in table 3.

The moment of inertia dependency-the sample to be tested of the albedo image intensity of table 2 each department

Table 3 differentiates result

From table 2 and table 3, the differentiation accuracy rate of this discrimination model is fine, and this model is ratio for the identification of hail cloud Accurate, accuracy rate is 100%.

Claims (5)

1. method based on the hail cloud recognition estimating excavation, is characterized in that, described method comprises the following steps:

The first step, gathers radar RGB cloud atlas, utilizes cloud atlas to obtain dBZ,

Second step, is converted into gray level image the RGB image of first step gained,

3rd step, is divided into 15 magnitudes the reflectance of the gray level image of second step gained,

4th step, calculates each magnitude proportion in the picture,

5th step, calculates average x of each magnitude₁, variance x₂, degree of bias x₃, kurtosis x₄, angle second moment x₅, entropy x₆, the moment of inertia x₇、 Dependency x₈,

6th step, average x of the 5th step gained₁, variance x₂, degree of bias x₃, kurtosis x₄, angle second moment x₅, entropy x₆, the moment of inertia x₇, phase Closing property x₈:

Y=879.83840 x₁+186.90265·x₂-1335·x₃+770.04074·x₄+2145·x₅+59.31601· x₆-13.01776·x₇-6022·x₈

If therefore have: y ＞ y₀, then X ∈ G₁Otherwise X ∈ G₂, G₁Conjunction, G is converged for hail shooting₂For converging conjunction without hail.

2. method based on the hail cloud recognition estimating excavation as claimed in claim 1, is characterized in that: described y₀=158.23206.

3. method based on the hail cloud recognition estimating excavation as claimed in claim 1, is characterized in that: also include:

Utilize formula:

Hail shooting converges conjunction G₁In meansigma methods,Hail shooting converges conjunction G₂In meansigma methods, n₁Conjunction G is converged for hail shooting₁In unit Element number, n₂For converging conjunction G without hail₂In element number,

Obtain y0.

4. method based on the hail cloud recognition estimating excavation as claimed in claim 1, is characterized in that: described reflectance, amount The relation that level, rgb value, gray value are corresponding is:

Reflectance is :-5, and corresponding magnitude is: 1, and corresponding rgb value is :-201,201,201, and corresponding gray value is: 201；

Reflectance is: 0, and corresponding magnitude is: 2, and corresponding rgb value is :-118,118,118, and corresponding gray value is: 118；

Reflectance is: 5, and corresponding magnitude is: 3, and corresponding rgb value is :-255,170,170, and corresponding gray value is: 196；

Reflectance is: 10, and corresponding magnitude is: 4, and corresponding rgb value is :-238,140,140, and corresponding gray value is: 170；

Reflectance is: 15, and corresponding magnitude is: 5, and corresponding rgb value is :-201,112,112, and corresponding gray value is: 139；

Reflectance is: 20, and corresponding magnitude is: 6, and corresponding rgb value is: 0,251,144, and corresponding gray value is: 164；

Reflectance is: 25, and corresponding magnitude is: 7, and corresponding rgb value is: 0,187,0, and corresponding gray value is: 110；

Reflectance is: 30, and corresponding magnitude is: 8, and corresponding rgb value is :-255,255,112, and corresponding gray value is: 239；

Reflectance is: 35, and corresponding magnitude is: 9, and corresponding rgb value is: 208,208,96, and corresponding gray value is: 195；

Reflectance is: 40, and corresponding magnitude is: 10, and corresponding rgb value is: 255,96,96, and corresponding gray value is: 144；

Reflectance is: 45, and corresponding magnitude is: 11, and corresponding rgb value is: 218,0,0, and corresponding gray value is: 65；

Reflectance is: 50, and corresponding magnitude is 12, and corresponding rgb value is: 174,0,0, and corresponding gray value is: 52；

Reflectance is: 55, and corresponding magnitude is: 13, and corresponding rgb value is: 0,0,255, and corresponding gray value is: 29；

Reflectance is: 60, and corresponding magnitude is: 14, and corresponding rgb value is :-160,255,255, and corresponding gray value is: 227；

Reflectance is: 65, and corresponding magnitude is: 15, and corresponding rgb value is: 231,0,255, and corresponding gray value is: 98.

5. method based on the hail cloud recognition estimating excavation as claimed in claim 1, is characterized in that: by RGB and gray value Conversion relation formula

Gray=R × 0.2999+G × 0.587+B × 0.114, obtains corresponding gray value.