CN100399048C - Multiplexing accounting method and device for linear-variable frequency-adjusting scale imaging algorithm factor - Google Patents

Abstract

The present invention discloses a multiplexing computational device for converting an imaging algorithm factor of the linear frequency modulation dimension, which comprises an input interface module 1, a precomputed module 2, a linear correlation amount computation module 3, a point correlation amount computation module 4, an output interface module 5 and a control module 6. The present invention also discloses a multiplexing computational method for converting the imaging algorithm factor of the linear frequency modulation dimension. The present invention comprises the following steps that firstly, three kinds of basic variables which comprise frame correlation amount, line correlation amount and point correlation amount in a computational expression of a CS compensation factor, a distance compensation factor and an orientation compensation factor are separated; subsequently, the basic variables in the factor computational expression are analyzed numerically and simplified; the variables with the same computational direction in the factor computational expression are combined to generate new variables; the computational formula is simplified according to the computational formula for three factors so that an unified factor computational model A1*R1+A2*R2 is obtained; finally, the computational model is used for figuring out the factorial values.

Description

The reusable computing method and the device of conversion linear frequency modulation scale imaging algorithm factor

Technical field

The present invention relates to a kind of computing method of the reusable factor, the computing method of particularly CS compensation in the conversion linear frequency modulation scale imaging algorithm, compensated distance, three factors of orientation compensation.

Background technology

Synthetic-aperture radar (SAR, Synthetic Aperture Radar) can realize that round-the-clock, round-the-clock, large tracts of land observe and high-resolution imaging over the ground, in fields such as military, economy and environment significant application value and potentiality is arranged.Synthetic-aperture radar imaging algorithm commonly used is apart from range and Doppler (RD, Range Doppler) and conversion linear frequency modulation yardstick (CS, Chirp Scaling) algorithm.The CS algorithm is directly from echo, and the echoed signal of accurately deriving is in the expression formula in distance Doppler territory, avoided in the RD algorithm the synthetic aperture time than the bigger defective of distance translation correction interpolation calculated amount increase under the elongate member.

As shown in Figure 1, conversion linear frequency modulation scaling algorithm comprises: import raw data, export to inverse fourier transform, complex image to inverse fourier transform, azran data transposition, orientation to Fourier transform, distance apart from bearing data transposition, fourier transform of azimuth, distance.The whole flow process of this algorithm is as follows: earlier in the orientation to carrying out Fourier transform, raw data is transformed to the territory apart from Doppler, by carrying out complex multiplication with Chirp Scaling (CS) factor, being corrected on the reference curve of certain preliminary election of all range gate, thereby make the identical of all range gate apart from displacement curve apart from displacement curve.Then carry out distance to Fourier transform, transform the data into two-dimensional frequency, it is carried out compensated distance factor complex multiplication, finish apart from displacement correction and distance compression; By apart to inverse fourier transform, become again again, carry out complex multiplication, carry out the orientation compression after the orientation obtains the SAR image to inverse fourier transform with the orientation compensating factor apart from behind the Doppler territory.

From above-mentioned flow process as seen: between Fourier transform, need to add the CS compensating factor at fourier transform of azimuth and distance, distance needs to add the compensated distance factor to Fourier transform and distance between inverse fourier transform, azran data transposition and orientation need to add the orientation compensating factor between inverse fourier transform.

Calculating to above-mentioned three factors has had ready-made computing formula, and computing formula is as follows: CS compensating factor: Φ ₁(τ, f, R _Ref)=exp{-j π B _R0(f, R _Ref) C _s(f) [τ-τ _Ref(f)] ²(1)

The compensated distance factor: ${\mathrm{\Φ}}_2(f_c,f,R_{\mathrm{ref}})=\exp \left\{j\frac{\mathrm{\π}{f}_c^2}{B_{r0}(f,R_{\mathrm{ref}})[1+C_s\left(f\right)]}\right\}\·\exp \left\{j\frac{4\mathrm{\π}}{c}{f}_{\mathrm{\τ}}{R}_{\mathrm{ref}}{C}_s\left(f\right)\right\}---\left(2\right)$

The orientation compensating factor: ${\mathrm{\Φ}}_3(\mathrm{\τ},f)=\exp \{-j\frac{4\mathrm{\πR}}{\mathrm{\λ}}[1-\sin \mathrm{\φ}\·D\left(f\right)]+(-j\frac{2\mathrm{\π}\mathrm{Rf}\cos \mathrm{\φ}}{v})+\mathrm{j\Θ}\left(f\right)\}---\left(3\right)$

Wherein,

$C_s\left(f\right)=\frac{\sin \mathrm{\φ}}{\sqrt{1-{\left(\frac{\mathrm{\λf}}{2v}\right)}^2}}-1$

$B_{r0}(f;R_{\mathrm{ref}})=\frac{b}{1+b{R}_{\mathrm{ref}}\sin \mathrm{\φ}\frac{2\mathrm{\λ}}{c^2}\frac{{\left(\frac{\mathrm{\λf}}{2v}\right)}^2}{{[1-{\left(\frac{\mathrm{\λf}}{2v}\right)}^2]}^{\frac{3}{2}}}}$

${\mathrm{\τ}}_{\mathrm{ref}}\left(f\right)=\frac{2}{c}{R}_{\mathrm{ref}}[1+C_s\left(f\right)]$

R _RefBe reference distance, adopt image center usually;

φ is the platform motion angular separation of radar wave speed center and radar;

B is a transponder pulse frequency modulation rate;

C is the light velocity;

λ is the radar emission wavelength;

V is the radar ground speed;

The computing formula of above-mentioned variable is:

$v=\sqrt{\frac{\mathrm{\λ}{R}_{\mathrm{ref}}{f}_{\mathrm{dr}}}{2}+{\left(\frac{\mathrm{\λ}{f}_{\mathrm{dc}}}{2}\right)}^2}$

$\mathrm{\φ}=\arccos [-\frac{{\mathrm{\λf}}_{\mathrm{dc}}}{2v}]$

$D\left(f\right)={[1-{\left(\frac{\mathrm{\λf}}{2v}\right)}^2]}^{\frac{1}{2}}$

$\mathrm{\Θ}\left(f\right)=\frac{4\mathrm{\π}}{c^2}{B}_{r0}(f,R_{\mathrm{ref}})[1+C_s\left(f\right)]C_s\left(f\right){[R\frac{\sin \mathrm{\φ}}{\sin {\mathrm{\φ}}_{\mathrm{ref}}}-R_{\mathrm{ref}}]}^2$

In above-mentioned basic variable, f be the orientation to variable, and τ, f _τ, R is that distance is to variable.

In three factor expressions, reference distance R _Ref, transponder pulse frequency modulation rate b, light velocity c, radar emission wavelength λ are basic variables known or that can directly be obtained by instrument; And radar ground speed v, radar beam center and Texas tower direction of motion included angle need be asked for by calculating.In the computing formula of v, f _DrAnd f _DcRepresent doppler frequency rate and doppler centroid respectively, the f in other formula, τ, f _τRepresent the orientation to Doppler frequency respectively, distance is to constantly, and distance is to frequency on the two-dimensional frequency.Can obtain C by above-mentioned basic variable _s(f), B _R0Variable such as (f).But to further understand the information list of references 1 that calculates about three kinds of factors: Huangyan, " high-resolution spaceborne synthetic aperture radar image-forming Processing Technology Research ", BJ University of Aeronautics ﹠ Astronautics's PhD dissertation, in June, 1999.

In existing synthetic aperture radar image-forming CS algorithm, utilize the aforementioned calculation formula to Chirp Scaling (CS) factor, the compensated distance factor, when compensating factor these three kinds of factors in orientation are calculated, because computing formula itself is comparatively complicated, therefore the data volume of calculating is huge, very high and be not easy to guarantee the real-time calculated to the requirements such as storage resources, Internet resources and computational resource of system, if adopt specialized hardware to realize, then its numerous high computing just allows the people forbidding.Therefore in existing factor computing method, adopt software programming usually, this method calculation of complex, speed is lower, is difficult to satisfy the requirement of calculating in real time.If can be based on the principle of numerical analysis, variable in three factor formula is carried out combinatory analysis, the computing method that adopt numerical value to approach in certain accuracy rating have carried out simplifying processing to variable and combination thereof, just can improve the efficient of calculating, make hardware realize becoming possibility, desired real-time when guaranteeing factor calculating.In addition, need the calculating that different software and hardwares is realized correlation factor, the height that not only assesses the cost, and complicated operation, the current calculating that does not also have a kind of method or equipment can be applicable to three kinds of factors simultaneously at the different factors in the prior art.Therefore study a kind of calculating that can be applicable to three kinds of factors, and special-purpose faster factor reusable computing method and the device of arithmetic speed is very important.

Summary of the invention

One object of the present invention is that in conversion linear frequency modulation scale imaging algorithm unified CS compensation, compensated distance, three factor calculation process of orientation compensation provide a kind of reusable factor computing method and device, reduce the design complexities of factor computing system.

Another object of the present invention is in conversion linear frequency modulation scale imaging algorithm, simplifies factor Calculation Method, realizes the real-time calculating to the factor.

To achieve these goals, the invention provides a kind of reusable calculation element of conversion linear frequency modulation scale imaging algorithm factor, comprising: input interface module 1, precalculation module 2, line correlative computing module 3, spot correlation amount computing module 4, output interface module 5 and control module 6; Wherein input interface module 1 is used for enter factor and calculates needed initial variate-value; Precalculation module 2 is used for calculating in advance the parameter that will use when calculated factor; Line correlative computing module 3 is used to realize the calculating to weak real-time line correlative; Spot correlation amount computing module 4 is used to realize the calculating to the spot correlation amount of hard real time; Output interface module 5 is used to export factor result calculated; Control module 6 is used for the workflow of control module and the synchronous and communication issue between the module.

In the technique scheme, the function of described precalculation module 2 is by software programming.

In the technique scheme, described spot correlation amount computing module 4 adopts level Four flowing water floating-point adder.

A kind of reusable calculation element of using conversion linear frequency modulation scale imaging algorithm factor carries out the reusable Calculation Method, may further comprise the steps:

1) computing formula of three factors is carried out abbreviation, obtain the computation model that a unified factor is calculated: A ₁* R ₁+ A ₂* R ₂

2), utilize the preceding original computing formula of this factor abbreviation to computation model A to the some factors in three factors ₁* R ₁+ A ₂* R ₂In component A ₁Ask a plurality of sampled points, utilize sampled point to realize that a certain polynomial function approaches the numerical value of component, obtain the form and the correlation parameter of this polynomial function;

3) replace original computing formula of component with polynomial function, by the polynomial function evaluation being obtained the value of component;

4) use and step 2) and 3) similarly method obtain other components R in the computation model ₁, A ₂, R ₂Value;

5) value of resulting each component is done linear operation and data and go whole the processing, obtain the value of computation model, also just obtain the value of this factor;

6) use the same method and ask the value of other factors.

In the described step 1), the A in the computation model ₁, A ₂Be weak real-time calculated amount, R ₁, R ₂Be the hard real time calculated amount; For CS compensating factor, A ₁Be (τ-τ _Ref) ², R ₁Be B _R0C _S, A ₂And R ₂Be 0; For the compensated distance factor, A ₁Be 1/[B _R0(1+C _S)], R ₁Be f _τ ², and A ₂Be R _RefC _S(f)/and c, R ₂Be f _τTo orientation compensating factor, A ₁Be 4B _R0C _s(1+C _s)/c ², R ₁For

A ₂Be 4 R ₂Be R.

In step 2) in, described when doing numerical value and approach with polynomial function, preferably adopt the polynomial function of segmentation to do numerical value and approach.

In step 5), described data go to put in order processing can be to A ₁* R ₁And A ₂* R ₂Carry out respectively or to A ₁* R ₁+ A ₂* R ₂The result do data and go whole the processing.

The invention has the advantages that:

1, the present invention has realized the reusable of conversion linear frequency modulation scale imaging algorithm factor is calculated, and makes and the calculating that can be used for the different factors with a kind of device has improved the integrated level that the factor is calculated.

2, the present invention has simplified the design of factor computing system in the conversion linear frequency modulation scale imaging algorithm, helps the realization of hardware design and software-hardware synergism.

3, in factor calculation process, adopted polynomial expression to approach or the polynomial expression approach method of segmentation to aleatory variable, reduced the complexity that the factor is calculated, improved the efficient that the factor is calculated.

Description of drawings

Fig. 1 is existing conversion linear frequency modulation scaling algorithm process flow diagram;

Fig. 2 is the process flow diagram of the reusable computing method of conversion linear frequency modulation scale imaging algorithm factor of the present invention;

Fig. 3 is the collaborative realization flow figure of conversion linear frequency modulation scaling algorithm soft or hard of the present invention;

Fig. 4 is a C-band SAR image transformation linear frequency modulation scaling algorithm floating-point adder data path of the present invention;

Fig. 5 is that shown position is a Beijing's Imperial Palace based on C-band SAR image transformation linear frequency modulation scaling algorithm part imaging results of the present invention;

Fig. 6 is that shown position is the YiHeYuan,BeiJing based on C-band SAR image transformation linear frequency modulation scaling algorithm part imaging results of the present invention.

Embodiment

Below in conjunction with the drawings and specific embodiments the present invention is described in further detail.

With the C-band is example, and as shown in Figure 2, the operation steps of reusable factor computing method comprises:

In step 10, three class basic variables in separation of C S compensating factor, the compensated distance factor, the orientation compensating factor calculating formula: frame correlative, line correlative, spot correlation amount.Wherein, the frame correlative is meant the variable that every frame need upgrade, it is a non real-time calculated amount, when non real-time is meant every frame image data is handled, only need calculate the calculated amount of fixed number of times, its value remains constant in the data processing of every two field picture, computation complexity is a constant, and is irrelevant with the image size; The variable that need upgrade when the line correlative is meant every line in the computed image, it is weak real-time calculated amount, a little less than be meant that in real time arbitrary line to pending image only need carry out the calculating of fixed number of times, its computation complexity is directly proportional with the line number of processing image; The spot correlation amount is meant all needs the variable that upgrades to each pixel in the image, it is the calculated amount of hard real time, hard real time is meant any pixel for pending view data, and this value all needs to calculate separately, and its computation complexity is directly proportional with counting of processing image.

The normal distance that adopts is to always representing calculated direction with the orientation in imaging processing, and in different factor computing formula, the calculated direction of the factor is different, in the CS compensating factor, calculated direction be the orientation to, in the compensated distance factor and orientation compensating factor, calculated direction be distance to.Stipulate that again in the factor was calculated, the variable vertical with calculated direction was as weak real-time variable, variable that will be identical with factor calculated direction is as the hard real time variable.Therefore, the orientation is the spot correlation variable of hard real time to variable in the CS compensating factor, and distance is weak real-time line correlated variables to variable; And in the compensated distance factor and orientation compensating factor, distance is the spot correlation variable of hard real time to variable, and the orientation is weak real-time line correlated variables to variable.

With the CS compensating factor is example, calculated direction be the orientation to, orientation to frequency f be the orientation to variable, distance to moment τ be distance to variable, so f and make up and calculate as the hard real time amount, τ and combination thereof are then calculated as weak real-time amount.And τ _Ref(f) be the orientation upward reference time,, can be used as the non real-time variable quantity, be approximately in same frame image data and remain constant because its value is very little in the variation range that whole orientation makes progress.

In step 20, the basic variable in the calculation expression of three factors is carried out numerical analysis, and according to circumstances Partial Variable is wherein simplified.In a frame image data, by to doppler centroid f _Dc, doppler frequency rate f _DrRadar ground speed v, the numerical analysis of variablees such as radar beam center and Texas tower direction of motion included angle finds that the slope of these variablees is very low, be reduced to behind the constant not quite to the influence of factor computational accuracy within the specific limits, therefore above-mentioned variable can be considered constant.

In step 30, the variable that some calculated direction are identical in the factor calculation expression is merged, generate new variable, and adopt the numerical value approach method to simplify processing the calculating of new variables.As in the CS compensating factor, the orientation is to variable B _R0With the orientation to variable C _SAll be the variable about f, the calculated direction of two variablees is identical, so these two variablees can be merged into a new orientation to variable.If to synthetic new variables B _R0C _SWhen calculating with existing computing formula, because existing computing formula is too complicated, computing velocity and computational resource are difficult to satisfy actual needs, therefore will be to synthetic new variables B _R0C _SComputing formula simplified.From the numerical analysis of previous step as can be known, in C-band SAR image transformation linear frequency modulation scaling algorithm, the f dynamic range is smaller, therefore to B _R0C _SCan adopt and carry out polynomial expression (adopt binomial usually, numerical analysis shows for the C-band data selects for use linear function also can obtain the goodish effect of approaching) function numerical value based on least square method and approach, used polynomial function was exactly B when numerical value approached _R0C _SNew computing formula.Compare with original computing formula, new computing formula is simplified greatly, and precision of calculation results is suitable.

In step 40, according to the computing formula of three factors, computing formula is carried out abbreviation, obtain the computation model A that a unified factor is calculated ₁* R ₁+ A ₂* R ₂A wherein ₁, A ₂Be weak real-time calculated amount, R ₁, R ₂Be the hard real time calculated amount.For three factors, A ₁, A ₂, R ₁, R ₂The implication of its expression has nothing in common with each other, after the calculating formula abbreviation with three factors, and can be in the hope of A ₁, A ₂, R ₁, R ₂Expression formula separately: for CS compensating factor, A ₁Be (τ-τ _Ref) ², R ₁Be B _R0C _S, τ wherein _RefIn certain accuracy rating, can be considered constant, and A ₂And R ₂Be 0; For the compensated distance factor, A ₁Be 1/[B _R0(1+C _S)], R ₁Be f _τ ², and A ₂Be R _RefC _S(f)/and c, R ₂Be f _τTo orientation compensating factor, A ₁Be 4B _R0C _s(1+C _s)/c ², R ₁For And A ₂Be 4 R ₂Be R.

In step 50, for reducing the complexity of calculating, to computation model A ₁* R ₁+ A ₂* R ₂In single component, as A ₁, A ₂, R ₁, R ₂, adopt based on the polynomial function of least square method and realize respectively the approaching of each component done linear operation to each component at last, obtain computation model A ₁* R ₁+ A ₂* R ₂Value, thereby obtain the value of three factors.

With the R in the CS compensating factor ₁=B _R0C _SBe example, its implementation method is:

A) to R ₁Adopt parabolic function to carry out approximate treatment, establish its approximant R of being ₁(f)=a ₁f ²+ b ₁F+c ₁, this is approximant to be to be the parabolic function of variable with Doppler frequency f.

B) according to CS compensating factor B _R0C _SComputing formula

$B_{r0}{C}_s=(\frac{\sin \mathrm{\φ}}{\sqrt{1-{\left(\frac{\mathrm{\λf}}{2v}\right)}^2}}-1)\·\left(\frac{b}{1+b{R}_{\mathrm{ref}}\sin \mathrm{\φ}\frac{2\mathrm{\λ}}{c^2}\frac{{\left(\frac{\mathrm{\λf}}{2v}\right)}^2}{{[1-{\left(\frac{\mathrm{\λf}}{2v}\right)}^2]}^{\frac{3}{2}}}}\right)$

Obtain R ₁3 above sampled points, with the value substitution of sampled point about R ₁Approximant in, thereby obtain a ₁, b ₁, c ₁Numerical value.With a ₁, b ₁, c ₁Value substitution R ₁Approximant in, utilize this approximantly can ask for R in the real-time calculating easily ₁Value.

C) utilize similar method to ask A ₁, A ₂, R ₂, at last according to computation model A ₁* R ₁+ A ₂* R ₂Obtain the value of orientation compensating factor.

Can use the same method to orientation compensating factor and the compensated distance factor and to ask their value.

In the methods of the invention, in step 20 and step 50, use polynomial function to realize the numerical value of each component in aleatory variable in the calculating formula and the computation model is approached.When these polynomial functions are calculated, can adopt totalizer to realize.But using totalizer to carry out in the calculation process, the single step accumulation of error of polynomial function can amplified.In order to eliminate the accumulated error that totalizer may be brought, adopt piecewise function to carry out numerical value and approach.For convenience of description, be example with one-dimensional linear function y=ex+t, when this function is carried out hardware calculating, need a multiplier and a totalizer; If this function adopts recursion formula y when calculating _I+1-y _i=e (x _I+1-x _i), during owing to the SAR Flame Image Process, the difference dx=(x of adjacent 2 independents variable _I+1-x _i) be constant, irrelevant with the position i of point, therefore when calculating y _I+1=y _iOnly need a totalizer to get final product during+dy (wherein dy=e*dx), compare, on hardware is realized, become simple, reduced the complexity of system with the method that needs multiplier and totalizer simultaneously.But totalizer has error inevitably in the process that at every turn linear function is added up, and the input value of the result that totalizer will last time add up when adding up as next time, thereby additive errors can be amplified by accumulation with the increase of accumulative frequency.In order to eliminate this cumulative errors, the linear function that adopts segmentation is to linear function y _i=e _ix _i+ t _iThe variable that (wherein i represents different segmentations) approached carries out numerical evaluation, promptly adopts many broken lines that variable is approached, in the linear function of each segmentation, and the slope e of linear function _iDifference is arranged slightly.The accumulation length of totalizer be restricted on the calculated direction count and segments between the merchant, adopt many broken lines that variable is approached also than adopting single straight line that variable is carried out the error that numerical value approaches simultaneously and further reduce.Y=ex+t is similar with the one-dimensional linear function, when using polynomial function to do numerical value to approach, also can adopt the polynomial function of segmentation to realize the numerical value of variable or component is approached, to improve computational accuracy.In addition, when polynomial function is done segmentation, also can reduce the dimension of piecewise function, adopt lower function do numerical value approach can reach with the higher-dimension function of not segmentation to approach effect suitable.

Enumerated the one-dimensional linear function in the present embodiment, be to be understood that when adopting the higher-dimension function to realize, also can improve precision the approaching of variable for those of ordinary skill in the art.

Utilize method of the present invention can adopt the calculating of a unified computation model realization to three factors.Owing to be to adopt based on the polynomial function of least square method to realize approaching in factor computation process to each component, and for each factor, the multinomial coefficient that obtains by least square method is only along with the frame correlation parameter changes, and with frame in location independent, these multinomial coefficients are as a ₁, b ₁, c ₁, also be regarded as the frame correlative.For the calculating of frame correlative, because its calculating is very flexible, calculative data volume seldom can adopt the precalculated method of software to ask for.In factor computation process, also to calculate real-time amount, owing to, therefore adopt the method for hardware or software and hardware combining to realize to these calculating of amount in real time to the having relatively high expectations of the real-time of real-time amount.In sum, the reusable calculation element of the factor adopts the soft or hard layered cooperative to realize.

As shown in Figure 3, the reusable calculation element of conversion linear frequency modulation scale imaging algorithm factor comprises: input interface module 1, precalculation module 2, line correlative computing module 3, spot correlation amount computing module 4, output interface module 5 and control module 6.

The function of input interface module 1 is that enter factor calculates needed initial variate-value, comprising: reference distance R _Ref, transponder pulse frequency modulation rate b, light velocity c, radar emission wavelength λ, doppler centroid f _Dc, doppler frequency rate f _DrThe value of needed its dependent variable can obtain by the calculating of above-mentioned variable when the factor was calculated.

The function of precalculation module 2 is to calculate the parameter that will use in advance by the method for programming in subsequent step, comprising: calculate each component A be used for realizing to computation model ₁, A ₂, R ₁, R ₂Carry out numerical value used polynomial function when approaching, as the R in the CS compensating factor ₁(f)=a ₁f ²+ b ₁F+c ₁The related coefficient of polynomial function is as a ₁, b ₁, c ₁Needed frame correlation parameter when carrying out line correlation parameter or spot correlation calculation of parameter is as R in the CS compensating factor ₁The initial value etc. of f.Known from the acquiring method of the factor, the related coefficient of polynomial function is to utilize existing complicated formulas that variable is asked the sampled point of some, obtains in the value substitution polynomial function with sampled point.The computation complexity height of aforesaid operations, the consumed time of wanting longer, and since the formula complexity be difficult to realize with hardware.But these parameters mostly are non real-time frame correlative, and are lower to the requirement of real-time, therefore ask for these parameters and can realize in software by the method for programming.

The function of line correlative computing module 3 is the calculating that realizes weak real-time line correlative, and the line correlative comprises: factor unified calculation model A ₁* R ₁+ A ₂* R ₂In A ₁And A ₂, for the different factors, A ₁And A ₂Represented implication difference, but can realize calculating with same module to the different factors.Line correlative computing module 3 utilizes by the frame correlative that obtains in the precalculation module 2, calculates in inside modules, obtains desired line correlative.Because the line correlative is weak real-time amount, it is not very high to the requirement of real-time, so line correlative computing module 3 can adopt the method for software and hardware combining, use less hardware resource, and the method realization by software and State Control is to the calculating of line correlative.Hardware resource is minimum only to need a multiplier, a totalizer, and some registers and state machine.According to the concrete condition of hardware resource be used for the polynomial function that numerical value approaches, edit different software, be used for the calculating of control hardware realization to the line correlative.Compare with the method for simple use hardware, utilize the method for software and hardware combining can reach the purpose that exchanges space and hardware resource with the time for.

The function of spot correlation amount computing module 4 is the calculating that realizes the spot correlation amount of hard real time, and the spot correlation amount comprises factor unified calculation model A ₁* R ₁+ A ₂* R ₂In R ₁And R ₂And to the multiply-add operation of four components.

For the different factors, R ₁And R ₂Represented implication difference, but can realize calculating with same module to the different factors.Spot correlation amount computing module 4 input precalculation module 2 resulting frame correlatives and line correlative computing module 3 resulting line correlatives are utilized the calculating of hardware realization to the spot correlation amount.Because the spot correlation amount is a hard real time, therefore, in the present invention, mainly adopt the calculating of the method realization of hardware to the spot correlation amount.To spot correlation amount R ₁And R ₂Calculating the time adopted the method for Floating-point Computation, and calculating in real time requires result of every bat output, therefore in an embodiment, floating-point adder can adopt level Four flowing water floating-point adder as shown in Figure 4.Level Four in the level Four flowing water floating-point adder is meant needs four bats just can finish an additive operation in this floating-point adder, although flowing water is meant that finishing an additive operation needs four beats, but every bat can both be exported a result, streamline just as in the industry generation utilizes the expansion in space to come the overlapping of deadline.For example, when first additive operation was in the second level of level Four flowing water, second additive operation was in the first order of level Four flowing water.Four beats in the level Four flowing water floating-point adder finish respectively complement code to rank, complement code add/subtract, complement code normalizing and the processing of rounding off.Level Four flowing water floating-point adder has two kinds of inputs, and a kind of input is the result of calculation of totalizer output, and another kind is input as the streamline characteristic parameter that adds up.The streamline characteristic parameter that adds up is the relevant parameter of frame, and it comprises the initial value of streamline and the yield value of streamline, and the yield value of streamline is relevant with the slope of polynomial function.This totalizer is the level Four streamline, therefore needs four initial values, supposes to use respectively s0, s1, and s2, s3 represents.From the input of totalizer, input sequence is respectively s0, s1, s2, s3, s0+4, s1+4, s2+4 ..., the output of totalizer is respectively s0+4, s1+4, s2+4, s3+4, s0+4+4 in proper order ...Obvious s0, s1, s2, s3 can't obtain from the output of streamline, can only obtain by precomputation, that is to say that the initialization of streamline need be by the result of precomputation, and in a single day streamline flows, and its output just can be gone down as the input operation endlessly.Mention in the aforesaid method, error when utilizing piecewise function to realize that numerical value to variable approaches more single function and does numerical value and approach is littler, and that the yield value of streamline is the slope of and function is relevant, therefore when realizing that with piecewise function numerical value to variable approaches, the yield value of streamline also will be made corresponding change.In the flowing water calculation process, under the control of control state machine, can reset the yield value of streamline, to improve factor precision of calculation results.

Try to achieve factor unified calculation model A ₁* R ₁+ A ₂* R ₂In R ₁And R ₂After, utilize multiplier to obtain A respectively ₁* R ₁And A ₂* R ₂Value, then resulting data are done whole the processing, utilize totalizer will go whole later data addition to obtain final value at last, thereby obtain the value of correlation factor.Wherein, the whole processing of going of data is meant: because factor result of calculation is angle value, if angle surpasses 360 °, this value is nonsensical, and going whole the processing is exactly to remove to surpass 360 ° part in the angle, and angle value is controlled in 360 °.In the present invention, calculate and adopted floating-point format, the index that needs only according to floating data when therefore going whole the processing carries out shifting processing to data, abandons its integral part then and gets final product.In the prior art, going whole the processing is just to carry out after all calculating is all finished, and will go whole the processing to finish before last add operation in the methods of the invention, can will be used for A ₁* R ₁And A ₂* R ₂The original floating-point adder of doing the addition processing is reduced to the fixed point totalizer.

The function of output interface module 5 is the factor result of calculation output that will obtain in the abovementioned steps.Include corresponding output interface circuit in the output interface module 5.

Control module 6 be used between the workflow of control module and the module synchronously and communication issue, under the effect of the control signal of control module 6, system can realize the calculating to the different factors.Simultaneously, control module is responsible for the total system state is controlled and monitored, and is responsible for the synchronous and communication issue between native system and the other system.

Factor computing system has versatility to CS compensating factor, the compensated distance factor and orientation compensating factor, by can realize the calculating to the different factors to the different initial value of factor computing system input.

The part imaging results that utilization realizes based on C-band SAR image transformation linear frequency modulation scaling algorithm of the present invention can be referring to Fig. 5 and Fig. 6, and the imaging region among Fig. 5 is a Beijing's Imperial Palace, and the imaging region among Fig. 6 is the YiHeYuan,BeiJing.

SAR image transformation linear frequency modulation scaling algorithm for its all band just has subtle difference on the inner structure of concrete hardware component, its Calculation Method, and aspects such as the structure of system do not have the difference of internal.So method of the present invention has adaptive widely.

Claims (3)

1. the reusable calculation element of a conversion linear frequency modulation scale imaging algorithm factor comprises: input interface module (1), precalculation module (2), line correlative computing module (3), spot correlation amount computing module (4), output interface module (5) and control module (6); Wherein input interface module (1) is used for enter factor and calculates needed initial variate-value; Precalculation module (2) is used for calculating in advance the parameter that will use when calculated factor; Line correlative computing module (3) is used to realize the calculating to weak real-time line correlative; Spot correlation amount computing module (4) is used to realize the calculating to the spot correlation amount of hard real time; Output interface module (5) is used to export factor result calculated; Control module (6) is used for the workflow of control module and the synchronous and communication issue between the module.

2. the reusable calculation element of conversion linear frequency modulation scale imaging algorithm factor according to claim 1 is characterized in that, the function of described precalculation module (2) is by software programming.

3. the reusable calculation element of conversion linear frequency modulation scale imaging algorithm factor according to claim 1 is characterized in that, described spot correlation amount computing module (4) adopts level Four flowing water floating-point adder.

Images