CN101957331A - Phase change thermography display system based on image processing algorithm - Google Patents

Abstract

The invention provides a phase change thermography display system based on an image processing algorithm, which comprises an initial image preprocessing module, a phase change line extraction module and a heat map display module. The initial image preprocessing module also comprises a sequence image position matching module used for adjusting the initial sequence image position offset generated by model offset; the phase change line extraction module adopts a phase change line tracking algorithm with fusion shape priori characteristics to extract a model phase change line from a model initial contour; the algorithm adopts a C-V active contour model improved from the traditional C-V active contour model, based on differential message and having the fusion shape priori characteristics. The accuracy of the image inter-frame differential is improved by using the sequence image position matching module of the invention, energy curve evolution of the differential change information is carried out by using the phase change line tracking algorithm with the fusion shape priori characteristics, and therefore the phase change lines in sequence images can be effectively extracted from various model types.

Description

Phase transformation thermal map display system based on image processing algorithm

Technical field

The present invention relates to a kind of phase transformation thermal map display system, relate in particular to a kind of phase transformation thermal map display system based on image processing algorithm.

Background technology

Complicated along with international situation, China develops high performance strategic arms and aerospacecraft has become and urgent task.When aircraft flew in the atmospheric envelope high speed, because shock wave compression and surface viscosity friction, the part of the huge kinetic energy of aircraft was transformed into the heat energy of air, and aircraft ambient air temperature can be very high, and aircraft surface is heated, promptly pneumatic heating.Long-time high-speed flight in atmospheric envelope, the local pneumatic heating that aircraft is suffered and total heating are all very big, so aircraft must carry out the thermal protection struc ture design.The basis of thermal protection struc ture design is accurately to understand pneumatic heating parameters, and wind tunnel test is the important means of understanding pneumatic heating parameters, carrying out pneumatic heat study.

In the various calorimetric means of wind-tunnel, conventional spot measurement technology such as thermopair can not provide abundant point to describe pneumatic heating parameters all sidedly, so phase transformation thermal map experimental technique is generally adopted.Phase transformation thermal map experimental technique not only can realize large-area heat mapping according to the temperature of the phase transformation judgement body surface of coating, and can generally investigate the pneumatic heat rate distribution of whole test thing.With respect to infrared technique, this technology has that use cost is low, convenient, flexible characteristic, and is not subjected to the influence of environment temperature.Phase transformation thermal map experimental technique test macro block diagram is seen accompanying drawing 1.With high-resolution CCD camera and video capture card, with the speed collection model surface phase transformation image of per second 25 frames, captured in real time model surface phase transformation course.The CCD camera is selected the camera lens of proper focal length for use, makes the model image suitable size that collects, and can obtain figure very clearly after the focusing.

The detailed process of heat of transformation wind tunnel test is: be coated with last layer phase transformation lacquer (fixedly melting temperature T equably on the surface of experiment thing _o), phase transformation lacquer is a White-opalescent when not undergoing phase transition, and (because coating is very thin, can think coating phase transition temperature T when temperature is elevated to phase transition temperature _oBe mould walls surface temperature T _w), just undergo phase transition and become the transparent body.Each phase transformation lacquer, its phase transition temperature is fixed.If the phase transformation lacquer is applied to the model surface that isotropic thermal insulation material is made, model surface will become white so, in process of the test, because be in the model surface nonuniform heating in the wind-tunnel, the position that has reaches phase transition temperature earlier, undergoes phase transition, show the true colors of model, the position that has does not have phase transformation because temperature is low, still is white.Therefore transparent part and opaque section form the separatrix, and this line promptly is a phase change line, and the temperature of its representative is a phase transition temperature, and as time passes, other position of model also the series of phase transitions line will occur.T on phase change line _w=T _oNote the image (promptly different phase change line position constantly) that each position of model surface undergoes phase transition course by image capturing system, carry out Flame Image Process and extract phase change line, again according to parameters such as the pneumatic heating coefficient on thermal map relational expression computation model surface and hot-fluids, what obtain model surface pneumaticly adds heat distribution.

Doctors Wang Xiaonian of Xi'an Communications University etc. have proposed a kind of extraction method at the comparatively simple heat of transformation sequence image of model I surface condition.This method is transformed into one to the original series image by the time-space conversion and is combined into image, in the composograph one row have write down the phase transition process of any on the characteristic image, and wherein the position (position of cut-off rule) of vertical gray scale acute variation is exactly the phase transformation position on the characteristic image.But this method is based on following prerequisite: at first, phase change line does not have change in topology, and its direction of motion is simple one-way movement.Secondly, the image sequence conversion regime is easy to artificial definite, and composograph has the phase change line feature that is better than former figure.Therefore, present algorithm can only be realized effectively cutting apart at a certain class phase transformation thermal map, and needs the artificial image transitions mode of determining.Thereby limited the diversified demand of phase transformation thermal map test macro model.

Summary of the invention

The invention provides a kind of system that can accurately carry out the demonstration of phase transformation thermal map to the multiclass model of phase transformation thermal map test macro.

A kind of phase transformation thermal map display system based on image processing algorithm, comprise: the initial pictures pretreatment module, the display module of phase change line extraction module, thermal map spectrum, wherein, the initial pictures pretreatment module is used for the initiation sequence image of the performance model surface phase transformation of CCD camera and video capture card collection is carried out pre-service, comprising:

Sequence image location matches module is used to adjust the initiation sequence picture position skew that is produced because of the model skew, forms normalized sequence image;

Initial profile is chosen module, is used for choosing the model initial profile from normalized sequence image;

Described phase change line extraction module adopts the phase change line track algorithm that merges the shape prior characteristic, extraction model phase change line from the model initial profile; The phase change line track algorithm of described fusion shape prior characteristic adopts the C-V active contour model based on difference information and fusion shape prior characteristic, by phase change line extraction algorithm extraction model phase change line from the model initial profile;

The display module of described thermal map spectrum comprises:

The model parameter load module, by the corresponding relation of model and model initial profile size, the adjustment of implementation model parameter input;

The thermal map spectrum is drawn and output module, draws out the thermal map of model by model phase change line and model parameter and composes the line output of going forward side by side.

In wind tunnel test, the phase transition process of model (trier) is very short, generally need make the wind tunnel test environment reach steady state (SS) earlier, and the drop-test thing is to testing the position again.And in launch process, also there is phase transformation in trier.Therefore the video data that obtains at experimentation is the initiation sequence image that comprises the trier motion, so several frames of beginning that need earlier trier to be transferred in the process carry out the position adjustment, guarantees the consistance of test object location in the sequence image.

Sequence image location matches module of the present invention is carried out position adjustment by the image matching algorithm based on architectural feature to the initiation sequence image according to the architectural feature of model itself, and forms normalized sequence image through rename.Can override the initiation sequence image because adjust the sequence image of position, need the sequence image of adjusting good position is carried out rename.The initiation sequence image is the most initial do not have the frame of phase transformation be first frame as normalized sequence image sequence number 1 (frame that does not have phase transformation fully), the normalized sequence image sequence number of postorder is 2.

Described image matching algorithm based on architectural feature is: establish start frame image characteristic matrix p _uExpression, a certain eigenmatrix q in zone to be matched _u(q _u, p _uHave identical space size) expression, this paper puies forward the edge feature statistical indicator and is:

$\mathrm{\θ}(p,q)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}(\sqrt{{({\mathrm{\ϵ}}_1*p_u)}^2+{({\mathrm{\ϵ}}_2*p_u)}^2}-\sqrt{{({\mathrm{\ϵ}}_1*q_u^i)}^2+{({\mathrm{\ϵ}}_2*q_u^i)}^2})---\left(4\right)$

Wherein, ${\mathrm{\ϵ}}_1=\left[\begin{array}{ccc}-1& -1& -1\\ 0& 0& 0\\ 1& 1& 1\end{array}\right],$ ${\mathrm{\ϵ}}_2=\left[\begin{array}{ccc}-1& 0& 1\\ -1& 0& 1\\ -1& 0& 1\end{array}\right],$ M is that zone to be matched contains q _uNumber, matrix p _uAnd q _uWith ε ₁, ε ₂Multiply each other and can be split as a series of 3 * 3 matrix.Two correspond respectively to trier marginal information in target area and the zone to be matched in the formula (4).When the pixel that is characterized as u all appears at q in the zone to be matched _uThe time, (p q) gets minimum value to θ, and this moment, feature u was the highest to this regional degree of support.

The target of the sequence image that is obtained in the heat of transformation project is exactly the contour curve of model, and other parts all are background, and the relative background of target has specific gray-scale value scope.The purpose of image segmentation is separated interested part in the sequence image exactly from background, to carry out follow-up processing.The degree of cutting apart depends on the problem that will solve.In the present invention, need obtain the initial profile of model in the binaryzation picture.In order to acquire the contour curve of binaryzation picture, adopt the edge finding algorithm that profile extracts and profile is followed the tracks of.Profile extracts and profile is followed the tracks of very simple for binary picture profile extraction algorithm, the method that profile extracts is emptied internal point exactly: if having among the former figure a bit for black, and when its 8 consecutive point all are black (this moment, this point was an internal point), then with this point deletion.The basic skills that profile is followed the tracks of is: find out the pixel on the target object profile according to some strictness " detection criterion " earlier, find out other pixels on the target object according to some features of these pixels with certain " tracking criterion " again, the specific algorithm that the binary picture profile is followed the tracks of is: at first find first boundary pixel, " detection criterion " is: according to from left to right, sequential search from top to bottom, first stain that finds must be the frontier point of lower left, is designated as A.Have at least one to be frontier point in its right side, upper right, last, upper left four adjoint points, be designated as B.Begin to look for from B, find the frontier point C in the consecutive point by right, upper right, last, upper left, left, lower-left, order following, the bottom right.If C is exactly A, then show to make a circle EOP (end of program); Otherwise continue to look for from the C point, till finding A.Judge whether frontier point is easy to: if its four adjoint points up and down are not that stain promptly is frontier point (promptly following the tracks of criterion).

This algorithm will judge that calculated amount is bigger to eight points around each boundary pixel, and the algorithm that another kind of binary picture profile is followed the tracks of is: the frontier point that at first finds lower left according to top said " detection criterion ".Initial with this frontier point, suppose to have found all frontier points around entire image one circle along clockwise direction.Because the border is continuous, all each frontier points can be represented with the angle that this frontier point is opened previous frontier point.Therefore can use following tracking criterion: from first frontier point, defining the initial direction of search is along the upper left side: if upper left point is a stain, then be frontier point, and 45 degree otherwise the direction of search turns clockwise.Like this until find till first stain, then this stain as new frontier point, on the basis of current search direction, be rotated counterclockwise 90 degree, continue to use the same method and continue the next stain of search, know and return till the initial frontier point.

The phase transition process of complicated phase transformation trier is not single progressive formation, there is tangible change in topology in phase change line, Level Set Method is the effective ways that solve the change in topology problem, but its curve evolvement and search speed are unsatisfactory, consider the curve evolvement feature that phase change line changes, the present invention improves on the basis of the C-V active contour model of binding curve evolution and level set theory, proposed a kind ofly based on difference information and merge the C-V active contour model of shape prior feature, adopted based on difference information and the phase change line track algorithm of fusion shape prior characteristic of C-V active contour model that merges the shape prior feature to improve processing speed.Because,, very consuming time if respectively to every two field picture individual processing.In fact, have shape similarity and relevance between the sequence image phase change line, the phase change line of prior image frame provides priori for the postorder phase change line.So, make this algorithm become possibility.

Described C-V active contour model based on difference information and fusion shape prior characteristic is:

Suppose that Ω is R ²In the bounded opener,

Be the border of opener, u ₀: Ω → R is given image, supposes image u ₀Be by there being two gray-scale values to be similar to constant c respectively ₁And c ₂The zone form, further the contours of objects that will cut apart of supposition is C _o, the point value of object is u ₀(x y), in object inside, has u so ₀=c ₁, and u is arranged in the object outside ₀=c ₂, then be based on difference information and the energy function that merges the C-V active contour model of shape prior characteristic:

$E(c_1,c_2,\mathrm{\φ})={\mathrm{\μ}}_2\∫\frac{1}{2}{\left(\right|\▿\mathrm{\φ}(x,y)|-1)}^2\mathrm{dxdy}+{\mathrm{\μ}}_1\underset{\mathrm{\Ω}}{\∫}\mathrm{\δ}\left(\mathrm{\φ}(x,y)\right)|\▿\mathrm{\φ}(x,y)|\mathrm{dxdy}+\mathrm{\α}\∫{(H\left(\mathrm{\φ}\right)-H\left({\mathrm{\φ}}_l\right))}^2\mathrm{dx}---\left(1\right)$

$+{\mathrm{\λ}}_1\underset{\mathrm{\Ω}}{\∫}{|I_0(x,y)-c_1|}^2\left(H\left(\mathrm{\φ}(x,y)\right)\right)\mathrm{dxdy}+{\mathrm{\λ}}_2\underset{\mathrm{\Ω}}{\∫}{|I_0(x,y)-c_2|}^2(1-H\left(\mathrm{\φ}(x,y)\right))\mathrm{dxdy}$

Wherein, (x y) is level set function, I to φ ₀(x, y)=((x, y)-k (x, y)), (x y) is initial pictures in the normalized sequence image to k to g to γ, and (x y) is postorder phase transformation image in the normalized sequence image to g, and γ is the coefficient that contrast strengthens, u ₁, u ₂, α, λ ₁, λ ₂Be every coefficient constant.According to the Euler-Lagrange equation partial differential equation that the level set function φ of formula (1) minimization satisfies of sening as an envoy to of deriving:

$\left\{\begin{array}{c}{c}_1\left(\mathrm{\φ}\right)=\frac{\underset{\mathrm{\Ω}}{\∫}{I}_0(x,y)H\left(\mathrm{\φ}(x,y)\right)\mathrm{dxdy}}{\underset{\mathrm{\Ω}}{\∫}H\left(\mathrm{\φ}(x,y)\right)\mathrm{dxdy}}\\ c_2\left(\mathrm{\φ}\right)=\frac{\underset{\mathrm{\Ω}}{\∫}{I}_0(x,y)(1-H\left(\mathrm{\φ}(x,y)\right))\mathrm{dxdy}}{\underset{\mathrm{\Ω}}{\∫}(1-H\left(\mathrm{\φ}(x,y)\right))\mathrm{dxdy}}\\ \frac{\∂\mathrm{\φ}}{\∂t}=\mathrm{\δ}\left(\mathrm{\φ}\right)({\mathrm{\μ}}_1\mathrm{div}\left(\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)-{\mathrm{\λ}}_1{(I_0(x,y)-c_1)}^2+{\mathrm{\λ}}_2{(I_0(x,y)-c_2)}^2)\\ +u_2({\▿}^2\mathrm{\φ}-\mathrm{div}\left(\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right))+\mathrm{\α}\·\mathrm{\δ}\left(\mathrm{\φ}\right)\·(H\left(\mathrm{\φ}\right)-H\left({\mathrm{\φ}}_l\right)),(0,\∞)\×\mathrm{\Ω}\\ \mathrm{\φ}(0,x,y)={\mathrm{\φ}}_0(x,y),(x,y)\∈\mathrm{\Ω}\end{array}\right.---\left(2\right)$

Wherein, φ is the level set function of present frame, φ _lBe the level set function of preceding frame, φ ₀Be the level set function of start frame, H (x), δ (x) is respectively Heaviside function and Dirac function, in actual computation, gets respectively:

ε is a variable, and concrete value is 1;

The phase change line track algorithm of described fusion shape prior characteristic may further comprise the steps:

1) chooses the start frame of normalized sequence image;

2) by interval N frame, choose the postorder frame automatically,, obtain the effective information of differential variation, finish initialization level set function φ by the adjustment of difference and contrast;

3) calculate c according to formula (2) ₁(φ _{I, j} ⁿ), c ₂(φ _{I, j} ⁿ), then by calculating

${\mathrm{\φ}}_{i,j}^{n+1}={\mathrm{\φ}}_{i,j}^n+\mathrm{\τ}(\mathrm{\δ}\left({\mathrm{\φ}}_{i,j}^n\right)({\mathrm{\μ}}_1\mathrm{div}\left(\frac{\▿{\mathrm{\φ}}_{i,j}^n}{|\▿{\mathrm{\φ}}_{i,j}^n|}\right)-{\mathrm{\λ}}_1{(I_0(x,y)-c_1)}^2+$

${\mathrm{\λ}}_2{(I_0(x,y)-c_2)}^2)+u_2({\▿}^2{\mathrm{\φ}}_{i,j}^n-\mathrm{div}\left(\frac{\▿{\mathrm{\φ}}_{i,j}^n}{|\▿{\mathrm{\φ}}_{i,j}^n|}\right))+\mathrm{\α}\·\mathrm{\δ}\left({\mathrm{\φ}}_{i,j}^n\right)\·(H\left({\mathrm{\φ}}_{i,j}^n\right)-H\left({\mathrm{\φ}}_{i,j}^{n-1}\right)))---\left(3\right)$ Obtain the level set function φ of back frame _{I, j} ^N+1τ is an iteration step length, evolution level set function, φ _{I, j} ⁿBe the level set function of present frame, φ _{I, j} ^N+1Be the level set function of back frame, φ _{I, j} ^N-1Level set function for preceding frame.

4) calculate condition of convergence Q=∑ _{I, j}| φ _{I, j} ^N+1-φ _{I, j} ⁿ|, this formula is represented the situation of change on the profile.Get threshold xi, if Q＜ξ, then think convergence, iteration stopping; Otherwise then returning step 3) continues to handle.

5) differentiating the postorder frame is last frame, if then withdraw from, obtains the phase transition process and the accurate model phase change line of all required normalized sequence images; Be not then to return step 2) continue to move.

Wherein, step 2) difference in is meant inter-frame difference, is undertaken between the initiation sequence picture frame having guaranteed that the change procedure of heat of transformation sequence image is normalized to the disappearance change procedure of trier surface phase transformation lacquer after the location matches by sequence image location matches module.So can take first frame with normalized sequence image is start frame, the method for postorder picture and its difference obtains the effective information that normalized sequence image changes.

Step 2) the contrast adjustment in be because, if under under-exposed or excessive situation since the brightness range of image is not enough or non-linear meeting to make the contrast of image be not very desirable, gradation of image may be confined in the very little scope.At this moment people see will be one smudgy, as if do not have the image of level, the method that the available pixel amplitude is redistributed is improved the contrast of image.The gray scale correction is divided into following three kinds: linear transformation, piecewise linear transform and nonlinear transformation.(referring to Seeking Truth science and technology .Visual C++ Digital Image Processing typical algorithm and realization [M]. the People's Telecon Publishing House, 2007, P117-131)

Adopt traditional C-V active contour model to need every two field picture reconstruct level set function φ in the calculation process, this is a step step very consuming time.In actual applications, there is association in the phase change line of frame before and after the phase transformation sequence image, and preceding frame phase change line can provide prior imformation for the latter.For this reason, the present invention is from initialization, the reconstruct of level set function φ and how to merge phase change line shape facility aspect the C-V active contour model is improved to improve the extraction rate of phase change line.

Initialization φ: the phase transformation hot line that former frame extracted is a benchmark, fast initialization level set function φ.Extract the phase change line that obtains if η is preceding frame, calculate each pixel coordinate point and closed curve η in the current image frame apart from d (x, y), by calculate φ=± (x, y) (preceding frame profile is inner for negative, outside for just) obtains level set function to d.Because heat of transformation graphic sequence is a progressive formation, between the single frames difference very little, frame is initial profile φ before selecting, and has significantly reduced the iterations of search, has reduced operation time.

After preceding frame phase change line is obtained, can be used as the priori of present frame.Structure phase change line shape prior fundamental function [11]:

E _initial(φ，φ _l)＝∫(H(φ)-H(φ _l)) ²dx (5)

φ is the level set function of present frame in the formula, φ _lLevel set function for preceding frame.

Need not reconstruct φ: in order to improve the evolution speed of algorithm, CM Li has proposed a level set coordination function based on Level Set function, can need not to consider the reconstruct problem of level set function, and its formula is as follows:

$\mathrm{\ρ}\left(\mathrm{\φ}\right)=\frac{1}{2}{\left(\right|\▿\mathrm{\φ}(x,y)|-1)}^2\mathrm{dxdy}---\left(6\right)$

Can solve level set function reconstruct problem promptly according to this calculation function: level set function needn't be reconstructed into level set function, improves the speed of curve evolvement.

This paper convolution (5) and formula (6) have proposed (1) formula.

The drafting of described phase transformation thermal map collection of illustrative plates is specially: every two field picture is calculated, obtain on the phase change line each pixel at the coordinate figure X on realistic model surface, Y, parameters such as the pneumatic heating coefficient value of Z, pixel position, heat flow value, integrated all phase transformation images, and pneumatic heating coefficient value in model surface position of not noting down of match, heat flow value etc.If repeatedly local test result inlays local result.Thermal map spectrum paint type:

(1) full model phase transformation test thermal map spectrum is drawn;

(2) model surface localized heat collection of illustrative plates is drawn;

(3) selected characteristic phase transformation thermal image collection of illustrative plates is drawn;

(4) a plurality of localized heat collection of illustrative plates of model surface are inlayed the combination drafting on the full model surface;

(5) the localized heat collection of illustrative plates is mounted to full model surface heat collection of illustrative plates.

Because of model is made with thermal insulation material, and dope layer is very little with respect to model diameter, so phase transformation thermometric problem can be summed up as end heating, the semiinfinitepiston One dimensional unsteady heat conduction problem of sidewall thermal insulation, and its differential equation is:

$\frac{{\∂}^{2}T}{\∂x^2}=\mathrm{\ω}\frac{\∂T}{\∂t}---\left(8\right)$

Starting condition: t=0, T=T _i

Boundary condition: $x=0,-k\frac{\∂T}{\∂x}=h(T_i-T_w);$ $x\→\∞,\frac{\∂T}{\∂x}=0$

ω=ρ c/k wherein; ρ, c, k are respectively density, specific heat and the coefficient of heat conductivity of model; The t express time; T represents temperature; Vertical and the trier surface of x direction.

Given certain boundary condition and starting condition, can obtain the One dimensional unsteady heat-conduction equation separate for:

$\frac{T_w-T_i}{T_{\mathrm{aw}}-T_i}=1-e^{{\mathrm{\β}}^2}\mathrm{erfc}\left(\mathrm{\β}\right)---\left(9\right)$

Wherein the complementary error function definition is: $\mathrm{erfc}\left(\mathrm{\β}\right)=1-\frac{2}{\sqrt{\mathrm{\π}}}{\∫}_0^{\mathrm{\β}}{e}^{-t^2}\mathrm{dt}---\left(10\right)$

The complementary error function needs this expression formula is launched:

$\mathrm{erfc}\left(\mathrm{\β}\right)=1-\frac{2}{\sqrt{\mathrm{\π}}}\left(\underset{i=0}{\overset{N}{\mathrm{\Σ}}}{e}^{-{\mathrm{\λ}}_i^2}\mathrm{\Δ}{\mathrm{\λ}}_i\right)---\left(11\right)$

Wherein ${\mathrm{\λ}}_i=\frac{i}{N}\mathrm{\β},$ $\mathrm{\Δ}{\mathrm{\λ}}_i=\frac{1}{N}\mathrm{\β}.$ Then formula (9) can become:

$1-\frac{T_w-T_i}{T_{\mathrm{aw}}-T_i}-e^{{\mathrm{\β}}^2}(1-\frac{2}{\sqrt{\mathrm{\π}}}\left(\underset{i=0}{\overset{N}{\mathrm{\Σ}}}{e}^{-{\mathrm{\λ}}_i^2}\mathrm{\Δ}{\mathrm{\λ}}_i\right))=0---\left(12\right)$

The method of utilizing numerical value to approach:

$f(1-\frac{T_w-T_i}{T_{\mathrm{aw}}-T_i}-e^{{\mathrm{\&beta;}}^2}(1-\frac{2}{\sqrt{\mathrm{\&pi;}}}\left(\underset{i=0}{\overset{N}{\mathrm{\&Sigma;}}}{e}^{-{\mathrm{\&lambda;}}_i^2}\mathrm{\&Delta;}{\mathrm{\&lambda;}}_i\right)))<0.0000000001---\left(13\right)$

When satisfying formula (13), the β value that has calculated realistic size is described.

$\mathrm{\&beta;}=h\sqrt{t}/\sqrt{\mathrm{\&rho;ck}}---\left(14\right)$

H represents hot change of current coefficient; T _wBe the model surface temperature, can think the phase transition temperature that equals material; T _iInitial temperature for model; Model adiabatic wall temperature T _Aw=0.9T ₀T wherein _Aw, T _w, T _i, ρ, c, k are all known, use the method for iteration to obtain the optimum coefficient of heat transfer.

Can draw instantaneous surface heat flow q (t) and T according to following formula again _wRelational expression.

q＝h(T _aw-T _w) (15)

Pneumatic heating coefficient value such as every of thermal map spectrum or heat flow value line (comprise information: frame number, time, h, q, h/h with line style or color differentiating _s(T ₀), h/h _s(0.9T ₀), q/q _s(T ₀), q/q _s(0.9T ₀) or other parameter (can choose)); Each thermal map spectrum mark T _PC, the test condition parameter, as (can choose) such as test train number, Mach number, Reynolds number, stagnation temperature, stagnation pressures.

The present invention adopts sequence image location matches module, the initiation sequence image is carried out the normalization adjustment, so that improve the accuracy of image inter-frame difference, the parted pattern that merges priori based on difference information carries out energy trace to differential variation information and develops, and can be effectively multiple version be extracted phase change line in the sequence image.

Description of drawings

Fig. 1 is a phase transformation thermal map experimental technique test macro block diagram.

Fig. 2 is a phase transformation thermal map display system block diagram of the present invention.

Fig. 3 is the extraction result schematic diagram of model phase change line of the present invention.

Fig. 4 is model thermal map spectrum synoptic diagram of the present invention.

Embodiment

As shown in Figure 1, be phase transformation thermal map experimental technique test macro block diagram, comprise dummy vehicle, high-resolution CCD camera, video capture card and image processing system, thermal map spectral analysis system, wherein, CCD camera and video capture card are with the speed collection model surface phase transformation image of per second 25 frames, captured in real time model surface phase transformation course.The CCD camera is selected the camera lens of proper focal length for use, makes the model initiation sequence image suitable size that collects, and can obtain figure very clearly after the focusing.The photo that each model collected among the embodiment about the 2G size, about 80 seconds consuming time of 2000 plurality of pictures.

Accompanying drawing 2 is phase transformation thermal map display system block diagram of the present invention, and phase transformation thermal map display system comprises: initial pictures pretreatment module, the display module of phase change line extraction module, thermal map spectrum; Wherein the initial pictures pretreatment module comprises that sequence image location matches module, initial profile choose module; The display module of described thermal map spectrum comprises: model parameter load module and thermal map spectrum are drawn and output module.

At first use sequence image location matches module that the model initiation sequence image of CCD camera and video capture card collection is carried out the normalization adjustment, obtain normalized sequence image.

Described image matching algorithm based on architectural feature is: establish start frame eigenmatrix p _uExpression, a certain eigenmatrix q in zone to be matched _uExpression, q _u, p _uHave identical space size, the edge feature statistical indicator is:

$\mathrm{\&theta;}(p,q)=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}(\sqrt{{({\mathrm{\&epsiv;}}_1*p_u)}^2+{({\mathrm{\&epsiv;}}_2*p_u)}^2}-\sqrt{{({\mathrm{\&epsiv;}}_1*q_u^i)}^2+{({\mathrm{\&epsiv;}}_2*q_u^i)}^2})---\left(4\right)$

Wherein ${\mathrm{\&epsiv;}}_1=\left[\begin{array}{ccc}-1& -1& -1\\ 0& 0& 0\\ 1& 1& 1\end{array}\right],$ ${\mathrm{\&epsiv;}}_2=\left[\begin{array}{ccc}-1& 0& 1\\ -1& 0& 1\\ -1& 0& 1\end{array}\right],$ M is that zone to be matched contains q _uNumber, matrix p _uAnd q _uWith ε ₁, ε ₂Multiply each other and can be split as a series of 3 * 3 matrix.Two correspond respectively to trier marginal information in target area and the zone to be matched in the formula (1).When the pixel that is characterized as u all appears at q in the zone to be matched _uThe time, (p q) gets minimum value to θ, and this moment, feature u was the highest to this regional degree of support.

Carry out the debugging test result in conjunction with raw data and show that algorithm has very strong robustness, and computational solution precision height very, to coincide with the actual data that obtain, the coupling deviation is very little.

The utilization initial profile is chosen module, adopts the algorithm that profile extracts and profile is followed the tracks of to choose the model initial profile from normalized sequence image.

Use the phase change line extraction module to adopt the phase change line track algorithm that merges the shape prior characteristic, extraction model phase change line from the model initial profile again.

The phase change line track algorithm of described fusion shape prior characteristic adopts the C-V active contour model based on difference information and fusion shape prior characteristic, and described C-V active contour model based on difference information and fusion shape prior characteristic is:

Suppose that Ω is R ²In the bounded opener, Be the border of opener, u ₀: Ω → R is given image, supposes image u ₀Be by there being two gray-scale values to be similar to constant c respectively ₁And c ₂The zone form, further the contours of objects that will cut apart of supposition is C _o, the point value of object is u ₀(x y), in object inside, has u so ₀=c ₁, and u is arranged in the object outside ₀=c ₂, then be based on difference information and the energy function that merges the C-V active contour model of shape prior characteristic:

$E(c_1,c_2,\mathrm{\&phi;})={\mathrm{\&mu;}}_2\&Integral;\frac{1}{2}{\left(\right|\&dtri;\mathrm{\&phi;}(x,y)|-1)}^2\mathrm{dxdy}+{\mathrm{\&mu;}}_1\underset{\mathrm{\&Omega;}}{\&Integral;}\mathrm{\&delta;}\left(\mathrm{\&phi;}(x,y)\right)|\&dtri;\mathrm{\&phi;}(x,y)|\mathrm{dxdy}+\mathrm{\&alpha;}\&Integral;{(H\left(\mathrm{\&phi;}\right)-H\left({\mathrm{\&phi;}}_l\right))}^2\mathrm{dx}---\left(1\right)$

$+{\mathrm{\&lambda;}}_1\underset{\mathrm{\&Omega;}}{\&Integral;}{|I_0(x,y)-c_1|}^2\left(H\left(\mathrm{\&phi;}(x,y)\right)\right)\mathrm{dxdy}+{\mathrm{\&lambda;}}_2\underset{\mathrm{\&Omega;}}{\&Integral;}{|I_0(x,y)-c_2|}^2(1-H\left(\mathrm{\&phi;}(x,y)\right))\mathrm{dxdy}$

Wherein, (x y) is level set function, I to φ ₀(x, y)=((x, y)-k (x, y)), (x y) is initial pictures in the normalized sequence image to k to g to γ, and (x y) is postorder phase transformation image in the normalized sequence image to g, and γ is the coefficient that contrast strengthens, u ₁, u ₂, α, λ ₁, λ ₂Be every coefficient constant.According to the Euler-Lagrange equation partial differential equation that the level set function φ of formula (1) minimization satisfies of sening as an envoy to of deriving:

$\left\{\begin{array}{c}{c}_1\left(\mathrm{\&phi;}\right)=\frac{\underset{\mathrm{\&Omega;}}{\&Integral;}{I}_0(x,y)H\left(\mathrm{\&phi;}(x,y)\right)\mathrm{dxdy}}{\underset{\mathrm{\&Omega;}}{\&Integral;}H\left(\mathrm{\&phi;}(x,y)\right)\mathrm{dxdy}}\\ c_2\left(\mathrm{\&phi;}\right)=\frac{\underset{\mathrm{\&Omega;}}{\&Integral;}{I}_0(x,y)(1-H\left(\mathrm{\&phi;}(x,y)\right))\mathrm{dxdy}}{\underset{\mathrm{\&Omega;}}{\&Integral;}(1-H\left(\mathrm{\&phi;}(x,y)\right))\mathrm{dxdy}}\\ \frac{\&PartialD;\mathrm{\&phi;}}{\&PartialD;t}=\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)({\mathrm{\&mu;}}_1\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)-{\mathrm{\&lambda;}}_1{(I_0(x,y)-c_1)}^2+{\mathrm{\&lambda;}}_2{(I_0(x,y)-c_2)}^2)\\ +u_2({\&dtri;}^2\mathrm{\&phi;}-\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right))+\mathrm{\&alpha;}\&CenterDot;\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)\&CenterDot;(H\left(\mathrm{\&phi;}\right)-H\left({\mathrm{\&phi;}}_l\right)),(0,\&infin;)\&times;\mathrm{\&Omega;}\\ \mathrm{\&phi;}(0,x,y)={\mathrm{\&phi;}}_0(x,y),(x,y)\&Element;\mathrm{\&Omega;}\end{array}\right.---\left(2\right)$

Wherein, φ is the level set function of present frame, φ _lBe the level set function of preceding frame, φ ₀Be the level set function of start frame, H (x), δ (x) is respectively Heaviside function and Dirac function, in actual computation, gets respectively:

ε is a variable, and concrete value is 1;

The phase change line track algorithm of described fusion shape prior characteristic may further comprise the steps:

1) chooses the start frame of normalized sequence image;

2) by interval N frame, choose the postorder frame automatically,, obtain the effective information of differential variation, finish initialization level set function φ by the adjustment of difference and contrast;

3) calculate c according to formula (2) ₁(φ _{I, j} ⁿ), c ₂(φ _{I, j} ⁿ), then by calculating

${\mathrm{\&phi;}}_{i,j}^{n+1}={\mathrm{\&phi;}}_{i,j}^n+\mathrm{\&tau;}(\mathrm{\&delta;}\left({\mathrm{\&phi;}}_{i,j}^n\right)({\mathrm{\&mu;}}_1\mathrm{div}\left(\frac{\&dtri;{\mathrm{\&phi;}}_{i,j}^n}{|\&dtri;{\mathrm{\&phi;}}_{i,j}^n|}\right)-{\mathrm{\&lambda;}}_1{(I_0(x,y)-c_1)}^2+$

${\mathrm{\&lambda;}}_2{(I_0(x,y)-c_2)}^2)+u_2({\&dtri;}^2{\mathrm{\&phi;}}_{i,j}^n-\mathrm{div}\left(\frac{\&dtri;{\mathrm{\&phi;}}_{i,j}^n}{|\&dtri;{\mathrm{\&phi;}}_{i,j}^n|}\right))+\mathrm{\&alpha;}\&CenterDot;\mathrm{\&delta;}\left({\mathrm{\&phi;}}_{i,j}^n\right)\&CenterDot;(H\left({\mathrm{\&phi;}}_{i,j}^n\right)-H\left({\mathrm{\&phi;}}_{i,j}^{n-1}\right)))---\left(3\right)$

Obtain the level set function φ of back frame _{I, j} ^N+1τ is an iteration step length, evolution level set function, φ _{I, j} ⁿBe the level set function of present frame, φ _{I, j} ^N+1Be the level set function of back frame, φ _{I, j} ^N-1Level set function for preceding frame;

4) calculate condition of convergence Q=∑ _{I, j}| φ _{I, j} ^N+1-φ _{I, j} ⁿ|, this formula is represented the situation of change on the profile.Get threshold xi, if Q＜ξ, then think convergence, iteration stopping; Otherwise then returning step 3) continues to handle;

5) differentiating the postorder frame is last frame, if then withdraw from, obtains the phase transition process and the accurate model phase change line of all required normalized sequence images; Be not then to return step 2) continue to move.

Wherein the gray scale correction adopts linear single-valued function that each pixel in the image is done linear expansion, and each parameter value is as follows: λ ₁=λ ₂=1, μ ₁=0.02 * 255 ², μ ₂=1, γ=2, α=1, τ=0.1, ξ gets 200000, and N is 5, and the model phase change line of gained is as shown in Figure 3.Model shown in Fig. 3 is divided into former and later two parts, among the figure before and after the phase change line that obtains of two parts there is no mutual phenomenon, and the each several part phase change line has and only has one, meets realistic meaning.

Adopt the model parameter load module again, by the corresponding relation of model and model initial profile size, input model parameter;

Draw and output module by the thermal map spectrum, the thermal map spectrum that model phase change line that utilization is obtained and model parameter are drawn out model, the line output of going forward side by side, the model thermal map spectrum of gained as shown in Figure 4, from accompanying drawing 4 as can be seen, the change procedure of thermal map spectrum is the process of going forward one by one forward monotonously slowly and changing, realistic phase transition process.

The described content of this instructions embodiment only is enumerating the way of realization of inventive concept; protection scope of the present invention should not be regarded as only limiting to the concrete form that embodiment states, protection scope of the present invention also reach in those skilled in the art conceive according to the present invention the equivalent technologies means that can expect.

Claims (5)

1. phase transformation thermal map display system based on image processing algorithm, the display module that comprises initial pictures pretreatment module, phase change line extraction module, thermal map spectrum, it is characterized in that: described initial pictures pretreatment module comprises sequence image location matches module, be used to adjust the initiation sequence picture position skew that is produced because of the model skew, form normalized sequence image; Described phase change line extraction module adopts the phase change line track algorithm that merges the shape prior characteristic, extraction model phase change line from the model initial profile; The phase change line track algorithm of described fusion shape prior characteristic adopts the C-V active contour model based on difference information and fusion shape prior characteristic, and described C-V active contour model based on difference information and fusion shape prior characteristic is:

Suppose that Ω is R ²In the bounded opener,

Be the border of opener, u ₀: Ω → R is given image, supposes image u ₀Be by there being two gray-scale values to be similar to constant c respectively ₁And c ₂The zone form, further the contours of objects that will cut apart of supposition is C _o, the point value of object is u ₀(x y), in object inside, has u so ₀=c ₁, and u is arranged in the object outside ₀=c ₂, then be based on difference information and the energy function that merges the C-V active contour model of shape prior characteristic:

$E(c_1,c_2,\mathrm{\&phi;})={\mathrm{\&mu;}}_2\&Integral;\frac{1}{2}{\left(\right|\&dtri;\mathrm{\&phi;}(x,y)|-1)}^2\mathrm{dxdy}+{\mathrm{\&mu;}}_1\underset{\mathrm{\&Omega;}}{\&Integral;}\mathrm{\&delta;}\left(\mathrm{\&phi;}(x,y)\right)|\&dtri;\mathrm{\&phi;}(x,y)|\mathrm{dxdy}+\mathrm{\&alpha;}\&Integral;{(H\left(\mathrm{\&phi;}\right)-H\left({\mathrm{\&phi;}}_l\right))}^2\mathrm{dx}---\left(1\right)$

$+{\mathrm{\&lambda;}}_1\underset{\mathrm{\&Omega;}}{\&Integral;}{|I_0(x,y)-c_1|}^2\left(H\left(\mathrm{\&phi;}(x,y)\right)\right)\mathrm{dxdy}+{\mathrm{\&lambda;}}_2\underset{\mathrm{\&Omega;}}{\&Integral;}{|I_0(x,y)-c_2|}^2(1-H\left(\mathrm{\&phi;}(x,y)\right))\mathrm{dxdy}$

Wherein, (x y) is level set function, I to φ ₀(x, y)=((x, y)-k (x, y)), (x y) is initial pictures in the normalized sequence image to k to g to γ, and (x y) is postorder phase transformation image in the normalized sequence image to g, and γ is the coefficient that contrast strengthens, u ₁, u ₂, α, λ ₁, λ ₂Be every coefficient constant.According to the Euler-Lagrange equation partial differential equation that the level set function φ of formula (1) minimization satisfies of sening as an envoy to of deriving:

$\left\{\begin{array}{c}{c}_1\left(\mathrm{\&phi;}\right)=\frac{\underset{\mathrm{\&Omega;}}{\&Integral;}{I}_0(x,y)H\left(\mathrm{\&phi;}(x,y)\right)\mathrm{dxdy}}{\underset{\mathrm{\&Omega;}}{\&Integral;}H\left(\mathrm{\&phi;}(x,y)\right)\mathrm{dxdy}}\\ c_2\left(\mathrm{\&phi;}\right)=\frac{\underset{\mathrm{\&Omega;}}{\&Integral;}{I}_0(x,y)(1-H\left(\mathrm{\&phi;}(x,y)\right))\mathrm{dxdy}}{\underset{\mathrm{\&Omega;}}{\&Integral;}(1-H\left(\mathrm{\&phi;}(x,y)\right))\mathrm{dxdy}}\\ \frac{\&PartialD;\mathrm{\&phi;}}{\&PartialD;t}=\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)({\mathrm{\&mu;}}_1\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)-{\mathrm{\&lambda;}}_1{(I_0(x,y)-c_1)}^2+{\mathrm{\&lambda;}}_2{(I_0(x,y)-c_2)}^2)\\ +u_2({\&dtri;}^2\mathrm{\&phi;}-\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right))+\mathrm{\&alpha;}\&CenterDot;\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)\&CenterDot;(H\left(\mathrm{\&phi;}\right)-H\left({\mathrm{\&phi;}}_l\right)),(0,\&infin;)\&times;\mathrm{\&Omega;}\\ \mathrm{\&phi;}(0,x,y)={\mathrm{\&phi;}}_0(x,y),(x,y)\&Element;\mathrm{\&Omega;}\end{array}\right.---\left(2\right)$

Wherein, φ is the level set function of present frame, φ _lBe the level set function of preceding frame, φ ₀Be the level set function of start frame, H (x), δ (x) is respectively Heaviside function and Dirac function, in actual computation, gets respectively:

ε is a variable, and concrete value is 1;

The phase change line track algorithm of described fusion shape prior characteristic may further comprise the steps:

1) chooses the start frame of normalized sequence image;

2) by interval N frame, choose the postorder frame automatically,, obtain the effective information of differential variation, finish initialization level set function φ by the adjustment of difference and contrast;

3) calculate c according to formula (2) ₁(φ _{I, j} ⁿ), c ₂(φ _{I, j} ⁿ), then by calculating

${\mathrm{\&phi;}}_{i,j}^{n+1}={\mathrm{\&phi;}}_{i,j}^n+\mathrm{\&tau;}{\left(\mathrm{\&delta;}\left({\mathrm{\&phi;}}_{i,j}^n\right)({\mathrm{\&mu;}}_1\mathrm{div}\left(\frac{\&dtri;{\mathrm{\&phi;}}_{i,j}^n}{|\&dtri;{\mathrm{\&phi;}}_{i,j}^n}\right)-{\mathrm{\&lambda;}}_1(I_0(x,y)-c_1\right)}^2+$

${\mathrm{\&lambda;}}_2(I_0(x,y)-c_1{)}^2)+u_2({\&dtri;}^2{\mathrm{\&phi;}}_{i,j}^n-\mathrm{div}\left(\frac{\&dtri;{\mathrm{\&phi;}}_{i,j}^n}{|\&dtri;{\mathrm{\&phi;}}_{i,j}^n}\right))+\mathrm{\&alpha;}\&CenterDot;\mathrm{\&delta;}\left({\mathrm{\&phi;}}_{i,j}^n\right)\&CenterDot;(H\left({\mathrm{\&phi;}}_{i,j}^n\right)-H\left({\mathrm{\&phi;}}_{i,j}^{n-1}\right)))---\left(3\right)$

Obtain the level set function φ of back frame _{I, j} ^N+1τ is an iteration step length, evolution level set function, φ _{I, j} ⁿBe the level set function of present frame, φ _{I, j} ^N+1Be the level set function of back frame, φ _{I, j} ^N-1Level set function for preceding frame;

4) calculate the condition of convergence

This formula is represented the situation of change on the profile.Get threshold xi, if Q＜ξ, then think convergence, iteration stopping; Otherwise then returning step 3) continues to handle;

5) differentiating the postorder frame is last frame, if then withdraw from, obtains the phase transition process and the accurate model phase change line of all required normalized sequence images; Be not then to return step 2) continue to move.

2. phase transformation thermal map display system as claimed in claim 1, it is characterized in that: described sequence image location matches module adopts the image matching algorithm based on architectural feature, and described image matching algorithm based on architectural feature is: establish start frame eigenmatrix p _uExpression, a certain eigenmatrix q in zone to be matched _uExpression, q _u, p _uHave identical space size, the edge feature statistical indicator is:

$\mathrm{\&theta;}(p,q)=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}(\sqrt{{({\mathrm{\&epsiv;}}_1*p_u)}^2+{({\mathrm{\&epsiv;}}_2*p_u)}^2}-\sqrt{{({\mathrm{\&epsiv;}}_1*q_u^i)}^2+{({\mathrm{\&epsiv;}}_2*q_u^i)}^2})---\left(4\right)$

Wherein

M is that zone to be matched contains q _uNumber, matrix p _uAnd q _uWith ε ₁, ε ₂Multiply each other and can be split as a series of 3 * 3 matrix.Two correspond respectively to trier marginal information in target area and the zone to be matched in the formula (1).When the pixel that is characterized as u all appears at q in the zone to be matched _uThe time, (p q) gets minimum value to θ, and this moment, feature u was the highest to this regional degree of support.

3. phase transformation thermal map display system as claimed in claim 1 is characterized in that: described initial pictures pretreatment module comprises that initial profile chooses module, is used for choosing the model initial profile from normalized sequence image.

4. phase transformation thermal map display system as claimed in claim 3 is characterized in that: described initial profile is chosen module and is adopted the edge finding algorithm that profile extracts and profile is followed the tracks of.

5. phase transformation thermal map display system as claimed in claim 1 is characterized in that: the display module of described thermal map spectrum comprises:

The model parameter load module, by the corresponding relation of model and model initial profile size, the adjustment of implementation model parameter input;

The thermal map spectrum is drawn and output module, draws out the thermal map of model by model phase change line and model parameter and composes the line output of going forward side by side.

Images