CN104635221B - A kind of subband joining method based on internal calibration data - Google Patents

Abstract

The invention discloses a kind of subband joining method based on internal calibration data, including step one：Extract error and interchannel time delay in band；Step 2：Process range error and the time delay that each subband internal calibration signal is extracted；Step 3：Make the internal calibration signal compression position of different sub-band be moved to zero moment, process the phase error of each subband；Step 4：Read the radar return data of subband, carry out error compensation in list band band；Step 5：Error in the band of each subband signal is compensated, calculates the time delay of each subband：Step 6：The mode that the frequency-region signal of each subband in step 5 is carried out with afterbody zero padding carries out a liter sampling；Step 7：Carry out splicing in frequency domain；Step 8：Obtain splicing the time-domain signal completing.The inventive method not only can compensate the amplitude phase error in subband, and can compensate the time delay between the amplitude phase error of intersubband and different sub-band, realizes the relevant fusion treatment of signal between different sub-band.

Description

A kind of subband joining method based on internal calibration data

Technical field

The invention belongs to field of signal processing, it is related to subband joining method, particularly to a kind of based on internal calibration data Synthetic aperture radar subband joining method.

Background technology

Synthetic aperture radar (SAR) satellite quickly grew in the last few years, because SAR satellite is not subject to weather, geography, time etc. The restriction of factor, can carry out the observation of round-the-clock over the ground, and have certain penetration power, thus be widely used in military affairs The aspects such as scouting, mapping, resource detection, oceanographic observation, ecological monitoring, Natural calamity monitoring, Quick rescue.

High-resolution is the eternal pursuit of synthetic aperture radar.SAR imaging resolution is higher to mean target scattering characteristics Understanding is more, more abundant to the acquisition of information of imaging region, target recognition accurate performance more can be guaranteed.Additionally, it is high Resolution also will be helpful to suppress the speckle noise phenomenon of coherence imaging system, and image visual effect will be substantially improved.SAR is imaged Process is an orientation and distance process to two-dimension focusing, its resolution be also classified into orientation and resolution and distance to point Resolution.The lifting of azimuth resolution depends on the increase of synthetic aperture time, and the lifting of range resolution depends on and sends out Penetrate the raising of receipt signal bandwidth.However, when transmitted signal bandwidth is very big, signal digitized will be carried out and then requires higher adopting Flat characteristic in sample rate and broader receiver band, this can cause higher requirement to hardware.For solving problem above, can adopt With band shared method, echo-signal is carried out with the many subbands of frequency domain receive.This method is to increase system hardware complexity as generation Valency, reduces the requirement to radar receiver sample rate and reception bandwidth for the system.Need when being received using subband in radar two dimension Before focal imaging, each subband signal is carried out splicing, obtain and transmitted signal bandwidth identical broadband signal.In theory, when In the band of subband receiver no amplitude-phase error when, the inverse process that receives for subband of process of subband splicing.But in reality In engineering, in each subband receiver band, have different amplitude phase error, and the frequency modulation characteristic according to linear FM signal, Different receivers receive different center frequency subband signal between also have the regular hour postpone.And subband splicing is one Individual extremely fine processing procedure, small error also can cause very big impact to splicing effect, can affect follow-up simultaneously Two-dimension focusing process.Therefore, during splicing, need in subband and the error of intersubband compensates.

During radar work, the radar separation regular hour launches and receives internal calibration signal.Its process such as Fig. 1 Shown, radar emission broadband signal, this signal without antenna array transmitting and receives, only through radar home loop.Internal calibration Signal can reflect the characteristic of radar inner link effectively, can extract in subband by internal calibration data and intersubband width Phase error, can effectively compensate for the error in subband.And not only need to consider the error in subband in subband splicing, It is also contemplated that the amplitude phase error of intersubband, realize the relevant fusion of subband signal.

Content of the invention

The invention aims to solving the above problems, a kind of subband joining method based on internal calibration data is proposed, The method, based on the magnitude-phase characteristics of linear FM signal, can effectively utilizes internal calibration data be extracted in subband and intersubband Error, and process is compensated to radar initial data, while not affecting echo data orientation characteristic, completes subband Splicing, to bandwidth, range of lift is to resolution for the distance increasing echo-signal.

A kind of subband joining method based on internal calibration data, including following step：

Step one：Read in frame in calibration data of list band, extracted in band by mistake by this frame in calibration data of list band Poor and interchannel time delay；

Step 2：By the method for average superposition denoising process the range error that each subband internal calibration signal extracts and when Between postpone, improve the method estimated accuracy；

Step 3：Remove each subband first order corresponding with time delay, make the internal calibration signal compressed-bit of different sub-band Set moves to zero moment, processes the phase error of each subband by the method for phase gradient and average superposition denoising；

Step 4：Read the radar return data of k-th subband, carry out error compensation in list band band；

Step 5：Repeat step four, error in the band of each subband signal is compensated；Now each, in subband Phase error has removed, and with the first subband as time reference, calculates the time delay of each subband：

Step 6：The mode that the frequency-region signal of each subband in step 5 is carried out with afterbody zero padding carries out a liter sampling, zero padding Count as (N-1) × ns, wherein N is subband total number；Inverse Fourier transform is done to signal, transforms to time domain, then now signal Sample rate be Fs, Fs=N × fs；It is multiplied by a phase compensation function in time domain and carry out frequency displacement so that subband signal is located at it Corresponding frequency position；

Step 7：After each subband signal is carried out frequency displacement process, if k-th subband signal is Sub_k；Enter in frequency domain Row splicing, joining method is added for frequency domain, and the part overlapping for intersubband takes its average；

Step 8：Refill distance to frequency modulation frequency modulation item, obtain splicing the time-domain signal completing.

It is an advantage of the current invention that：

(1) the inventive method not only can compensate the amplitude phase error in subband, and can compensate the width phase of intersubband by mistake Time delay between difference and different sub-band, realizes the relevant fusion treatment of signal between different sub-band；

(2) the splicing process of the inventive method enters row distance to process only for signal, and error extraction process need not Carry out two-dimension focusing process, the orientation being not related to subband is processed, and therefore has efficient feature；

(3) the inventive method does not destroy the frequency modulation characteristic of original transmission signal, does not affect follow-up two-dimension focusing Processing Algorithm Selection；

(4) in the extraction process of the amplitude phase error of the inventive method, average value processing is taken using multiframe internal calibration, can be effective Ground weakens effect of noise, therefore has good robustness.

Brief description

Fig. 1 is default mark system operating diagram

Fig. 2 is method of the present invention flow chart；

Fig. 3 is the method flow diagram extracting error and intersubband time delay in subband of the present invention；

Fig. 4 is the splicing of subband frequency domain and the signal relationship between frequency and time schematic diagram of the present invention；

Fig. 5 is that the list band internal calibration data of the present invention is directly compressed and adopted the inventive method result comparison diagram；

Fig. 6 is two subband internal calibration data direct splicing of the present invention and is contrasted using the inventive method result Figure；

Fig. 7 is three subband internal calibration data direct splicing of the present invention and contrasts using by inventive method result Figure；

Fig. 8 is the list band radar data imaging results figure of the present invention；

Fig. 9 is the three subband radar data direct compensation splicing result figures of the present invention.

Figure 10 is that the present invention three subband radar data adopts the inventive method splicing result figure

Specific embodiment

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

The present invention is a kind of subband joining method based on internal calibration data, as shown in Fig. 2 including following eight steps： First pass through step one and arrive step 3, error is extracted by internal calibration data, as shown in Figure 3；Passing through step 4 again will to step 8 The error that first three step is extracted compensates：

Step one：Read in frame in calibration data of list band, extracted in band by mistake by this frame in calibration data of list band Poor and interchannel time delay.

Concrete steps are divided into again：

A, k-th subband internal calibration data of reading, for ease of carrying out fast Fourier transform, mend to time-domain signal Zero, in original time domain signal afterbody zero padding so as to the power side always counted as 2, now the mid frequency of kth subband signal is f_c K (), sample rate is f_s, signal counts as ns.List band signal is carried out with fast Fourier transform (FFT) to frequency domain, it is entered Row matched filtering, removes frequency modulation quadratic term.Matched filtering function is:

Wherein：f_τBe distance to discrete sampling frequency,K_rFor frequency modulation rate.

The inverse of the range value of signal in each frequency for the signal after B, extraction matched filtering, is designated as A_inv：

A_inv_k=1/A_k(f_τ)

Wherein：A_k(f_τ) for discrete sampling frequency be f_τThe range value at place.

C, the frequency-region signal afterbody zero padding after matched filtering is carried out time domain and rise sampling, zero padding is counted as Ns-ns, after zero padding Signal is counted as Ns, Ns=r × ns, and wherein r is to rise a sample rate, carries out fast Fourier transform for convenience, r is also taken as 2 power Power, and r >=32.Then rising the sample rate after sampling is f_up, f_up=r × f_s.Convert the signal into time domain, obtain now pulse pressure The position n of signal maximum after contracting_max, obtain the pulse compression maximum moment t of k-th subband_k=n_max/f_up.

Step 2：By the method for average superposition denoising process the range error that each subband internal calibration signal extracts and when Between postpone, improve the method estimated accuracy.

Concrete steps are divided into again：

(1), the i-th frame in calibration signal of k-th subband signal is processed according to A, B two step of step one, obtain The reciprocal function A_inv of range value_{k_i}.

(2), process the i-th frame in calibration signal of k-th subband signal according to C in setting steps one, obtain pulse compression The maximum moment is t_{k_i}.

(3), according to A, B, C in step 2, the calibration signal of k-th subband signal is processed, total coprocessing m is set Frame, obtains the range value reciprocal function A_inv of each frame signal_{k_i}With time delay t_k(i), wherein i=1,2...m.To m In signal amplitude inverse, each frequency element and time delay are averaged the average amplitude reciprocal function obtaining k-th subbandWith the pulse compression maximum momentThe impact of white Gaussian noise in signal can be reduced, for lifting estimated accuracy m ≥100.

Wherein:

(4), each subband signal is processed according to step 2 (1), (2), (3), obtain the average of all subbands Amplitude reciprocal functionWith the pulse compression maximum moment

Step 3：Remove each subband first order corresponding with time delay, make the internal calibration signal compressed-bit of different sub-band Set moves to zero moment.Process the phase error of each subband by the method for phase gradient and average superposition denoising.

Concretely comprise the following steps：

<1>, the i-th frame in calibration signal of k-th subband signal is carried out at process of pulse-compression according to A in step one Reason, remove signal distance to quadratic phase.

<2>, by k-th subband signal after pulse compression i-th frame in calibration frequency-region signal be multiplied by linear phase ginseng Examine signal, make the internal calibration signal pulse compression peak of different sub-band be moved to zero moment.This goes linear phase with reference to letter Number it is：

<3>, setting signal now be S_{k_i}(f_τ), calculate phase gradient and remove phase error, the phase place of setting frequency Gradient is

Wherein：Arg represents and takes phase operation.

<4>, phase error that every frame signal of k-th subband signal extracted ask for phase again after being overlapped summation Position, reduces effect of noise in signal

<5>, phase gradient is integrated sue for peace, obtain the phase place of each frequency phase error composition of k-th subband signal Error, can be expressed as Φ_k：

Step 4：Read the radar return data of k-th subband, carry out error compensation in list band band.

Concretely comprise the following steps：

(A), carry out afterbody zero padding along distance to time-domain signal, so that distance is reached in step one to discrete sampling points ns.

(B), radar signal is entered with row distance to Fourier transformation, and be multiplied by the matched filtering function S in step one_refEnter Horizontal pulse is compressed.The echo-signal arranging now l-th orientation time-ofday signals is S_l,

S_l=[S_l(f_τ)]

The thermal compensation signal arranging k-th subband is S_comref_k, then

Both multiplications obtain the list band signal S_com after phase error compensation_k=S_l×S_comref_k.

Step 5：Repeat step four, error in the band of each subband signal is compensated.Now each, in subband Phase error has removed, and residual error is mostly derived from the time delay error of intersubband.With the first subband as time reference, meter Calculate the time delay of each subband：

K-th subband signal is carried out with Fourier transformation to frequency domain, be multiplied by the frequency-domain linear phase place that the removal time prolongs to every Individual subband signal carries out time delay and processes, and removes time delay function and is：

Step 6：The mode that the frequency-region signal of each subband in step 5 is carried out with afterbody zero padding carries out a liter sampling, zero padding Count as (N-1) × ns, wherein N is subband total number.Signal is done with inverse Fourier transform (IFFT), transforms to time domain, then this When signal sample rate be Fs (Fs=N × fs).It is multiplied by phase compensation function in time domain and carry out frequency displacement so that subband signal Positioned at its corresponding frequency position.For k-th subband, its frequency displacement function is

S_fmove=exp { j2 π f_c(k)τ}

Wherein：τ be distance to the fast time,

Step 7：After each subband signal is carried out frequency displacement process, if k-th subband signal is Sub_k.Enter in frequency domain Row splicing, joining method is added for frequency domain, and the part overlapping for intersubband takes its average, as shown in Figure 4.

Step 8：Process for not affecting follow-up imaging algorithm, refill distance to frequency modulation frequency modulation item, penalty function is：

Signal is entered row distance to inverse Fourier transform (IFFT), transforms to time domain, you can obtain splicing the time domain completing Signal.

Embodiment：Using the inventive method, process one group of three subband and splice SAR on-board data, radar parameter such as table 1 institute Show.

Table 1 radar parameter

Step one：Read in frame in calibration data of list band, extracted in band by mistake by this frame in calibration data of list band Poor and interchannel time delay.Concrete steps are divided into again：

A, k-th subband internal calibration data of reading, for ease of carrying out fast Fourier transform, mend to time-domain signal Zero, in original time domain signal afterbody zero padding so as to the power side always counted as 2, now the mid frequency of kth subband signal is f_c K (), sample rate is f_s, signal counts as ns.List band signal is carried out with fast Fourier transform (FFT) to frequency domain, it is entered Row matched filtering, removes frequency modulation quadratic term.Matched filtering function is:

Wherein：f_τBe distance to discrete sampling frequency,K_rFor frequency modulation rate.

In the present embodiment, design parameter is：Fs=320Mhz, each subband points ns=16384, Kr=6E13Hz/ after zero padding S, fc (1)=9.34GHz, fc (2)=9.63GHz, fc (3)=9.92GHz.

The inverse of the range value of signal in each frequency for the signal after B, extraction matched filtering, is designated as A_inv：

A_inv_k=1/A_k(f_τ)

Wherein：A_k(f_τ) for discrete sampling frequency be f_τThe range value at place.

C, the frequency-region signal afterbody zero padding after matched filtering is carried out time domain and rise sampling, zero padding is counted as Ns-ns, after zero padding Signal is counted as Ns (Ns=r × ns), and wherein r is to rise a sample rate, carries out fast Fourier transform for convenience, r is also taken as 2 power Power, and r >=32.Then rising the sample rate after sampling is f_up(f_up=r × f_s).Convert the signal into time domain, obtain now pulse The position n of signal maximum after compression_max, obtain the pulse compression maximum moment t of k-th subband_k=n_max/f_up.This enforcement In example, the design parameter of a frame in calibration signal is：R=32, t1=4.500426136363636 μ s, t2= 10.334505208333334 μ s, t3=17.16783854166667 μ s.

Step 2：By the method for average superposition denoising process the range error that each subband internal calibration signal extracts and when Between postpone, improve the method estimated accuracy.Concrete steps are divided into again：

(1), the i-th frame in calibration signal of k-th subband signal is processed according to A, B two step of step one, obtain The reciprocal function A_inv of range value_{k_i}, A_inv_{k_i}=1/A_{k_i}(f_τ),

(2), process the i-th frame in calibration signal of k-th subband signal according to C in setting steps one, obtain pulse compression The maximum moment is t_{k_i}.

(3), according to A, B, C in step 2, the calibration signal of k-th subband signal is processed, total coprocessing m is set Frame, obtains the range value reciprocal function A_inv of each frame signal_{k_i}With time delay t_k(i), wherein i=1,2...m.To m In signal amplitude inverse, each frequency element and time delay are averaged the average amplitude reciprocal function obtaining k-th subbandWith the pulse compression maximum momentThe impact of white Gaussian noise in signal can be reduced, for lifting estimated accuracy m ≥100.

Wherein:

Design parameter in the present embodiment is：

(4), each subband signal is processed according to step 2 (1), (2), (3), obtain the average of all subbands Amplitude reciprocal functionWith the pulse compression maximum moment

Step 3：Remove each subband first order corresponding with time delay, make the internal calibration signal compressed-bit of different sub-band Set moves to zero moment.Process the phase error of each subband by the method for phase gradient and average superposition denoising.

Concretely comprise the following steps：

<1>, the i-th frame in calibration signal of k-th subband signal is carried out at process of pulse-compression according to A in step one Reason, remove signal distance to quadratic phase.

<2>, by k-th subband signal after pulse compression i-th frame in calibration frequency-region signal be multiplied by linear phase ginseng Examine signal, make the internal calibration signal pulse compression peak of different sub-band be moved to zero moment.This goes linear phase with reference to letter Number it is：

<3>, setting signal now be S_{k_i}(f_τ), calculate phase gradient and remove phase error, the phase place of setting frequency Gradient is

Wherein arg represents and takes phase operation.

<4>, phase error that every frame signal of k-th subband signal extracted ask for phase again after being overlapped summation Position, reduces effect of noise in signal

<5>, phase gradient is integrated sue for peace, obtain the phase place of each frequency phase error composition of k-th subband signal Error, can be expressed as Φ_k：

Step 4：Read the radar return data of k-th subband, carry out error compensation in list band band.Concretely comprise the following steps：

(A), carry out afterbody zero padding along distance to time-domain signal, so that distance is reached in step one to discrete sampling points ns.

(B), radar signal is entered with row distance to Fourier transformation, and be multiplied by the matched filtering function S in step one_refEnter Horizontal pulse is compressed.The echo-signal arranging now l-th orientation time-ofday signals is S_l,

The thermal compensation signal arranging k-th subband is S_comref_k, then

Both multiplications obtain the list band signal S_com after phase error compensation_k=S_l×S_comref_k.

Step 5：Repeat step four, error in the band of each subband signal is compensated.Now each, in subband Phase error has removed, and residual error is mostly derived from the time delay error of intersubband.With the first subband as time reference, meter Calculate the time delay of each subband：

K-th subband signal is carried out with Fourier transformation to frequency domain, be multiplied by the frequency-domain linear phase place that the removal time prolongs to every Individual subband signal carries out time delay and processes, and removes time delay function and is：

Step 6：The mode that the frequency-region signal of each subband in step 5 is carried out with afterbody zero padding carries out a liter sampling, zero padding Count as (N-1) × ns, wherein N is subband total number.Signal is done with inverse Fourier transform (IFFT), transforms to time domain, then this When signal sample rate be Fs (Fs=N × fs).It is multiplied by phase compensation function in time domain and carry out frequency displacement so that subband signal Positioned at its corresponding frequency position.For k-th subband, its frequency displacement function is

S_fmove=exp { j2 π f_c(k)τ}

Wherein τ be distance to the fast time,

Step 7：After each subband signal is carried out frequency displacement process, if k-th subband signal is Sub_k.Enter in frequency domain Row splicing, joining method is added for frequency domain, and the part overlapping for intersubband takes its average, as shown in Figure 4.

Step 8：Process for not affecting follow-up imaging algorithm, refill distance to frequency modulation frequency modulation item, penalty function is

Signal is entered row distance to inverse Fourier transform (IFFT), transforms to time domain, you can obtain splicing the time domain completing Signal.

In order to the effectiveness of the method is described, verified by processing one group of three subband splicing SAR on-board data.Logical Cross error triple channel internal calibration data processing being extracted in subband and the time delay of intersubband, carry out two-dimension focusing it Between error compensation and splicing are carried out to echo data.Obtain height finally by two-dimension focusing process is carried out to signal after splicing Image in different resolution.

The echo-signal of the point target that internal calibration data can be 1 as scattering coefficient.By Fig. 5, dotted line in Fig. 6 and Fig. 7 The shown error compensation that do not carry out directly carries out the result of matched filtering process it is found that through radar emission receives link Afterwards, the focusing effect of its amplitude phase error heavy damage subband signal, its impact becomes when subband splices more to project.By scheming 5, Fig. 6 and Fig. 7 shown in solid carries out according to method in the present invention that error is extracted and compensation deals obtain result and understand, the present invention Method can efficiently extract interior in internal calibration signal subband and intersubband amplitude phase error.

From Fig. 8, Fig. 9, Figure 10 can be seen that the error that not only can effectively extract inband signaling by the inventive method, Realize effective focusing of list band signal, the error of intersubband can also be obtained, thus realizing the relevant fusion of subband, lifting SAR system range resolution.

Claims (2)

1. a kind of subband joining method based on internal calibration data, including following step：

Step one：Read in list band frame in calibration data, by this frame in calibration data of list band extract band in error and Interchannel time delay；

Concrete steps are divided into again：

A, read k-th subband internal calibration data, zero padding is carried out to time-domain signal, in original time domain signal afterbody zero padding so as to The power side always counted as 2, now the mid frequency of kth subband signal is f_cK (), sample rate is f_s, signal counts as ns；Right List band signal carries out fast Fourier transform to frequency domain, and it is carried out with matched filtering, removes frequency modulation quadratic term；

Matched filtering function is:

$S_{filter}=\exp \left\{\frac{{{\mathrm{j\&pi;f}}_{\mathrm{\&tau;}}}^2}{K_r}\right\}$

Wherein：f_τBe distance to discrete sampling frequency,K_rFor frequency modulation rate；

The inverse of the range value of signal in each frequency for the signal after B, extraction matched filtering, is designated as A_inv：

A_inv_k=1/A_k(f_τ)

Wherein：A_k(f_τ) for discrete sampling frequency be f_τThe range value at place；

C, the frequency-region signal afterbody zero padding after matched filtering is carried out time domain and rise sampling, zero padding counts as Ns-ns, signal after zero padding Count as Ns, Ns=r × ns, wherein r is to rise a sample rate, and r is taken as 2 power side, and r >=32；Then rise the sample rate after sampling For f_up, f_up=r × f_s；Convert the signal into time domain, obtain the position n of signal maximum after now pulse compression_max, obtain kth The pulse compression maximum moment t of individual subband_k=n_max/f_up；

Step 2：The range error of each subband internal calibration signal extraction is processed by the method for average superposition denoising and the time prolongs Late, improve the estimated accuracy of the method；

Concrete steps are divided into again：

(1), the i-th frame in calibration signal of k-th subband signal is processed according to A, B two step of step one, obtain amplitude The reciprocal function A_inv of value_{k_i}；

(2), process the i-th frame in calibration signal of k-th subband signal according to C in setting steps one, obtain pulse compression maximum The value moment is t_{k_i}；

(3), according to A, B, C in step 2, the calibration signal of k-th subband signal is processed, total coprocessing m frame is set, obtains Range value reciprocal function A_inv to each frame signal_{k_i}With time delay t_k(i), wherein i=1,2...m；To m signal width In degree inverse, each frequency element and time delay are averaged the average amplitude reciprocal function obtaining k-th subbandAnd arteries and veins The punching press contracting maximum moment

Wherein:

$\begin{array}{c}\stackrel{\&OverBar;}{A\_{\mathrm{inv}}_k}=\stackrel{\&OverBar;}{A\_{\mathrm{inv}}_k}\left(f_{\mathrm{\&tau;}}\right)\\ =\&lsqb;\frac{1}{m}\underset{i=1}{\overset{m}{\&Sigma;}}A\_{\mathrm{inv}}_{k\_i}\left(f_{\mathrm{\&tau;}}\right)\&rsqb;\end{array}$

$\stackrel{\&OverBar;}{t_k}=\frac{1}{m}\underset{i=1}{\overset{m}{\&Sigma;}}{t}_{k\_i}$

(4), each subband signal is processed according to step 2 (1), (2), (3), obtain the average amplitude of all subbands Reciprocal functionWith the pulse compression maximum moment

Step 3：Remove each subband first order corresponding with time delay, make the internal calibration signal compressed-bit set of different sub-band Move to zero moment, process the phase error of each subband by the method for phase gradient and average superposition denoising；

Concretely comprise the following steps：

<1>, according to A in step one, process of pulse-compression is carried out to the i-th frame in calibration signal of k-th subband signal, remove letter Number distance to quadratic phase；

<2>, by k-th subband signal after pulse compression i-th frame in calibration frequency-region signal be multiplied by linear phase with reference to letter Number, make the internal calibration signal pulse compression peak of different sub-band be moved to zero moment, this goes the linear phase reference signal to be：

$S_{k\_delinear}=\exp \{j2\mathrm{\&pi;}\&lsqb;f_{\mathrm{\&tau;}}+f_c\left(k\right)\&rsqb;\stackrel{\&OverBar;}{t_k}\}$

<3>, setting signal now be S_{k_i}(f_τ), calculate phase gradient and remove phase error, the phase gradient of each frequency is set For

${\stackrel{\&CenterDot;}{\mathrm{\&Phi;}}}_{k\_i}=\arg \left\{S_{k\_i}\left(f_{\mathrm{\&tau;}}\right)S_{k\_i}^{*}(f_{\mathrm{\&tau;}}+\frac{f_s}{ns})\right\}$

Wherein：Arg represents and takes phase operation；

<4>, phase error that every frame signal of k-th subband signal extracted ask for phase place again after being overlapped summation：

${\stackrel{\&CenterDot;}{\mathrm{\&Phi;}}}_k\left(f_{\mathrm{\&tau;}}\right)=\arg \left\{\underset{i=1}{\overset{m}{\&Sigma;}}{S}_{k\_i}\left(f_{\mathrm{\&tau;}}\right)S_{k\_i}^{*}(f_{\mathrm{\&tau;}}+\frac{f_s}{ns})\right\}$

<5>, phase gradient is integrated sue for peace, obtain the phase error of each frequency phase error composition of k-th subband signal, It is expressed as Φ_k(f_τ)：

$\mathrm{\&Phi;}(-\frac{f_s}{2})=0$

${\mathrm{\&Phi;}}_k\left(f_{\mathrm{\&tau;}}\right)=\underset{i=1}{\overset{l}{\&Sigma;}}\stackrel{\&CenterDot;}{\mathrm{\&Phi;}}(-\frac{f_s}{2}+l\frac{f_s}{ns}),l=1...(ns-1)$

Step 4：Read the radar return data of k-th subband, carry out error compensation in list band band；

Concretely comprise the following steps：

(A), carry out afterbody zero padding along distance to time-domain signal, make distance reach the ns in step one to discrete sampling points；

(B), radar signal is entered with row distance to Fourier transformation, and be multiplied by the matched filtering function S in step one_filterCarry out Pulse compression；The echo-signal arranging now l-th orientation time-ofday signals is S_l,

S_l=[S_l(f_τ)]

The thermal compensation signal arranging k-th subband is S_comref_k, then

$S\_{\mathrm{comref}}_k=\stackrel{\&OverBar;}{A\_inv}\left(f_{\mathrm{\&tau;}}\right)\exp (-{\mathrm{\&Phi;}}_k(f_{\mathrm{\&tau;}}\left)\right)$

Both multiplications obtain the list band signal S_com after phase error compensation_k=S_l×S_comref_k；

Step 5：Repeat step four, error in the band of each subband signal is compensated；Now the phase place in each subband is missed Difference has removed, and with the first subband as time reference, calculates the time delay of each subband：

$\stackrel{\&OverBar;}{{\mathrm{\&Delta;t}}_k}=\stackrel{\&OverBar;}{t_k}-\stackrel{\&OverBar;}{t_1}$

K-th subband signal is carried out with Fourier transformation to frequency domain, be multiplied by the frequency-domain linear phase place that the removal time prolongs to every height Band signal carries out time delay and processes, and removes time delay function and is：

$S_{delay}=\exp (-j2\mathrm{\&pi;}(f_{\mathrm{\&tau;}}+f_c\left(k\right)\left)\stackrel{\&OverBar;}{{\mathrm{\&Delta;t}}_k}\right)$

Step 6：The mode that the frequency-region signal of each subband in step 5 is carried out with afterbody zero padding carries out a liter sampling, and zero padding is counted For (N-1) × ns, wherein N is subband total number；Inverse Fourier transform is done to signal, transforms to time domain, then now the adopting of signal Sample rate is Fs, Fs=N × fs；It is multiplied by a phase compensation function in time domain and carry out frequency displacement so that subband signal is located at its correspondence Frequency position, for k-th subband, its frequency displacement function is

S_fmove=exp { j2 π f_c(k)τ}

Wherein：τ be distance to the fast time,

Step 7：After each subband signal is carried out frequency displacement process, if k-th subband signal is Sub_k；Spliced in frequency domain Process, joining method is added for frequency domain, the part overlapping for intersubband takes its average；

$S_{combine}=\underset{k=1}{\overset{N}{\&Sigma;}}{\mathrm{Sub}}_k$

Step 8：Refill distance to frequency modulation frequency modulation item, penalty function is：

$S_{ref}=\exp \left\{-\frac{{{\mathrm{j\&pi;F}}_{\mathrm{\&tau;}}}^2}{K_r}\right\},F_{\mathrm{\&tau;}}=-\frac{F_s}{2},(-\frac{F_s}{2}+\frac{Fs}{N\&times;ns})\mathrm{...}(\frac{Fs}{2}-\frac{Fs}{N\&times;ns})$

Signal is entered row distance to inverse Fourier transform, transforms to time domain, obtain splicing the time-domain signal completing.

2. a kind of subband joining method based on internal calibration data according to claim 1, in step 2, m >=100.