CN104537263A - Multi-resolution atmosphere data modeling method based on wavelet analysis - Google Patents

Abstract

The invention discloses a multi-resolution atmosphere data modeling method based on wavelet analysis. The multi-resolution atmosphere data modeling method includes the following steps: (1) obtaining an atmosphere environment area required by an air vehicle; (2) selecting a wavelet function and a boundary extension mode; (3) building a wavelet transformation primary function of two-dimensional space according to the wavelet function selected in the step (2); (4) carrying out wavelet transformation on atmosphere environment data; (5) determining a wavelet coefficient screening threshold value; (6) adding a detailed part of the wavelet coefficient to carry out wavelet inverse transformation, and generating the multi-resolution atmosphere environment data model. By means of the multi-resolution atmosphere data modeling method, multi-resolution atmosphere data modeling is effectively and rapidly achieved, the device storage efficiency is improved, and the operation efficiency is improved.

Description

Based on the multiresolution atmosphere data modeling method of wavelet analysis

Technical field

The invention belongs to aeronautical meteorology field of environmental technology, particularly relate to atmosphere data modeling technique field in aeronautical meteorology environment, specifically a kind of multiresolution atmosphere data modeling method based on wavelet analysis.

Background technology

Terrible weather environment can have a strong impact on the safe flight of aircraft and task completes.Due to the computing power of airborne computer and the requirement of sensor performance, and the polytrope of situation of battlefield, rapidity is the important performance that military routeing needs to consider, and the method for expressing of atmosphere data in routeing directly can affect memory space and the calculated amount of airborne computer, in order to reduce storage and computation burden, design a kind of multiresolution atmosphere data modeling method significant to actual Military Application.

Aircraft is different to the resolution requirement of atmosphere data under different flying condition.When aircraft carries out low-level penetration or be in landing or takeoff phase near airfield runway, because flying speed is lower, be subject to the impact of atmospheric environment, then need the atmosphere data that resolution is higher, when aircraft moves in subdued topography overhead, the impact change of landform on Atmospheric Flow is less, then can adopt the atmosphere data that resolution is lower.The multi-resolution models of atmospheric environment depends on its time space field characteristic of field to a great extent.Consider from the angle of the atmospheric environment situation reflected realistically in specific region and special time period, if atmospheric environment is relatively more violent with spatial variations in time, be then suitable for the environmental data adopting high-resolution.But in actual applications, be difficult to the atmospheric environment data of Real-time Obtaining highest resolution on the one hand, on the other hand, adopt the atmospheric environment data of highest resolution to reduce counting yield always, under many circumstances may it is desirable that the atmospheric environment data of multiresolution level of comparatively macroscopic view, how while not affecting atmosphere data modeling quality, to carry out suitable simplification to data and can realize storing queries is efficiently the problem needing in practical application to solve.

Summary of the invention

The object of the invention is to solve the problem, for existing aircraft in flight course to the practical problems that the quick storage of atmospheric environment data is inquired about, a kind of multiresolution atmosphere data modeling method based on wavelet analysis is proposed, utilize wavelet analysis technology, extract magnanimity, multiple dimensioned atmosphere data collection essential characteristic and expressed by wavelet coefficient, just can obtain the expression of logarithm according to the multiresolution of collection after data are correspondingly processed, improve aircraft storing queries efficiency to atmospheric environment data in flight course.

Based on a multiresolution atmosphere data modeling method for wavelet analysis, comprise following step:

Step one: obtain the ambient area that aircraft needs;

Step 2: choose wavelet function and boundary extension mode;

Step 3: according to the wavelet function chosen in step 2, builds the wavelet transformation basis function of two-dimensional space;

Step 4: wavelet transformation is carried out to atmospheric environment data;

Step 5: determine wavelet coefficient screening threshold value;

Step 6: the detail section adding wavelet coefficient carries out wavelet inverse transformation, generates multiresolution atmospheric environment data model;

The invention has the advantages that:

(1) realize the multi-resolution Modeling of atmosphere data fast and effectively, the storage efficiency of lifting means, improves operation efficiency;

(2) provide the multiresolution macroscopic view of ambient area to express, realize the rapid build of different levels resolution model;

(3) improve aircraft storing queries efficiency to atmospheric environment data in flight course.

Accompanying drawing explanation

Fig. 1 is the multiresolution atmosphere data modeling procedure figure based on wavelet analysis;

Fig. 2 is initial temperature field design sketch;

Fig. 3 is one-level resolution temperature field design sketch;

Fig. 4 is secondary resolution temperature field design sketch.

Embodiment

Below in conjunction with drawings and Examples, the present invention is described in further detail.

The main thought of the inventive method is: by the data structure definition of atmospheric environment in a slice atmosphere zone D, select appropriate coordinate system and after stress and strain model being carried out to region by X-axis and Y-axis, sampled data on corresponding net point (X, Y) is corresponding atmospheric properties Z.The atmospheric environment data of regular grid can regard the signal field of a two-dimentional relevant atmospheric properties as, and in signal field, some regional atmospheric attribute change is comparatively violent, and some regional change is comparatively slow.After carrying out wavelet transformation to atmospheric environment data, extract high-frequency signal and low frequency signal, its high frequency signal is the detail section that atmospheric properties becomes soon, and low frequency signal is the aggregate level part that atmospheric properties becomes slowly.The atmospheric environment data model that suitable filtering can access multiresolution is carried out to the detail section after conversion.

The present invention is a kind of multiresolution atmosphere data modeling method based on wavelet analysis, and flow process as shown in Figure 1, comprises following step:

Step one: obtain the ambient area that aircraft needs.

Step 1.1: by the three dimensional environmental space region residing for aircraft, is decomposed into the two-dimentional Euclidean space of different levels from short transverse, the two-dimentional Euclidean space obtained is required ambient area;

Step 1.2: can flying height according to landform altitude and aircraft by the two-dimentional Euclidean space that obtains, is divided into barrier and non-barrier, and determines initial position and the target location of aircraft, to i-th position X in original figure landform altitude figure _ibe expressed as follows:

Wherein, represent Obstacles, represent non-Obstacles, z _irepresent position X _ilandform altitude, what H represented aircraft can flying height;

Step 1.3: environmental area is normalized to unit square region, and carry out stress and strain model.

Step 2: choose wavelet function and boundary extension mode.

Step 2.1: the mode selecting wavelet transformation according to the continuity of required transform data or the reconfigurability of discreteness type and conversion and easy implementation;

Step 2.2: different wavelet filtering functions has different conversion characteristicss, determines the variation that small echo under different application background is chosen.By each character of clear and definite small echo on converting the impact that brings in conjunction with the consideration of practical application, from existing a large amount of wavelet function, determine the scope of candidate's small echo;

Step 2.3: within the scope of fixed candidate, in conjunction with practical application request, adopt different wavelet functions and boundary extension mode that raw data is carried out to wavelet transformation and reconstruct inverse transformation, compared the error of raw data and reconstruct data and calculate reconstruction accuracy, choose most suitable wavelet function and boundary extension mode.

Step 3: according to the wavelet function chosen in step 2, builds the wavelet transformation basis function of two-dimensional space.

For Haar small echo, specifically comprise following step:

Step 3.1: the Haar wavelet choosing the one-dimensional space, obtains one group shape similar small echo by flexible with translation by Haar wavelet, obtains the Haar wavelet basis function of the one-dimensional space:

$\begin{array}{c}{\mathrm{\φ}}_i^j\left(x\right)=\sqrt{2^j}\mathrm{\φ}(2^{j}x-i),{\mathrm{\ψ}}_i^j\left(x\right)=\sqrt{2^j}\mathrm{\ψ}(2^{j}x-j)\\ j=\mathrm{0,1},...J_{\max }\mathrm{and\; i}=\mathrm{0,1},...,2^j-1\end{array}---\left(2\right)$

Wherein, j represents level of resolution, 2 ^jrepresent contraction-expansion factor, i represents translational movement, represent the Haar scaling function after flexible and translation and Haar mother wavelet function respectively, x represents the location variable of the one-dimensional space;

Step 3.2: the Haar wavelet basis function being constructed two-dimensional space by the Haar wavelet basis function of the one-dimensional space is as follows:

$\begin{array}{c}{\mathrm{\Φ}}_{m,n}^j(x,y)={\mathrm{\φ}}_m^j\left(x\right){\mathrm{\φ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{1,j}(x,y){=\mathrm{\φ}}_m^j\left(x\right){\mathrm{\ψ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{2,j}(x,y)={\mathrm{\ψ}}_m^j\left(x\right){\mathrm{\φ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{3,j}(x,y)={\mathrm{\ψ}}_m^j\left(x\right){\mathrm{\ψ}}_n^j\left(y\right)\end{array}---\left(3\right)$

Wherein, (x, y) represents the location variable of two-dimensional space, and m, n are the location label of two-dimensional space both direction; represent the Haar scaling function of two-dimensional space, represent the Haar mother wavelet function in n direction, represent the Haar mother wavelet function in m direction, represent the Haar mother wavelet function in diagonal.

Step 4: wavelet transformation is carried out to atmospheric environment data.

In described step 4, the method for wavelet transformation is carried out specifically to atmospheric environment data:

Step 4.1: read raw data and piecemeal process, select interested ambient area, is shown as effective lattice point (x in current region by atmospheric environment tables of data _i, y _i) corresponding data value z _idata layout;

Step 4.2: choose when suitable boundary extension mode carries out boundary extension process to prevent from carrying out wavelet transformation to above-mentioned atmospheric environment data in data boundary locations generation distortion;

Step 4.3: realized by the tensor product of one dimension basis function the two-dimensional wavelet transformation of atmospheric environment data, obtains two-dimentional multiresolution function space from the multiresolution function space of one dimension by tensor product.Method detailed process is: first each row of atmospheric environment data matrix is carried out to once one dimension small echo line translation, the result of gained is carried out one dimension small echo rank transformation again, or what replace implements the wavelet coefficient that one-dimensional wavelet transform obtains atmospheric environment data on row and column, the general trend situation that wavelet coefficient not only reflects data also reflects the situation of change of data.

Step 5: determine wavelet coefficient screening threshold value.

Described step 5 determination wavelet coefficient screens the method for threshold value specifically:

Step 5.1: set up level of resolution according to raw data and lowest resolution data and divide, and determine corresponding wavelet coefficient screening threshold value;

Step 5.2: the severe degree according to atmospheric properties change carrys out Confirming model level of resolution, adopts the method for interpolation to obtain corresponding threshold value by level of resolution.

Step 6: the detail section adding wavelet coefficient carries out wavelet inverse transformation, generates multiresolution atmospheric environment data model.

The detail section that described step 6 adds wavelet coefficient carries out the method for wavelet inverse transformation specifically:

Step 6.1: according to the threshold value calculated in step 5, contrasts the wavelet coefficient of the atmospheric environment data of gained in step 4 and threshold value, retains the wavelet coefficient between threshold value;

Step 6.2: the detail section adding the whole wavelet coefficients remained carries out data reconstruction, generates the atmospheric environment data model of multiresolution.

Embodiment:

As shown in Figure 1, the multiresolution atmosphere data modeling method based on wavelet analysis of the present invention can complete flow process in accordance with the following steps:

Step one: obtain the environmental area required for aircraft, is loaded into atmospheric environment data.

Step 1.1: three dimensions can be decomposed into the two-dimensional space of different levels from short transverse, supposes that the environmental area required for miniature unmanned vehicle path planning is two-dimentional Euclidean space, is expressed as

Step 1.2: the Obstacles such as the massif that aircraft can not pass through, building are expressed as then non-Obstacles is expressed as path planning problem is abstract is: if the initial position point of unmanned vehicle is with source location be in space middle searching is a series of can flight position point, to meet certain constraint condition and goal condition.The environmental area selected in the embodiment of the present invention is the digital terrain elevation figure of 1024 × 1024 pixels.As landform altitude q _ican flying height H higher than aircraft, then this region is barrier region; As landform altitude q _ican flying height H lower than aircraft, this region is clear region, and formula (4) can be adopted to represent:

X _icertain some position in representative digit landform altitude figure, q _irepresent position X _ilandform altitude.Described aircraft can flying height H be the minimum altitude (relative to sea level) of miniature unmanned vehicle safe flight, then may cause danger lower than this altitude.

Step 1.3: in order to make the description of algorithm have readability, without loss of generality, environmental area be normalized to unit square region, be expressed as [0,1] × [0,1], the most fine grid blocks that can decompose is 2 ⁿ× 2 ⁿ, be spaced apart 1/2 ⁿ, because initial landform figure is 1024 × 1024 pixels, so the N=10 chosen.If best result distinguishes that grade is J _max, then/ _maxn can not be greater than.

Step 2: choose the basis function of wavelet transformation and build two dimensional basis functions.

Step 2.1: choose Haar wavelet, determines the Haar wavelet basis function of the one-dimensional space.

In multi-resolution Modeling process, environmental cost function approaches reconstruct by wavelet transformation basis function, different levels approach the resolution representing different brackets.Select suitable wavelet transformation basis function to approach environmental cost function, make wavelet transformation easily and accurately can describe the feature of multiresolution environment space.Select Haar wavelet basis function in the present invention, Haar wavelet basis function is the orthogonal wavelet function with compactly supported be the most simply suitable for.Because environmental area is two-dimentional Euclidean space, the wavelet basis function of two-dimensional space can be constructed by the Haar wavelet basis function of the one-dimensional space.Haar wavelet in the one-dimensional space is expressed as:

$\mathrm{\φ}\left(x\right)=\left\{\begin{array}{cc}1& x\∈\left[\mathrm{0,1}\right)\\ 0& x\∉\left[\mathrm{0,1}\right)\end{array}\right.,\mathrm{\ψ}\left(x\right)=\left\{\begin{array}{cc}1& x\∈[\mathrm{0,1}/2)\\ -1& x\∈[1/\mathrm{2,1})\\ 0& x\∉\left[\mathrm{0,1}\right)\end{array}\right.---\left(5\right)$

Wherein, φ (x) represents Haar scaling function, and ψ (x) represents Haar mother wavelet function, and x represents the location variable of the one-dimensional space.

Formula (5) is obtained the similar small echo of one group of shape by flexible with translation, obtains the Haar wavelet basis function of the one-dimensional space:

$\begin{array}{c}{\mathrm{\φ}}_i^j\left(x\right)=\sqrt{2^j}\mathrm{\φ}(2^{j}x-i),{\mathrm{\ψ}}_i^j\left(x\right)=\sqrt{2^j}\mathrm{\ψ}(2^{j}x-j)\\ j=\mathrm{0,1},...\mathrm{and\; i}=\mathrm{0,1},...,2^j-1\end{array}---\left(6\right)$

Wherein, j represents contraction-expansion factor 2 ^jindex, can characterize level of resolution, i represents translational movement, represent the Haar wavelet basis function after flexible and translation.

Step 2.2: the Haar wavelet basis function of structure two-dimensional space, the wavelet basis function of two-dimensional space is the fundamental element of multi-resolution representation.The Haar wavelet basis function of two-dimensional space is constructed such as formula shown in (7) according to formula (6):

$\begin{array}{c}{\mathrm{\Φ}}_{m,n}^j(x,y)={\mathrm{\φ}}_m^j\left(x\right){\mathrm{\φ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{1,j}(x,y){=\mathrm{\φ}}_m^j\left(x\right){\mathrm{\ψ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{2,j}(x,y)={\mathrm{\ψ}}_m^j\left(x\right){\mathrm{\φ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{3,j}(x,y)={\mathrm{\ψ}}_m^j\left(x\right){\mathrm{\ψ}}_n^j\left(y\right)\end{array}---\left(\text{7}\right)$

In above formula, subscript j can represent the grade of resolution, and j is larger, can comprise more detailed information, and the grade of resolution is higher; M, n are the location label of two-dimensional space both direction, and (x, y) represents the location variable of two-dimensional space, represent the Haar scaling function of two-dimensional space, represent the Haar mother wavelet function in n direction, represent the Haar mother wavelet function in m direction, represent the Haar mother wavelet function in diagonal.

Step 2.3: the wavelet basis function form of any one binary function two-dimensional space approached, the form that different levels are approached is the form differentiating expression more, and approach and be divided into two parts, Part I is rough approximation, and Part II is that details is similar to.Any one function by formula (7) close approximation be:

Wherein, for the coefficient of two-dimensional space Haar scaling function in close approximation, for the coefficient of two-dimensional space Haar mother wavelet function in close approximation, particular according to formula (9) obtains:

$\begin{array}{c}{\mathrm{\λ}}_{m,n}^{J_{\min }}={\∫}_{-\∞}^{\∞}{\∫}_{-\∞}^{\∞}f(x,y){\mathrm{\Φ}}_{m,n}^{J_{\min }}(x,y)\mathrm{dx\; dy}\\ {\mathrm{\ξ}}_{m,n}^{k,j}={\∫}_{-\∞}^{\∞}{\∫}_{-\∞}^{\∞}f(x,y){\mathrm{\Ψ}}_{-\∞}^{\∞}(x,y)\mathrm{dx\; dy}\end{array}---\left(9\right)$

Section 1 in formula (8) on the right of equation be the rough approximation to f (x, y), order of approximation is J _min, Section 2 contain the more details information of f (x, y).J _minfor lowest resolution grade, represent integer, for the function set that absolute square on two-dimensional space can be amassed.

Step 3: in described step 3, the method for wavelet transformation is carried out specifically to atmospheric environment data:

Step 3.1: read raw data and piecemeal process, select interested ambient area, is shown as effective lattice point (x in current region by atmospheric environment tables of data _i, y _i) corresponding data value z _idata layout;

Step 3.2: choose when suitable boundary extension mode carries out boundary extension process to prevent from carrying out wavelet transformation to above-mentioned atmospheric environment data in data boundary locations generation distortion;

Step 3.3: realized by the tensor product of one dimension basis function the two-dimensional wavelet transformation of atmospheric environment data, obtains two-dimentional multiresolution function space from the multiresolution function space of one dimension by tensor product.Method detailed process is: first each row of atmospheric environment data matrix is carried out to once one dimension small echo line translation, the result of gained is carried out one dimension small echo rank transformation again, or what replace implements the wavelet coefficient that one-dimensional wavelet transform obtains atmospheric environment data on row and column, the general trend situation that wavelet coefficient not only reflects data also reflects the situation of change of data.

Step 4: in described step 4, the method for wavelet transformation is carried out specifically to atmospheric environment data:

Step 4.1: read raw data and piecemeal process, select interested ambient area, is shown as effective lattice point (x in current region by atmospheric environment tables of data _i, y _i) corresponding data value z _idata layout;

Step 4.2: choose when suitable boundary extension mode carries out boundary extension process to prevent from carrying out wavelet transformation to above-mentioned atmospheric environment data in data boundary locations generation distortion;

Step 4.3: realized by the tensor product of one dimension basis function the two-dimensional wavelet transformation of atmospheric environment data, obtains two-dimentional multiresolution function space from the multiresolution function space of one dimension by tensor product.Method detailed process is: first each row of atmospheric environment data matrix is carried out to once one dimension small echo line translation, the result of gained is carried out one dimension small echo rank transformation again, or what replace implements the wavelet coefficient that one-dimensional wavelet transform obtains atmospheric environment data on row and column, the general trend situation that wavelet coefficient not only reflects data also reflects the situation of change of data.

Step 5: determine wavelet coefficient screening threshold value.

Described step 5 determines the method for wavelet coefficient screening threshold value specifically:

Step 5.1: set up level of resolution according to raw data and lowest resolution data and divide, and determine corresponding wavelet coefficient screening threshold value;

Step 5.2: the severe degree according to atmospheric properties change carrys out Confirming model level of resolution, adopts the method for interpolation to obtain corresponding threshold value by level of resolution.

The detail section that described step 6 adds wavelet coefficient carries out the method for wavelet inverse transformation specifically:

Step 6.1: according to the threshold value calculated in step 5, is undertaken the wavelet coefficient of the atmospheric environment data of gained in step 4 and threshold value contrasting to determine whether this coefficient is retained;

Step 6.2: the detail section adding the whole wavelet coefficients remained carries out data reconstruction, generates the atmospheric environment data model of multiresolution.

Above-mentioned method for designing is at IntelPentium (R) Dual 1.60GHz processor, 2.0G physical memory realizes, software condition is Matlab R2010a, what choose is atmospheric temperature attribute, result is as shown in Fig. 2 to Fig. 4 and table 1, can find out: the method effectively can reduce the valid data number of environmental data, greatly reduces data volume, improve device storage efficiency.

Table 1 initial temperature field, one-level resolution temperature field, the contrast of secondary resolution temperature field data amount

Data name	Initial temperature field	One yardstick approximate temperature field	Two yardstick approximate temperature fields
				Size of data	106＊140	55＊72	30＊38

The inventive method is by being applied in the multi-resolution representation to atmospheric environment by Haar Wavemeshs resolution theory, and the storage efficiency of effective lifting means, solves the problem that airborne computer computational resource is limited.The present invention has stronger practicality, and the multi-resolution Modeling for atmospheric field data provides a kind of simple and practical new method.

Claims (1)

1., based on a multiresolution atmosphere data modeling method for wavelet analysis, comprise following step:

Step one: obtain the ambient area that aircraft needs;

Step 1.1: by the three dimensional environmental space region residing for aircraft, is decomposed into the two-dimentional Euclidean space of different levels from short transverse, the two-dimentional Euclidean space obtained is required ambient area;

Step 1.2: can flying height according to landform altitude and aircraft by the two-dimentional Euclidean space that obtains, is divided into barrier and non-barrier, and determines initial position and the target location of aircraft, to i-th position X in original figure landform altitude figure _ibe expressed as follows:

Wherein, represent Obstacles, represent non-Obstacles, z _irepresent position X _ilandform altitude, what H represented aircraft can flying height;

Step 1.3: environmental area is normalized to unit square region, and carry out stress and strain model;

Step 2: choose wavelet function and boundary extension mode;

Choose Haar small echo, choose wavelet function and boundary extension mode;

Step 3: according to the wavelet function chosen in step 2, builds the wavelet transformation basis function of two-dimensional space;

Specifically comprise following step:

Step 3.1: the Haar wavelet choosing the one-dimensional space, obtains one group shape similar small echo by flexible with translation by Haar wavelet, obtains the Haar wavelet basis function of the one-dimensional space:

$\begin{array}{c}{\mathrm{\φ}}_i^j\left(x\right)=\sqrt{2^j}\mathrm{\φ}(2^{j}x-i),{\mathrm{\ψ}}_i^j\left(x\right)=\sqrt{2^j}\mathrm{\ψ}(2^{j}x-i)\\ j=\mathrm{0,1},...J_{\max }\mathrm{and\; i}=\mathrm{0,1},...,2^j-1\end{array}---\left(2\right)$

Wherein, j represents level of resolution, 2 ^jrepresent contraction-expansion factor, i represents translational movement, represent the Haar scaling function after flexible and translation and Haar mother wavelet function respectively, x represents the location variable of the one-dimensional space;

Step 3.2: the Haar wavelet basis function being constructed two-dimensional space by the Haar wavelet basis function of the one-dimensional space is as follows:

$\begin{array}{c}{\mathrm{\Φ}}_{m,n}^j(x,y)={\mathrm{\φ}}_m^j\left(x\right){\mathrm{\φ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{1,j}(x,y)={\mathrm{\φ}}_m^j\left(x\right){\mathrm{\ψ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{2,j}(x,y)={\mathrm{\ψ}}_m^j\left(x\right){\mathrm{\φ}}_n^j\left(y\right)\\ {\mathrm{\Ψ}}_{m,n}^{3,j}(x,y)={\mathrm{\ψ}}_m^j\left(x\right){\mathrm{\ψ}}_n^j\left(y\right)\end{array}---\left(3\right)$

Wherein, (x, y) represents the location variable of two-dimensional space, and m, n are the location label of two-dimensional space both direction; represent the Haar scaling function of two-dimensional space, represent the Haar mother wavelet function in n direction, represent the Haar mother wavelet function in m direction, represent the Haar mother wavelet function in diagonal;

Step 4: wavelet transformation is carried out to atmospheric environment data;

In described step 4, the method for wavelet transformation is carried out specifically to atmospheric environment data:

Step 4.1: read raw data and piecemeal process, select interested ambient area, is shown as effective lattice point (x in current region by atmospheric environment tables of data _i, y _i) corresponding data value z _idata layout;

Step 4.2: choose boundary extension mode and boundary extension process is carried out to above-mentioned atmospheric environment data;

Step 4.3: the two-dimensional wavelet transformation of atmospheric environment data is realized by the tensor product of one dimension basis function, two-dimentional multiresolution function space is obtained by tensor product from the multiresolution function space of one dimension, method detailed process is: first each row of atmospheric environment data matrix is carried out to once one dimension small echo line translation, the result of gained is carried out one dimension small echo rank transformation again, or replace on row and column, implement the wavelet coefficient that one-dimensional wavelet transform obtains atmospheric environment data;

Step 5: determine wavelet coefficient screening threshold value;

Described step 5 determination wavelet coefficient screens the method for threshold value specifically:

Step 5.1: set up level of resolution according to raw data and lowest resolution data and divide, and determine corresponding wavelet coefficient screening threshold value;

Step 5.2: the severe degree according to atmospheric properties change carrys out Confirming model level of resolution, adopts the method for interpolation to obtain corresponding threshold value by level of resolution;

Step 6: the detail section adding wavelet coefficient carries out wavelet inverse transformation, generates multiresolution atmospheric environment data model;

The detail section that described step 6 adds wavelet coefficient carries out the method for wavelet inverse transformation specifically:

Step 6.1: according to the threshold value calculated in step 5, contrasts the wavelet coefficient of the atmospheric environment data of gained in step 4 and threshold value, retains the wavelet coefficient between threshold value;

Step 6.2: the detail section adding the whole wavelet coefficients remained carries out data reconstruction, generates the atmospheric environment data model of multiresolution.