CN102081737B

CN102081737B - Multi-scale modeling method for pneumatic heat radiation images and application thereof

Info

Publication number: CN102081737B
Application number: CN201010614054A
Authority: CN
Inventors: 张天序; 关静; 余铮; 陈建冲; 武道龙; 杨卫东
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2010-12-30
Filing date: 2010-12-30
Publication date: 2012-10-10
Anticipated expiration: 2030-12-30
Also published as: CN102081737A

Abstract

The invention discloses a multi-scale modeling method for aerothermal radiation images, which includes: (1) registering each aerothermal radiation degradation image respectively, and calculating the difference between each degradation image and a reference image; The difference image is fitted in the whole image area, that is, the first scale, and the fitting surface at the first scale is obtained; (3) The last fitting scale is refined, and the difference image d _k is multi-scaled The surface approximation fitting is obtained to obtain the fitted surface polynomials at each scale, which constitutes the window thermal radiation fingerprint library of the aerothermal radiation degradation image sequence in the corresponding aerothermal environment. The invention also discloses an application of image correction by applying the above method. The contrast of the target area after correction in the present invention is obviously improved, and the contrast is improved more obviously with the refinement of the scale, and can be widely used in image correction.

Description

Pneumatic heat radiation image multi-scale modeling method and application thereof

Technical Field

The invention belongs to the technical field of cross science combining pneumatic optics and image processing, and particularly relates to a pneumatic thermal radiation multi-scale modeling method and application thereof in image correction.

Background

The aerooptical system is a discipline for researching the influence of a high-speed streaming flow field on the imaging detection of a high-speed aircraft. When the high-speed aircraft with the optical imaging detection system flies in the atmosphere, the interaction between the optical window and the incoming flow forms a complex flow field. Due to the viscous action of the air, the air flow in contact with the surface of the optical window will be retarded so that the air flow velocity decreases and a boundary layer is formed near the surface of the window. The layers with a large velocity gradient in the boundary layer generate strong friction, and the kinetic energy of the air flow is irreversibly changed into heat energy, which causes the temperature of the wall surface of the window to rise. The high temperature air flow will continuously transfer heat to the low temperature wall surface, causing strong pneumatic heating. The optical window is pneumatically heated and is in a severe pneumatic thermal environment, thermal radiation noise is generated, and the signal-to-noise ratio and the image quality of the photoelectric detection system are reduced.

The greater the flight speed, the more severe the airflow heats up on the aircraft surface. The irradiance of the air flow outside the window and the irradiance of the window are superposed with the irradiance of the background, and the imaging sensor enters a nonlinear region or is saturated, so that the loss of effective information of a scene or the reduction of a signal-to-noise ratio and a signal-to-noise ratio, the reduction of detection performance or functional failure are caused. Therefore, aerodynamic thermal radiation correction is needed to improve the signal-to-noise ratio.

Because the degradation model of the pneumatic thermal radiation is unknown and randomly changed, the degraded image also contains sensor noise, the difficulty of image recovery or correction is increased, and no relevant literature reports a pneumatic thermal radiation degraded image correction method at present.

Disclosure of Invention

The invention provides a multi-scale modeling method for an aerodynamic thermal radiation image. The invention also provides a method for correcting the pneumatic thermal radiation degraded image sequence by using the window thermal radiation fingerprint obtained by modeling, which can effectively correct and recover the pneumatic thermal radiation image and improve the signal-to-noise ratio and the image quality of the image.

In the invention, the optical window in the pneumatic thermal environment has a rule of changing along with the temperature and the pressure intensity, and is called as a window thermal radiation fingerprint.

The invention provides a multi-scale modeling method of a pneumatic thermal radiation image, which comprises the following specific steps:

(1) acquiring a sequence (f) of aero-thermal radiation degradation images in a certain aero-thermal environment (i.e. under certain temperature and pressure conditions) by an imaging device₁，f₂，f₃，......，f_n-2，f_n-1，f_n) And each frame of image in the image sequence is compared with a reference image f₀Grouped in pairs, each group of images constituting the basic processing object of the correction and restoration operation:

(f₀，f₁)，(f₀，f₂)，(f₀，f₃)，......，(f₀，f_n-1)，(f₀，f_n)

n is the number of frames of the degraded image sequence, f_kAt a temperature of T_kPressure of P_kK is the frame number of the pneumatic thermal radiation degradation image;

(2) for each set of images (f)₀，f_k) Registering to obtain registered image combination (f)₀，f′_k) And offset value (Vx)_k，Vy_k)，Vx_kIs the offset in the x-axis direction, Vy_kIs the offset in the y-axis direction;

(3) for each set of images after registration (f)₀，f′_k) Obtaining the difference d between the thermal radiation degradation image and the reference image_k＝f′_k-f₀；

(4) Using least square approximation principle to difference image d_kPerforming multi-scale curved surface approximation fitting to obtain a curved surface polynomial fitting under each scale

Namely, the window thermal radiation fingerprint, scale ═ max, mid, min, respectively representing a large scale (i.e., a first scale), a medium scale (i.e., a second scale), and a small scale (i.e., a third scale), p and q being the highest powers of polynomials, and if being a biquadratic polynomial, taking p ═ q ═ 3; the method specifically comprises the following steps:

large scale least squares approximation: setting the size of the pneumatic thermal radiation degradation image as N_max×M_maxFor N in the full-image region_max×M_maxSampling the difference values, and taking N'_max×M′_maxA difference point (x)_u，y_v)(u＝0，1，...，N′_max-1；v＝0，1，...，M′_max-1) fitting a large-scale polynomial surface to obtain a large-scale fitted surface polynomial

I.e. thermal radiation fingerprints at large scale;

mesoscale least squares approximation: thermal radiation fingerprint under large scale obtained by analysisFor errorsPerforming further mesoscale analysis on the regions with relatively large difference, and setting the size of the selected mesoscale region as N_mid×M_midFor N in the region_mid×M_midSampling the difference values, and taking N'_mid×M′_midA difference point (x)_u，y_v)(u＝0，1，...，N′_mid-1；v＝0，1，...，M′_mid-1) performing a mesoscale polynomial surface fit to obtain a surface polynomial fit at the mesoscale

I.e. thermal radiation fingerprint at the mesoscale;

small scale least squares approximation: thermal radiation fingerprint under mesoscale obtained by analysis

Performing further small-scale analysis on the region with relatively large error, and setting the size of the selected small-scale region as N_min×M_minFor N in the region_min×M_minSampling the difference values, and taking N'_min×M′_minA difference point (x)_u，y_v)(u＝0，1，...，N′_min-1；v＝0，1，...，M′_min-1) fitting a small-scale polynomial surface to obtain a small-scale fitted surface polynomial

I.e. thermal radiation fingerprints at small scale;

analyzing the obtained thermal radiation fingerprint under small scaleIf the area with relatively large error still exists, the selected area with relatively large error can be analyzed in a micro scale or other smaller scales according to the steps to obtain the thermal radiation fingerprint in the corresponding scale;

by combining the above processes, certain pneumatic heat can be obtainedOffset of optical axis (Vx) under environment_k，Vy_k) And thermal radiation fingerprints at a plurality of scales such as large scale, medium scale and small scale

(scale ═ max, mid, min), a multi-scale pneumatic thermal radiation image correction fingerprint library can be established.

The method for correcting the image by utilizing the pneumatic thermal radiation image correction fingerprint library obtained by the method comprises the following specific steps:

(1) inputting a sequence of aerodynamic thermal radiation degraded images (g)₁，g₂，g₃，......，g_m-2，g_m-1，g_m) M is the number of frames of the degraded image sequence, g_kIs a temperature T_kPressure P_kThe method comprises the following steps of (1) obtaining a pneumatic thermal radiation degradation image under a condition, wherein k is a frame number of the pneumatic thermal radiation degradation image;

(2) degradation image g for arbitrary aerodynamic heat radiation_kObtaining the temperature T in a pneumatic thermal radiation image correction fingerprint database_kAnd pressure P_kThermal radiation fingerprint of selected corresponding dimension under condition(scale max, mid, min) and corresponding optical axis offset (Vx)_k，Vy_k) Correcting the aerodynamic heat radiation degradation image to obtain correction results g 'of the aerodynamic heat radiation degradation image at a plurality of scales'_k，scale；

The target area contrast analysis and comparison are carried out on the pneumatic thermal radiation degradation images before and after correction, the contrast of the corrected target area is obviously improved, the contrast is obviously improved along with the scale thinning, and the effectiveness of the method is proved.

Drawings

FIG. 1 is a schematic view of the aerodynamic heat radiation effect of an optical window of a high speed aircraft;

FIG. 2 is a flow chart of the creation of a multi-scale pneumatic thermal radiation fingerprint library of the present invention;

FIG. 3 is a flow chart of a multi-scale aerodynamic thermal radiation image correction method of the present invention;

FIG. 4 is a flow chart of a multi-scale analysis of aerodynamic heat radiation images of the present invention;

FIG. 5 is a flow chart of the creation of a multi-scale aerodynamic heat radiation fingerprint of the present invention;

FIG. 6(a) is a reference image of a sequence of aerodynamic thermal radiation degraded images;

FIG. 6(b) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₁A frame image;

FIG. 6(c) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₂A frame image;

FIG. 6(d) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₃A frame image;

FIG. 6(e) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₄A frame image;

FIG. 6(f) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₅A frame image;

FIG. 6(g) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₆A frame image;

FIG. 6(h) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₇A frame image;

FIG. 6(i) is the f-th image of the sequence of aerodynamic thermal radiation degradation images in the modeling method of the present invention₈A frame image;

FIG. 7 is a graph of the target centroid position shift after registration of the aero thermal radiation degradation images of FIGS. 6(b) -6 (i);

FIG. 8(a) is a three-dimensional representation of the difference between the degraded image of FIG. 6(b) and the reference image of FIG. 6(a) after registration;

FIG. 8(b) is a three-dimensional representation of the difference between the degraded image of FIG. 6(c) and the reference image of FIG. 6(a) after registration;

FIG. 8(c) is a three-dimensional representation of the difference between the degraded image of FIG. 6(d) and the reference image of FIG. 6(a) after registration;

FIG. 8(d) is a three-dimensional representation of the difference between the degraded image of FIG. 6(e) and the reference image of FIG. 6(a) after registration;

FIG. 8(e) is a three-dimensional representation of the difference between the degraded image of FIG. 6(f) and the reference image of FIG. 6(a) after registration;

FIG. 8(f) is a three-dimensional representation of the difference between the degraded image of FIG. 6(g) and the reference image of FIG. 6(a) after registration;

FIG. 8(g) is a three-dimensional representation of the difference between the degraded image of FIG. 6(h) and the reference image of FIG. 6(a) after registration;

FIG. 8h is a three-dimensional representation of the difference between the degraded image of FIG. 6i after registration and the reference image of FIG. 6 a;

FIG. 9(a) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(b) at a large scale;

FIG. 9(b) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(c) at a large scale;

FIG. 9(c) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(d) at a large scale;

FIG. 9(d) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(e) at a large scale;

FIG. 9(e) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(f) at a large scale;

FIG. 9(f) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(g) at a large scale;

FIG. 9(g) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(h) at a large scale;

FIG. 9(h) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(i) at a large scale;

FIG. 10(a) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(b) at a mesoscale;

FIG. 10(b) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(c) at a mesoscale;

FIG. 10(c) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(d) at a mesoscale;

FIG. 10(d) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(e) at a mid-scale;

FIG. 10(e) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(f) at a mid-scale;

FIG. 10(f) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(g) at a mid-scale;

FIG. 10(g) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(h) at a mid-scale;

FIG. 10(h) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(i) at a mid-scale;

FIG. 11(a) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(b) at a small scale;

FIG. 11(b) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(c) at a small scale;

FIG. 11(c) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(d) at a small scale;

FIG. 11(d) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(e) at a small scale;

FIG. 11(e) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(f) at a small scale;

FIG. 11(f) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(g) at a small scale;

FIG. 11(g) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(h) at a small scale;

FIG. 11(h) is a three-dimensional display of the thermal radiation fingerprint of the degraded image of FIG. 6(i) at a small scale;

FIG. 12 is a schematic diagram of the data overlap region obtained by extending the boundary of the iso-halo region appropriately outward;

FIG. 13 is a schematic diagram of a data overlap region in one dimension with the appropriate extension of the isoplanar boundaries outward;

FIG. 14 is an actual aerodynamic thermal radiation degradation image that requires correction;

FIG. 15(a) is a thermal radiation corrected image at a large scale of the aerodynamic thermal radiation degradation image of FIG. 14;

FIG. 15(b) is a thermal radiation corrected image at the mesoscale of the aerodynamic thermal radiation degradation image of FIG. 14;

FIG. 15(c) is a thermal emission corrected image at a small scale of the aerodynamic thermal emission degradation image of FIG. 14;

FIG. 16(a) is a schematic drawing showing the selection of the target region ((170, 53), (237, 124));

fig. 16 b shows the results of evaluating the contrast of the corrected image of the aerodynamic thermal radiation degradation image in fig. 14 at a plurality of scales (fig. 16 a), in which 5 scales on the abscissa represent the reference image (fig. 6 a), the aerodynamic thermal radiation degradation image (fig. 14), the corrected image of the degradation image at the large scale (fig. 15 a), the corrected image at the medium scale (fig. 15 b), and the corrected image at the small scale (fig. 15 c) in this order from left to right, and the ordinate represents the contrast of the corresponding area of the images;

FIG. 17(a) is a schematic diagram showing the selection of the target region ((103, 154), (145, 195));

fig. 17 b shows the results of evaluating the contrast of the area (fig. 17 a) of the corrected image of the aerodynamic heat radiation degradation image in fig. 14 at a plurality of scales, in which 5 scales on the abscissa represent the reference image (fig. 6 a), the aerodynamic heat radiation degradation image (fig. 14), the corrected image of the degradation image at the large scale (fig. 15 a), the corrected image at the medium scale (fig. 15 b), and the corrected image at the small scale (fig. 15 c) in this order from left to right, and the ordinate represents the contrast of the area corresponding to the images.

Detailed Description

The present invention will be described in further detail with reference to the following drawings and specific examples.

A multi-scale modeling method for an aerodynamic heat radiation image comprises the following steps:

(1) acquiring a sequence (f) of aero-thermal radiation degradation images in a certain aero-thermal environment (i.e. under certain temperature and pressure conditions) by an imaging device₁，f₂，f₃，......，f_n-2，f_n-1，f_n) And compare it with a reference image f₀Grouped in pairs, each group of images constituting the basic processing object of the correction and restoration operation:

n is the number of frames of the degraded image sequence, f_kAt a temperature of T_kPressure of P_kThe k is the degradation of the aerodynamic thermal radiationThe frame number of the picture is quantized.

Degrading the pneumatic thermal radiation respectively by the f-th image sequence₁Frame degraded image, reference image, f₂The frame degraded image is grouped in pairs with the reference image, and the subsequent pneumatic thermal radiation degraded image is similarly processed.

(2) For each set of images (f)₀，f_k) Registering to obtain registered image combination (f)₀，f′_k) And offset value (Vx)_k，Vy_k)，Vx_kIs the offset in the x-axis direction, Vy_kIs the offset in the y-axis direction.

For the registered first set of images (f)₀，f′₁) Finding the f-th after registration₁Frame aerodynamic thermal radiation degradation image and reference image f₀Difference d of₁＝f′₁-f₀And so on, for the k-th group of images after registration (f)₀，f′_k) F after registration_kFrame aerodynamic thermal radiation degradation image and reference image f₀Has a difference of d_k＝f′_k-f₀Thereby obtaining a difference image sequence (d) of the aero-bolometric degradation image and the reference image₁，d₂，d₃，......，d_n-2，d_n-1，d_n)。

(4) Using least square approximation principle to difference image d_kPerforming multi-scale curved surface approximation fitting to obtain a curved surface polynomial fitted under each scale

scale is max, mid, min, which respectively represents large scale, medium scale and small scale, p and q are the highest power of the polynomial, and if the scale is a biquadratic polynomial, p is q is 3;

aligning the difference image d according to steps (4.1) to (4.4)_kCarrying out surface fitting, wherein the specific process is as follows:

(4.1) for N × M points (x) within the rectangular region of the image_u，y_v) (u-0, 1,. cndot., N-1; v-0, 1,. said, M-1; ) Value z of_uvLet the least squares fit polynomial be

Wherein p and q are the highest power of the polynomial;

(4.2) fixing y, and constructing M least square fitting polynomials for x:

wherein,

(i ═ 0, 1.. times, p) are mutually orthogonal polynomials constructed from the following recursion formula:

order to

Is provided with

β_i＝η_i/η_i-1According to the least square principle, the following can be obtained:

(4.3) constructing a least squares fit polynomial for y:

wherein psi_j(y) (j ═ 0, 1.. times, q) are mutually orthogonal polynomials constructed from the following recursion formula:

ψ₀(y)＝1

ψ₁(y)＝y-α′₀

ψ_j(y)＝(y-α′_j)ψ_j-1(y)-β′_jψ_j-2(y)

order to

Is provided with

β′_j＝δ_j/δ_j-1According to the least square principle, the following can be obtained:

(4.4) combining the results of the derivation in the steps (4.2) and (4.3) to obtain a polynomial for surface fitting:

the polynomial form converted to the standard is:

(5) large scale least squares approximation: let the image size be N_max×M_maxFor N in the full-image region_max×M_maxSampling the difference values, and taking N'_max×M′_maxA difference point (x)_u，y_v)(u＝0，1，...，N′_max-1；v＝0，1，...，M′_max-1), fitting the large-scale polynomial surface according to the steps (4.1) to (4.4) to obtain a large-scale fitted surface polynomial

I.e. degraded image f_kThermal radiation fingerprints at large scale.

(6) Mesoscale least squares approximation: thermal radiation fingerprint under large scale obtained by analysis

Performing further mesoscale analysis on the region with relatively large error, and setting the size of the selected mesoscale region as N_mid×M_midFor N in the region_mid×M_midSampling the difference values, and taking N'_mid×M′_midA difference point (x)_u，y_v)(u＝0，1，...，N′_mid-1；v＝0，1，...，M′_mid-1) performing a mesoscale polynomial surface fit to obtain a surface polynomial fit at the mesoscale

I.e. degraded image f_kThermal radiation fingerprints at a mesoscale;

and (3) carrying out mesoscale analysis on the aerodynamic heat radiation degradation image according to the steps (6.1) to (6.3), wherein the specific process is as follows:

(6.1) subjecting the large-scale thermal radiation fingerprint obtained in the step (5)Is divided into M₂×M₂Individual block (for mesoscale analysis, M)₂Generally 2 to 4) for each sub-block(s is the number of the sub-block, s is 1, 2, L, M)₂×M₂) Showing that the difference image d obtained in step (3) is similarly used_kIs correspondingly divided into M₂×M₂Individual sub-blocks, each sub-block being represented by d_k，mid(s)Showing that each sub-block can be approximately regarded as an isoplanatic zone, and the boundary of the isoplanatic zone is properly extended outwards for a certain pixel value;

(6.2) calculating N in the full-map region_max×M_maxDot (x)_u，y_v)(u＝0，1，...，N_max-1；v＝0，1，...，M_max-1) relative error of large-scale thermal radiation fingerprintsAnd N in each sub-block region_mid(s)×M_mid(s)Dot (x)_u，y_v)(u＝0，1，...，N_mid(s)-1；v＝0，1，...，M_mid(s)-1) relative error of large-scale thermal radiation fingerprints

If | is the absolute value sign

The mesoscale thermal radiation fingerprint of the corresponding sub-block

If it is notThen for difference sub-block d_k，mid(s)N within a region_mid(s)×M_mid(s)Sampling the difference values, and taking N'_mid(s)×M′_mid(s)A difference point (x)_u，y_v)(u＝0，1，...，N′_mid(s)-1；v＝0，1，...，M′_mid(s)-1), fitting the surface of the medium-scale polynomial according to the steps (4.1) to (4.4) to obtain the surface polynomial fitted under the medium-scaleNamely the thermal radiation fingerprint of the corresponding sub-block under the mesoscale;

(6.3) using a splicing algorithm to carry out thermal radiation fingerprint on each subblock under the mesoscale

(s＝1，2，L，M₂×M₂) Splicing to obtain a degraded image f_kThermal radiation fingerprint at mesoscale

The specific process comprises the following steps: and constructing a weighting coefficient according to the distance from the pixel of the overlapping area of each sub-block to the boundary, and finishing gradual change splicing by using the data of the overlapping area so as to remove visual fracture feeling of the spliced image.

As shown in fig. 12, region [1 ]][2][3]The part which is not overlapped with other areas under a certain level scale model does not need to be processed, and the original value is directly used, namely the area [4 ]]Is the overlapping part of two areas, area [5 ]]Is the overlapping portion of 4 regions; in the process of splicing and superposition, a weighted average method is adopted, and the weighting coefficients are simplified and illustrated by taking the overlapping in the one-dimensional direction as an example. The adjacent two image blocks are respectively represented by X, YThe images after the two are spliced are represented by Z, and as shown in FIG. 13, the size of X, Y is W_X、W_YThe boundary of the block region of the image boundary extends outwards by L, namely the boundary of the block region is at W-th of X_X-column L, at column L of Y. X, Y after the image correction process, the column L with block effect is removed, and the width 2d of the overlap area of the two image blocks is 2(L-L), i.e. the block boundary at this time is at the W-th of X_XAnd a column d, wherein in the column d of the Y, the weight coefficients of the transition elements in the image overlapping areas on the two sides are recurred according to 50%/d being 1/(2 d).

(7) Small scale least squares approximation: thermal radiation fingerprint under mesoscale obtained by analysis

I.e. degraded image f_kThermal radiation fingerprints at small scales;

and (3) carrying out small-scale analysis on the aerodynamic heat radiation degradation image according to the steps (7.1) to (7.3), wherein the specific process is as follows:

(7.1) subjecting the medium-scale thermal radiation fingerprint obtained in the step (6)

Is divided into M₃×M₃Individual block (for small scale analysis, M)₃Generally, 4 to 8) are selected for each sub-block

(s is the number of the sub-block, s is 1, 2, L, M)₃×M₃) Showing that the difference image d obtained in step (3) is similarly used_kIs correspondingly divided into M₃×M₃Individual sub-blocks, each sub-block being represented by d_k，min(s)Showing that each sub-block can be approximately regarded as an isoplanatic zone, and the boundary of the isoplanatic zone is properly extended outwards for a certain pixel value;

(7.2) calculating N in the full-map region_max×M_maxDot (x)_u，y_v)(u＝0，1，...，N_max-1；v＝0，1，...，M_max-1) relative error of the mesoscale thermal radiation fingerprint

And N in each sub-block region_min(s)×M_min(s)Dot (x)_u，y_v)(u＝0，1，...，N_min(s)-1；v＝0，1，...，M_min(s)-1) relative error of the mesoscale thermal radiation fingerprint

If | is the absolute value sign

Then the small-scale thermal radiation fingerprint of the corresponding sub-block

If it is notThen for difference sub-block d_k，min(s)N within a region_min(s)×M_min(s)Sampling the difference values, and taking N'_min(s)×M′_min(s)A difference point (x)_u，y_v)(u＝0，1，...，N′_min(s)-1；v＝0，1，...，M′_min(s)-1), fitting the small-scale polynomial surface according to the steps (4.1) to (4.4) to obtain a surface polynomial fitted under the small scale

Namely the thermal radiation fingerprint of the corresponding sub-block under the small scale;

(7.3) according to the splicing algorithm in the step (6.3), carrying out thermal radiation fingerprint on each subblock under a small scale

(s＝1，2，L，M₃×M₃) Splicing to obtain a degraded image f_kThermal radiation fingerprints at small scale.

(8) Analyzing the obtained thermal radiation fingerprint under small scale

If the area with relatively large error still exists, the selected area with relatively large error can be analyzed in a smaller scale such as a microscale according to the steps to obtain the thermal radiation fingerprint under the corresponding scale.

(9) The optical axis offset (Vx) under certain pneumatic thermal environment can be obtained by integrating the steps (2), (5), (6), (7) and (8)_k，Vy_k) And establishing a multi-scale pneumatic thermal radiation image correction fingerprint library by using thermal radiation fingerprints under a plurality of scales such as large scale, medium scale, small scale and the like.

(2) degradation image g for arbitrary aerodynamic heat radiation_kCorrection of fingerprints in pneumatic thermal radiation imagesTemperature T is obtained in the library_kAnd pressure P_kMultiple scale thermal radiation fingerprint under conditions

(scale max, mid, min) and corresponding optical axis offset (Vx)_k，Vy_k) Correcting the aerodynamic heat radiation degradation image to obtain correction results g 'of the aerodynamic heat radiation degradation image at a plurality of scales'_k，scale；

Performing multi-scale correction on the aerodynamic heat radiation degradation image according to the steps (2.1) to (2.2), wherein the specific process is as follows:

(2.1) utilizing the obtained optical axis offset (Vx)_k，Vy_k) Obtaining a registered aerodynamic heat radiation degradation image g'_k(x，y)＝g_k(x-Vx_k，y-Vy_k)；

(2.2) performing multi-scale correction on the pneumatic heat radiation degradation image according to requirements, and if the requirements on local details of the corrected image are not high, performing multi-scale correction on the pneumatic heat radiation degradation image g_kCorrecting for large scale, and mixing g'_kSubtracting the large-scale thermal radiation fingerprint obtained from the fingerprint library

Obtaining a pneumatic thermal radiation degradation image g_kCorrected image at large scaleThe image g can be degraded for aerodynamic thermal radiation if the requirements on the local detail of the corrected image are high_kCorrecting for medium scale or small scale, and mixing g'_kSubtracting the mesoscale thermal radiation fingerprint obtained from the fingerprint library

Obtaining a pneumatic thermal radiation degradation image g_kCorrected image at mesoscale

Or g'_kSubtracting the small-scale thermal radiation fingerprint obtained from the fingerprint libraryObtaining a pneumatic thermal radiation degradation image g_kCorrecting images at small scales

Claims

1. A multi-scale modeling method for an aerodynamic thermal radiation image is characterized in that a window thermal radiation fingerprint library under a corresponding aerodynamic thermal environment is obtained by performing curved surface approximation fitting on difference images of images in an aerodynamic thermal radiation degradation image sequence under multiple scales, and the method specifically comprises the following steps:

(1) respectively registering each pneumatic thermal radiation degraded image according to the reference image, and solving the difference value between each degraded image and the reference image to obtain the optical axis offset and the difference image of each degraded image;

(2) fitting each difference image in a full image area, namely a first scale to obtain a fitting curved surface in the first scale, namely a window thermal radiation fingerprint in the first scale;

(3) refining the last fitting scale, namely dividing the last fitting surface into a plurality of block areas, calculating the fitting error of each block area, if the fitting error of any one block area is not less than the preset error threshold value under the current refining scale, fitting the block areas again to obtain the fitting surface of the block area, namely the window thermal radiation fingerprint of the block area under the current refining scale, further obtaining the window thermal radiation fingerprint of the whole aerodynamic thermal radiation degradation image under the refining scale, and repeating the process until the fitting errors of the refined block areas are less than the error threshold values under the corresponding refining scales; otherwise, the window thermal radiation fingerprint under the last fitting scale is taken as the window thermal radiation fingerprint of the pneumatic thermal radiation degradation image under the refinement scale;

through the steps, the window thermal radiation fingerprints under multiple scales are obtained, namely a window thermal radiation fingerprint library of the pneumatic thermal radiation degradation image sequence under the corresponding pneumatic thermal environment is formed.

2. The method according to claim 1, wherein the fitting is performed by the following specific process: and sampling difference points of the difference image in a region corresponding to the fitting scale, and performing polynomial surface fitting approximation on the sampled difference points.

3. The method according to claim 1 or 2, wherein the fitting error of the block area refers to a relative error of the block area, which is a ratio of a sum of absolute differences between points of the fitted surface on the block area at the last fitting scale and corresponding points in the difference image to a sum of points on the corresponding reference image.

4. The method according to claim 1 or 2, wherein the error threshold value of the current refinement scale refers to a ratio of a sum of fitting errors of each partitioned area subjected to partitioning at the current refinement scale at the last fitting scale to the number of partitioned areas.

5. The method according to claim 1 or 2, wherein in the step (3), the window thermal radiation fingerprint of the block region with the fitting error smaller than the error threshold is the window thermal radiation fingerprint obtained by fitting the region at the last scale, and the window thermal radiation fingerprint of the whole aerodynamic thermal radiation degradation image at the refined scale is formed by mutually splicing the window thermal radiation fingerprints of the block regions at the refined scale.

6. The method of claim 1 or 2, wherein the fitting is a least squares fit.

7. Method for correcting an aerodynamic thermal radiation degradation image using a method for multi-scale modeling of an aerodynamic thermal radiation image according to any of claims 1 to 6, comprising the steps of:

(1) inputting an aerodynamic thermal radiation degradation image to be corrected;

(2) acquiring an aerodynamic thermal radiation image correction fingerprint library by using the method of any one of claims 1 to 6, acquiring a thermal radiation fingerprint of the degraded image in an aerodynamic thermal environment and a corresponding optical axis offset from the fingerprint library, and correcting the degraded image, namely firstly registering the degraded image by using the optical axis offset, and subtracting the thermal radiation fingerprint under the corresponding scale from the registered image to obtain a correction result under the corresponding scale.

8. The method according to claim 7, characterized in that the respective scale is determined according to the accuracy requirements of the image to be corrected.