最近在研究SAR成像算法时,需要用到Frequency Scaling算法,看了一段时间的资料,将FS算法做一个总结,有不到之处欢迎大家批评指正。
代码实现部分可以参考我的GitHub:
https://github.com/Huang-Zh-club/Frequency-Scaling
1.FS算法与CS算法的区别
首先,接触FS算法时一般都已经接触了CS算法和RD算法,RD算法在距离徙动校正时需要插值,与CS和FS算法差异较大,这里不深究。那么CSA与FSA都是用到了在频域(波数域)进行相乘以变标,那么其差别和联系在哪里呢?
无论是CSA或者FSA其本质都是要完成距离徙动校正来实现聚焦,只不过实现方式不同,我认为其差别体现在以下几个方面:
- CSA处理的信号需要时线性调频的;而FSA处理的信号则不要求是线性调频的,而是经过解线性频调的回波
- CSA在频域(多普勒频域以及二维频域)进行相位补偿操作;FSA则在波数域进行相位补偿操作
- 所以CSA将RCMC分为补余RCMC和一致RCMC,在完成补余RCMC之后,不同距离的RCM变成一致,而后再进行一致RCMC、距离压缩和方位压缩等等;而FSA则通过变标操作使得不同距离的RCM一致,而后在波数域进行RCM校正、SRC和方位压缩等等
由于由频率调制实现的变标或者平移不能太大,否则会引起不利的信号中心频率和带宽改变。所以两种算法的RCMC都分了两步进行
2.FS算法
FS算法是一种改进的线频调变标算法,该算法不要求信号在距离上是线性调频的,而是直接处理距离解线频调后的 SAR 数据。
以下用到的仿真参数设置如下
2.1解线频调
解线频调利用LFM信号的特征,用时间固定,频率和调频率相同的LFM信号作为参考信号,用它和回波做差频处理,也是一种脉冲压缩的方式。
发射LFM信号为
s
(
t
^
,
t
m
)
=
r
e
c
t
(
t
^
T
p
)
exp
(
j
2
π
(
f
c
t
^
+
1
2
γ
t
^
2
)
)
s\left( {\hat t,{t_m}} \right) = {\mathop{\rm rect}\nolimits} \left( {\frac{{\hat t}}{{{T_p}}}} \right)\exp \left( {j2\pi \left( {{f_c}\hat t + \frac{1}{2}\gamma {{\hat t}^2}} \right)} \right)
s(t^,tm)=rect(Tpt^)exp(j2π(fct^+21γt^2))
其中,
t
m
t_{m}
tm为慢时间(包含在斜距历程中),
t
^
\hat t
t^为快时间,
f
c
f_{c}
fc为中心频率,
γ
\gamma
γ为调频率,
T
p
T_{p}
Tp为脉宽。
则经过目标反射,回波信号为
s
r
(
t
^
,
t
m
)
=
A
r
e
c
t
(
t
^
−
2
R
t
/
c
T
p
)
exp
(
j
2
π
(
f
c
(
t
^
−
2
R
t
c
)
+
1
2
γ
(
t
^
−
2
R
t
c
)
2
)
)
{s_r}\left( {\hat t,{t_m}} \right) = {\mathop{\rm Arect}\nolimits} \left( {\frac{{\hat t - 2{R_t}/c}}{{{T_p}}}} \right)\exp \left( {j2\pi \left( {{f_c}\left( {\hat t - \frac{{2{R_t}}}{c}} \right) + \frac{1}{2}\gamma {{\left( {\hat t - \frac{{2{R_t}}}{c}} \right)}^2}} \right)} \right)
sr(t^,tm)=Arect(Tpt^−2Rt/c)exp(j2π(fc(t^−c2Rt)+21γ(t^−c2Rt)2))
参考距离为
R
r
e
f
R_{ref}
Rref,则参考信号为
s
r
(
t
^
,
t
m
)
=
A
r
e
c
t
(
t
^
−
2
R
r
e
f
/
c
T
p
)
exp
(
j
2
π
(
f
c
(
t
^
−
2
R
r
e
f
c
)
+
1
2
γ
(
t
^
−
2
R
r
e
f
c
)
2
)
)
{s_r}\left( {\hat t,{t_m}} \right) = {\mathop{\rm Arect}\nolimits} \left( {\frac{{\hat t - 2{R_{ref}}/c}}{{{T_p}}}} \right)\exp \left( {j2\pi \left( {{f_c}\left( {\hat t - \frac{{2{R_{ref}}}}{c}} \right) + \frac{1}{2}\gamma {{\left( {\hat t - \frac{{2{R_{ref}}}}{c}} \right)}^2}} \right)} \right)
sr(t^,tm)=Arect(Tpt^−2Rref/c)exp(j2π(fc(t^−c2Rref)+21γ(t^−c2Rref)2))
则解线频调就是将回波信号与参考信号的共轭相乘,
s
i
f
(
t
^
,
t
m
)
=
s
r
(
t
^
,
t
m
)
⋅
s
r
e
f
∗
(
t
^
,
t
m
)
{s_{if}}\left( {\hat t,{t_m}} \right) = {s_r}\left( {\hat t,{t_m}} \right) \cdot s_{ref}^*\left( {\hat t,{t_m}} \right)
sif(t^,tm)=sr(t^,tm)⋅sref∗(t^,tm)
则解线频调之后的信号即
s
i
f
(
t
^
,
t
m
)
=
A
r
e
c
t
(
t
^
−
2
R
t
/
c
T
p
)
exp
(
−
j
4
π
c
γ
(
t
^
−
2
R
r
e
f
c
)
R
Δ
)
exp
(
−
j
4
π
c
f
c
R
Δ
)
exp
(
j
4
π
γ
c
2
R
Δ
2
)
{s_{if}}\left( {\hat t,{t_m}} \right) = {\mathop{\rm Arect}\nolimits} \left( {\frac{{\hat t - 2{R_t}/c}}{{{T_p}}}} \right)\exp \left( { - j\frac{{4\pi }}{c}\gamma \left( {\hat t - \frac{{2{R_{ref}}}}{c}} \right){R_\Delta }} \right)\\ \exp \left( { - j\frac{{4\pi }}{c}{f_c}{R_\Delta }} \right)\exp \left( {j\frac{{4\pi \gamma }}{{{c^2}}}R_\Delta ^2} \right)
sif(t^,tm)=Arect(Tpt^−2Rt/c)exp(−jc4πγ(t^−c2Rref)RΔ)exp(−jc4πfcRΔ)exp(jc24πγRΔ2)
其中 R Δ = R t − R r e f {R_\Delta } = {R_t} - {R_{ref}} RΔ=Rt−Rref,可以看出在一个信号周期里,即距离差为定值,则差频处理后的信号变成了单频信号,差频输出的信号频率为
f = 2 γ R Δ / c f=2 \gamma R_{\Delta}/c f=2γRΔ/c
差频后的时间轴也不再表示距离信息,这时的时间轴与频谱频率的关系为
f
=
f
c
+
γ
τ
f=f_c+\gamma \tau
f=fc+γτ
由于时间和频率轴的含义发生了变化,因而也叫‘时频变换’。
这里借用参考书籍中的解线频调的示意图来说明这一过程
对于图(a),我们的发射信号是LFM,所以发射信号频率是关于时间的线性函数,并且调制到载频上所以信号的中心频率是 f c f_c fc,不同距离处的回波的时间延迟是不同的,所以图示的远距离回波、场景中心回波、近距离回波在时间轴上的位置不同。在一个信号周期里,快时间认为是定值,解调差频输出的信号频率为 2 γ R Δ / c 2 \gamma R_{\Delta}/c 2γRΔ/c ,对于某一个距离该值也是常数,所以图(b)的回波是平行于横轴的。
下面是对一个点目标的回波仿真结果,回波数据即已经解线频调后的回波(是一个Na*Nr的矩阵),然后抽取一行方位向数据,根据相位表达式,方位向时刻确定,信号相位是关于快时间的线性函数;抽取一列距离向数据,根据相位表达式,距离向向时刻确定,是关于慢时间的二次函数,所以中间震荡慢,两边震荡快。
2.2 FS算法
解线频调后的信号为
s
(
X
a
,
τ
)
=
σ
rec
t
[
X
a
−
X
a
c
L
]
⋅
rect
[
τ
−
2
c
(
R
t
−
R
r
e
f
)
T
p
]
⋅
exp
[
j
Φ
(
X
a
,
τ
)
]
Φ
(
X
a
,
τ
)
=
−
4
π
⋅
γ
c
⋅
τ
⋅
(
R
t
−
R
r
e
f
)
+
4
π
⋅
γ
c
2
⋅
(
R
t
−
R
r
e
f
)
2
−
4
π
⋅
f
c
c
⋅
(
R
t
−
R
r
e
f
)
\begin{array}{l} s\left(X_{a}, \tau\right)=\sigma \operatorname{rec} t\left[\frac{X_{a}-X_{a c}}{L}\right] \cdot \operatorname{rect}\left[\frac{\tau-\frac{2}{c}\left(R_{t}-R_{r e f}\right)}{T_{p}}\right] \cdot \exp \left[j \Phi\left(X_{a}, \tau\right)\right] \\ \Phi\left(X_{a}, \tau\right)=-\frac{4 \pi \cdot \gamma}{c} \cdot \tau \cdot\left(R_{t}-R_{r e f}\right)+\frac{4 \pi \cdot \gamma}{c^{2}} \cdot\left(R_{t}-R_{r e f}\right)^{2}\\ -\frac{4 \pi \cdot f_{c}}{c} \cdot\left(R_{t}-R_{r e f}\right) \end{array}
s(Xa,τ)=σrect[LXa−Xac]⋅rect[Tpτ−c2(Rt−Rref)]⋅exp[jΦ(Xa,τ)]Φ(Xa,τ)=−c4π⋅γ⋅τ⋅(Rt−Rref)+c24π⋅γ⋅(Rt−Rref)2−c4π⋅fc⋅(Rt−Rref)
X a X_a Xa和 X a c X_{ac} Xac分别为信号方位位置和孔径中心位置, τ \tau τ为距离时间(这里的 τ = t ^ − 2 R r e f / 2 R r e f c c \tau {\rm{ = }}\hat t - {{2{R_{ref}}} \mathord{\left/ {\vphantom {{2{R_{ref}}} c}} \right.} c} τ=t^−2Rref/2Rrefcc,即改变波束中心), γ \gamma γ为线性调频率, R r e f R_{ref} Rref 为参考距离。第一项和第三项为有用信息,分别用于距离分辨和方位分辨,第二项为剩余视频相位(RVP),实际处理中需要消除。其中瞬时斜距为
R t = ( X a − x 0 ) 2 + R B 2 {R_t} = \sqrt {{{\left( {{X_a} - {x_0}} \right)}^2} + R_B^2} Rt=(Xa−x0)2+RB2
大时间带宽条件下, s ( X a , τ ) s(X_a,\tau) s(Xa,τ)可以表示为参看附录证明1
s ( X a , τ ) = { A ⋅ exp [ j 4 π γ c ⋅ ( f c γ + τ ) ⋅ ( R r e f − R t ) ] } ⊗ exp [ − j π γ τ 2 ] A ( X a , τ ) = C ⋅ rect [ X a − X a c L ] ⋅ rec t [ τ T p ] \begin{array}{l} s\left(X_{a}, \tau\right)=\left\{A \cdot \exp \left[j \frac{4 \pi \gamma}{c} \cdot\left(\frac{f_{c}}{\gamma}+\tau\right) \cdot\left(R_{r e f}-R_{t}\right)\right]\right\} \otimes \exp \left[-j \pi \gamma \tau^{2}\right] \\ A\left(X_{a}, \tau\right)=C \cdot \operatorname{rect}\left[\frac{X_{a}-X_{a c}}{L}\right] \cdot \operatorname{rec} t\left[\frac{\tau}{T_{p}}\right] \end{array} s(Xa,τ)={A⋅exp[jc4πγ⋅(γfc+τ)⋅(Rref−Rt)]}⊗exp[−jπγτ2]A(Xa,τ)=C⋅rect[LXa−Xac]⋅rect[Tpτ]
即信号可以表示成有用部分和剩余相位的卷积。在未进行解线频调之前表示都是用时间表示距离,但是解线频调信号去斜后的基频回波用频率差值表示距离,所以解线频调又称为‘时频变换’。
在解线性频调之前可以知道,不同场景目标处的回波的时延
2
R
t
(
t
)
/
c
2{R_t}\left( t \right)/c
2Rt(t)/c不同,在距离时间轴上看起来,不同目标的回波中心不同,也就是‘斜的’;再将信号表示成(8)之后,信号的卷积符号前的部分
τ
=
t
^
−
2
R
r
e
f
/
2
R
r
e
f
c
c
\tau {\rm{ = }}\hat t - {{2{R_{ref}}} \mathord{\left/ {\vphantom {{2{R_{ref}}} c}} \right.} c}
τ=t^−2Rref/2Rrefcc,也就是中心全都在同样的距离位置,所以卷积符号后边部分相当于对信号完成了‘置斜’操作,而将信号表示成卷积的形式就是为了便于‘去斜’。
变标操作
由于FS是在波数域进行的,首先做变量替换把距离时间域转化为距离波数域
K R c = 4 π f c c Δ K R = 4 π γ τ c K R = K R c + Δ K R b = 8 π γ c 2 \begin{array}{c} K_{R c}=\frac{4 \pi f_{c}}{c} \\ \Delta K_{R}=\frac{4 \pi \gamma \tau}{c} \\ K_{R}=K_{R c}+\Delta K_{R} \\ b=\frac{8 \pi \gamma}{c^{2}} \end{array} KRc=c4πfcΔKR=c4πγτKR=KRc+ΔKRb=c28πγ
注意,这里仅仅做了变量替换,在代码中信号还是原来的信号,只不过表示形式在推导公式时有变化,则信号可以表示成
s ( X a , Δ K R ) = { A ⋅ exp [ − j K R ( R t − R r e f ) ] } ⊗ exp [ − j Δ K R 2 2 b ] A ( X a , Δ K R ) = C ⋅ rect [ X a − X a c L ] ⋅ rec [ 2 Δ K R b c T p ] \begin{array}{l} s\left(X_{a}, \Delta K_{R}\right)=\left\{A \cdot \exp \left[-j K_{R}\left(R_{t}-R_{r e f}\right)\right]\right\} \otimes \exp \left[-j \frac{\Delta K_{R}^{2}}{2 b}\right] \\ A\left(X_{a}, \Delta K_{R}\right)=C \cdot \operatorname{rect}\left[\frac{X_{a}-X_{a c}}{L}\right] \cdot \operatorname{rec}\left[\frac{2 \Delta K_{R}}{b c T_{p}}\right] \end{array} s(Xa,ΔKR)={A⋅exp[−jKR(Rt−Rref)]}⊗exp[−j2bΔKR2]A(Xa,ΔKR)=C⋅rect[LXa−Xac]⋅rec[bcTp2ΔKR]
做方位向傅里叶变换将信号变换到二维波数域参看附录证明2
∫ A exp [ − j K R ( ( X − x 0 ) 2 + R B 2 − R r e f ) ] exp ( − j K X X ) d X \int {A\exp \left[ { - j{K_R}\left( {\sqrt {{{\left( {X - {x_0}} \right)}^2} + {R_B}^2} - {R_{ref}}} \right)} \right]} \exp \left( { - j{K_X}X} \right)dX ∫Aexp[−jKR((X−x0)2+RB2−Rref)]exp(−jKXX)dX
得到信号二维波数域解析式
S ( K X , Δ K R ) = { A ⋅ exp [ j K R R r e f ] ⋅ exp [ − j K R 2 − K x 2 R B − j K X x 0 ] } ⊗ exp [ − j Δ K R 2 2 b ] S\left( {{K_X},\Delta {K_R}} \right) = \left\{ {A \cdot \exp \left[ {j{K_R}{R_{ref}}} \right] \cdot \exp \left[ { - j\sqrt {{K_R}^2 - {K_x}^2} {R_B} - j{K_X}{x_0}} \right]} \right\} \otimes \exp \left[ { - j\frac{{\Delta K_R^2}}{{2b}}} \right] S(KX,ΔKR)={A⋅exp[jKRRref]⋅exp[−jKR2−Kx2RB−jKXx0]}⊗exp[−j2bΔKR2]
做近似得到,参看附录
S ( K X , Δ K R ) = { A ⋅ exp [ − j ( R B A X − R r e f ) Δ K R ] ⋅ exp ( j K R c R r e f ) ⋅ exp [ − j A X K R c R B − j K X x 0 ] ⋅ exp [ j R B K X 2 2 K R c 3 A X 3 Δ K R 2 ] ⋅ exp [ − j R B K X 2 2 K R c 4 A X 5 Δ K R 3 ] } ⊗ exp [ − j Δ K R 2 2 b ] A ( K X , Δ K R ) = C ⋅ r e c t [ K X R B L K R 2 − K X 2 + X a c − x 0 L ] ⋅ r e c t [ 2 Δ K R b c T p ] A X = 1 − ( K X K R c ) 2 \begin{array}{c} S\left( {{K_X},\Delta {K_R}} \right) = \left\{ \begin{array}{c} A \cdot \exp \left[ { - j\left( {\frac{{{R_B}}}{{{A_X}}} - {R_{ref}}} \right)\Delta {K_R}} \right] \cdot \exp \left( {j{K_{Rc}}{R_{ref}}} \right)\\ \cdot \exp \left[ { - j{A_X}{K_{Rc}}{R_B} - j{K_X}{x_0}} \right] \cdot \exp \left[ {j\frac{{{R_B}K_X^2}}{{2K_{Rc}^3A_X^3}}\Delta K_R^2} \right]\\ \cdot \exp \left[ { - j\frac{{{R_B}K_X^2}}{{2K_{Rc}^4A_X^5}}\Delta K_R^3} \right] \end{array} \right\}\\ \otimes \exp \left[ { - j\frac{{\Delta K_R^2}}{{2b}}} \right]\\ A\left( {{K_X},\Delta {K_R}} \right) = C \cdot {\mathop{\rm rect}\nolimits} \left[ {\frac{{{K_X}{R_B}}}{{L\sqrt {K_R^2 - K_X^2} }} + \frac{{{X_{ac}} - {x_0}}}{L}} \right] \cdot {\mathop{\rm rect}\nolimits} \left[ {\frac{{2\Delta {K_R}}}{{bc{T_p}}}} \right]\\ {A_X} = \sqrt {1 - {{\left( {\frac{{{K_X}}}{{{K_{Rc}}}}} \right)}^2}} \end{array} S(KX,ΔKR)=⎩⎪⎪⎪⎨⎪⎪⎪⎧A⋅exp[−j(AXRB−Rref)ΔKR]⋅exp(jKRcRref)⋅exp[−jAXKRcRB−jKXx0]⋅exp[j2KRc3AX3RBKX2ΔKR2]⋅exp[−j2KRc4AX5RBKX2ΔKR3]⎭⎪⎪⎪⎬⎪⎪⎪⎫⊗exp[−j2bΔKR2]A(KX,ΔKR)=C⋅rect[LKR2−KX2KXRB+LXac−x0]⋅rect[bcTp2ΔKR]AX=1−(KRcKX)2
参看附录证明3这里的 A X A_X AX其实就是CS算法中的徙动因子。FSA通过在二维波数域乘以线性调频函数使得距离波数展缩一定的倍数。可以看出,二维波数域中 Δ K R \Delta {K_R} ΔKR的一次项系数为 R r e f − R B / R B A X A X {R_{ref}} - {{{R_B}} \mathord{\left/ {\vphantom {{{R_B}} {{A_X}}}} \right.} {{A_X}}} Rref−RB/RBAXAX,即不同斜距的距离徙动不同,为了消除这一现象,可以让一次项系数进行展缩,使得其中的 A X A_X AX消除,这样不同距离的RCM就一致了。并且斜视角越大, A X A_X AX(瞬时斜视角的余弦值)越小,即距离波数展宽越大,为了使展宽后波数仍在处理带宽内,引入一个实数 α \alpha α,使得 A X α A_X \alpha AXα尽量接近于1。一些文献是没有这个实数的,处理的时候根据参数设置具体分析
下图展示了三个不同距离的目标回波的距离徙动曲线,变标就是为了把近距以及远距的徙动曲线从虚线校正到实线位置,而参考位置的距离徙动也是一条曲线,就对应了 A X A_X AX是随着方位波数变化的。
H F S ( K X , Δ K R ) = exp [ j Δ K R 2 2 b ( 1 − α A X ) ] {H_{FS}}\left( {{K_X},\Delta {K_R}} \right) = \exp \left[ {j\frac{{\Delta K_R^2}}{{2b}}\left( {1 - \alpha {A_X}} \right)} \right] HFS(KX,ΔKR)=exp[j2bΔKR2(1−αAX)]
乘以变标函数后经过一定的变换,点目标的回波信号为,推导参见附录4
S 1 ( K X , Δ K R ) = { A 1 exp [ j α ( A X R r e f − R B ) Δ K R + j α 2 R B K X 2 2 K R c 3 A X Δ K R 2 ] ⋅ exp [ j K R C R r e f ] ⋅ exp [ j α A X Δ K R 2 2 b ( 1 − α A X ) ] ⋅ exp [ − j A X K R C R B − j K X x 0 ] ⋅ exp [ − j α 3 R B K X 2 2 K R c 4 A X 2 Δ K R 3 ] } ⊗ exp [ − j α A X Δ K R 2 2 b ] \begin{array}{c} {S_1}\left( {{K_X},\Delta {K_R}} \right) = \left\{ {{A_1}\exp \left[ {j\alpha \left( {{A_X}{R_{ref}} - {R_B}} \right)\Delta {K_R} + j\frac{{{\alpha ^2}{R_B}K_X^2}}{{2K_{Rc}^3{A_X}}}\Delta K_R^2} \right]} \right.\\ \cdot \exp \left[ {j{K_{RC}}{R_{ref}}} \right] \cdot \exp \left[ {j\frac{{\alpha {A_X}\Delta K_R^2}}{{2b}}\left( {1 - \alpha {A_X}} \right)} \right]\\ \cdot \exp \left[ { - j{A_X}{K_{RC}}{R_B} - j{K_X}{x_0}} \right]\left. { \cdot \exp \left[ { - j\frac{{{\alpha ^3}{R_B}K_X^2}}{{2K_{Rc}^4A_X^2}}\Delta K_R^3} \right]} \right\}\\ \otimes \exp \left[ { - j\frac{{\alpha {A_X}\Delta K_R^2}}{{2b}}} \right] \end{array} S1(KX,ΔKR)={A1exp[jα(AXRref−RB)ΔKR+j2KRc3AXα2RBKX2ΔKR2]⋅exp[jKRCRref]⋅exp[j2bαAXΔKR2(1−αAX)]⋅exp[−jAXKRCRB−jKXx0]⋅exp[−j2KRc4AX2α3RBKX2ΔKR3]}⊗exp[−j2bαAXΔKR2]
其中 A 1 = C ⋅ A ( K X , α A X Δ K R ) {A_1} = C \cdot A\left( {{K_X},\alpha {A_X}\Delta {K_R}} \right) A1=C⋅A(KX,αAXΔKR),这里可以看到一次项系数发生了改变 A X R r e f − R B {A_X}{R_{ref}} - {R_B} AXRref−RB,即这时的参考位置可以看成为 A X R r e f A_X R_{ref} AXRref。
在距离波数域的卷积即是距离域的相乘,所以对变标后的信号作距离逆傅里叶变换,剩余视频相位去除通过乘以距离参考函数参见附录证明5
H R V P C ( K X , Y S ) = exp [ − j b Y S 2 2 α A X ] {H_{RVPC}}\left( {{K_X},{Y_S}} \right) = \exp \left[ { - j\frac{{bY_S^2}}{{2\alpha {A_X}}}} \right] HRVPC(KX,YS)=exp[−j2αAXbYS2]
这时,二维波数域的信号形式即前面中卷积符号前的形式,即完成了‘去斜操作’。对比变标前后的信号形式可知变标操作引入了新的关于 Δ K R \Delta K_R ΔKR的二次相位误差,应该在后续操作前予以去除(在哪个域引入的就在那个域去除),对信号做距离傅里叶变换,在二维波数域乘以参考函数
H I F S ( K X , Δ K R ) = exp [ j α A X Δ K R 2 2 b ( α A X − 1 ) ] {H_{IFS}}\left( {{K_X},\Delta {K_R}} \right) = \exp \left[ {j\frac{{\alpha {A_X}\Delta K_R^2}}{{2b}}\left( {\alpha {A_X} - 1} \right)} \right] HIFS(KX,ΔKR)=exp[j2bαAXΔKR2(αAX−1)]
这相当于逆变标的过程。
距离徙动校正及压缩
完成变标操作之后,这时不同距离点目标回波的距离弯曲相等。这时信号的二维波数域为
S 1 ( K X , Δ K R ) = { A 1 exp [ j α ( A X R r e f − R B ) Δ K R + j α 2 R B K X 2 2 K R c 3 A X Δ K R 2 ] ⋅ exp [ − j A X K R C R B − j K x x 0 ] ⋅ exp [ − j α 3 R B K X 2 2 K R c 4 A X 2 Δ K R 3 ] } \begin{array}{c} {S_1}\left( {{K_X},\Delta {K_R}} \right) = \left\{ {{A_1}\exp \left[ {j\alpha \left( {{A_X}{R_{ref}} - {R_B}} \right)\Delta {K_R} + j\frac{{{\alpha ^2}{R_B}K_X^2}}{{2K_{Rc}^3{A_X}}}\Delta K_R^2} \right]} \right.\\ \cdot \exp \left[ { - j{A_X}{K_{RC}}{R_B} - j{K_x}{x_0}} \right]\left. { \cdot \exp \left[ { - j\frac{{{\alpha ^3}{R_B}K_X^2}}{{2K_{Rc}^4A_X^2}}\Delta K_R^3} \right]} \right\} \end{array} S1(KX,ΔKR)={A1exp[jα(AXRref−RB)ΔKR+j2KRc3AXα2RBKX2ΔKR2]⋅exp[−jAXKRCRB−jKxx0]⋅exp[−j2KRc4AX2α3RBKX2ΔKR3]}
在二维波数域乘以线性相位函数对距离徙动进行校正
H R M C ( K X , Δ K R ) = exp [ j α A X R r e f Δ K R − j α R s Δ K R ] {H_{RMC}}\left( {{K_X},\Delta {K_R}} \right) = \exp \left[ {j\alpha {A_X}{R_{ref}}\Delta {K_R} - j\alpha {R_s}\Delta {K_R}} \right] HRMC(KX,ΔKR)=exp[jαAXRrefΔKR−jαRsΔKR]
其中 R s R_s Rs为天线与场景中心的最近距离。二次距离压缩
H S R C ( K X , Δ K R ) = exp [ − j α 2 R B K X 2 2 K R c 3 A X Δ K R 2 ] ⋅ exp [ j α 3 R B K X 2 2 K R c 4 A X 2 Δ K R 3 ] {H_{SRC}}\left( {{K_X},\Delta {K_R}} \right) = \exp \left[ { - j\frac{{{\alpha ^2}{R_B}K_X^2}}{{2K_{Rc}^3{A_X}}}\Delta K_R^2} \right] \cdot \exp \left[ {j\frac{{{\alpha ^3}{R_B}K_X^2}}{{2K_{Rc}^4A_X^2}}\Delta K_R^3} \right] HSRC(KX,ΔKR)=exp[−j2KRc3AXα2RBKX2ΔKR2]⋅exp[j2KRc4AX2α3RBKX2ΔKR3]
进行距离向傅里叶逆变换完成距离压缩之后,乘以匹配函数进行方位压缩
H A S ( K X , Y s ) = exp [ j K R c A X R B ] {H_{AS}}\left( {{K_X},{Y_s}} \right) = \exp \left[ {j{K_{Rc}}{A_X}{R_B}} \right] HAS(KX,Ys)=exp[jKRcAXRB]
最后做方位向傅里叶逆变换得到时域解析形式。
整个FSA的流程框图如下图所示。
注意:编程问题,方位波数的范围以及采样间隔怎么设置?可以从波数的定义来考虑。振荡频率
f
c
f_c
fc的单频平面波(角频率为
ω
c
=
2
π
f
c
{\omega _c} = 2\pi {f_c}
ωc=2πfc)沿着
l
l
l方向传播,其时空表达式为
σ
e
j
(
ω
c
t
−
K
l
)
\sigma {e^{j\left( {{\omega _c}t - Kl} \right)}}
σej(ωct−Kl),振幅为常数,相位
ψ
(
t
,
l
)
=
ω
c
t
−
K
l
\psi \left( {t,l} \right) = {\omega _c}t - Kl
ψ(t,l)=ωct−Kl,
K
K
K其中称为波数或者空间(角)频率,
ω
c
=
∂
ψ
∂
t
{\omega _c} = \frac{{\partial \psi }}{{\partial t}}
ωc=∂t∂ψ,
K
=
−
∂
ψ
∂
l
K = - \frac{{\partial \psi }}{{\partial l}}
K=−∂l∂ψ ,电磁波传输速度
c
=
∂
l
∂
t
=
∂
l
∂
ψ
∂
ψ
∂
t
=
ω
c
K
c = \frac{{\partial l}}{{\partial t}}{\rm{ = }}\frac{{\partial l}}{{\partial \psi }}\frac{{\partial \psi }}{{\partial t}}{\rm{ = }}\frac{{{\omega _c}}}{K}
c=∂t∂l=∂ψ∂l∂t∂ψ=Kωc即
K
=
ω
c
c
=
2
π
f
c
c
=
2
π
λ
K = \frac{{{\omega _c}}}{c} = \frac{{2\pi {f_c}}}{c}{\rm{ = }}\frac{{2\pi }}{\lambda }
K=cωc=c2πfc=λ2π 。
距离波数的定义是这里的2倍
Δ
K
R
=
2
2
π
γ
τ
c
\Delta {K_R} = 2\frac{{2\pi \gamma \tau }}{c}
ΔKR=2c2πγτ,是为了方便处理。
那么在方位向,运动是传感器引起的,速度即是传感器的速度,则波数
k
a
=
2
π
f
a
/
2
π
f
a
V
V
{k_a} = {{2\pi {f_a}} \mathord{\left/ {\vphantom {{2\pi {f_a}} V}} \right.} V}
ka=2πfa/2πfaVV。
对点目标进行仿真,仿真结果如下
从最后的成像结果可以看出聚焦效果良好,完成了成像。
附录(公式推导)
1.卷积的证明
卷积公式为
y
(
t
)
=
x
(
t
)
⊗
h
(
t
)
=
∫
h
(
τ
)
x
(
t
−
τ
)
d
τ
y\left( t \right) = x\left( t \right) \otimes h\left( t \right) = \int {h\left( \tau \right)x\left( {t - \tau } \right)} d\tau
y(t)=x(t)⊗h(t)=∫h(τ)x(t−τ)dτ
欲证明信号可以表示成卷积的形式,只需要计算下式卷积结果(忽略包络)
s
(
X
a
,
τ
)
=
{
exp
[
j
4
π
γ
c
⋅
(
f
c
γ
+
τ
)
⋅
(
R
r
e
f
−
R
t
)
]
}
⊗
exp
[
−
j
π
γ
τ
2
]
s\left( {{X_a},\tau } \right) = \left\{ {\exp \left[ {j\frac{{4\pi \gamma }}{c} \cdot \left( {\frac{{{f_c}}}{\gamma } + \tau } \right) \cdot \left( {{R_{ref}} - {R_t}} \right)} \right]} \right\} \otimes \exp \left[ { - j\pi \gamma {\tau ^2}} \right]
s(Xa,τ)={exp[jc4πγ⋅(γfc+τ)⋅(Rref−Rt)]}⊗exp[−jπγτ2]
上式是对 τ \tau τ为变量进行卷积,所以可以写成
exp ( − j 4 π c f c R Δ ) ⋅ exp [ − j 4 π γ c ⋅ τ ⋅ R Δ ] ⊗ exp [ − j π γ τ 2 ] \exp \left( { - j\frac{{4\pi }}{c}{f_c}{R_\Delta }} \right) \cdot \exp \left[ { - j\frac{{4\pi \gamma }}{c} \cdot \tau \cdot {R_\Delta }} \right] \otimes \exp \left[ { - j\pi \gamma {\tau ^2}} \right] exp(−jc4πfcRΔ)⋅exp[−jc4πγ⋅τ⋅RΔ]⊗exp[−jπγτ2]
由于第一项中不含积分变量只计算后两项的积分即可
∫
exp
[
−
j
4
π
γ
c
⋅
(
τ
−
t
)
⋅
R
Δ
]
exp
(
−
j
π
γ
t
2
)
d
t
\int {\exp \left[ { - j\frac{{4\pi \gamma }}{c} \cdot \left( {\tau - t} \right) \cdot {R_\Delta }} \right]} \exp \left( { - j\pi \gamma {t^2}} \right)dt
∫exp[−jc4πγ⋅(τ−t)⋅RΔ]exp(−jπγt2)dt
这里利用驻定相位法计算积分,相位为 t h e t a ( t ) = − π γ t 2 − ( τ − t ) R Δ 4 π γ / 4 π γ c c theta \left( t \right) = - \pi \gamma {t^2} - \left( {\tau - t} \right){R_\Delta }{{4\pi \gamma } \mathord{\left/ {\vphantom {{4\pi \gamma } c}} \right.} c} theta(t)=−πγt2−(τ−t)RΔ4πγ/4πγcc ,求关于 t t t的导数等于0得到 t = 2 R Δ / 2 R Δ c c t = {{2{R_\Delta }} \mathord{\left/ {\vphantom {{2{R_\Delta }} c}} \right. } c} t=2RΔ/2RΔcc,将其带入相位中得
θ = − π γ ( 2 R Δ c ) 2 − 4 π γ c ( τ − 2 R Δ c ) R Δ = − 4 π γ R Δ 2 c 2 − 4 π γ c τ R Δ + 8 π γ R Δ 2 c 2 = 4 π γ R Δ 2 c 2 − 4 π γ c τ R Δ \begin{array}{c} \theta {\rm{ = }} - \pi \gamma {\left( {\frac{{2{R_\Delta }}}{c}} \right)^2} - \frac{{4\pi \gamma }}{c}\left( {\tau - \frac{{2{R_\Delta }}}{c}} \right){R_\Delta }\\ = - \frac{{4\pi \gamma {R_\Delta }^2}}{{{c^2}}} - \frac{{4\pi \gamma }}{c}\tau {R_\Delta } + \frac{{8\pi \gamma {R_\Delta }^2}}{{{c^2}}}\\ = \frac{{4\pi \gamma {R_\Delta }^2}}{{{c^2}}} - \frac{{4\pi \gamma }}{c}\tau {R_\Delta } \end{array} θ=−πγ(c2RΔ)2−c4πγ(τ−c2RΔ)RΔ=−c24πγRΔ2−c4πγτRΔ+c28πγRΔ2=c24πγRΔ2−c4πγτRΔ
所以卷积后相位为
4
π
γ
R
Δ
2
c
2
−
4
π
γ
c
τ
R
Δ
−
4
π
f
c
c
R
Δ
\frac{{4\pi \gamma {R_\Delta }^2}}{{{c^2}}} - \frac{{4\pi \gamma }}{c}\tau {R_\Delta } - \frac{{4\pi {f_c}}}{c}{R_\Delta }
c24πγRΔ2−c4πγτRΔ−c4πfcRΔ
Φ
(
X
a
,
τ
)
=
−
4
π
⋅
γ
c
⋅
τ
⋅
(
R
t
−
R
r
e
f
)
+
4
π
⋅
γ
c
2
⋅
(
R
t
−
R
r
e
f
)
2
−
4
π
⋅
f
c
c
⋅
(
R
t
−
R
r
e
f
)
\Phi \left( {{X_a},\tau } \right) = - \frac{{4\pi \cdot \gamma }}{c} \cdot \tau \cdot \left( {{R_t} - {R_{ref}}} \right) + \frac{{4\pi \cdot \gamma }}{{{c^2}}} \cdot {\left( {{R_t} - {R_{ref}}} \right)^2} - \frac{{4\pi \cdot {f_c}}}{c} \cdot \left( {{R_t} - {R_{ref}}} \right)
Φ(Xa,τ)=−c4π⋅γ⋅τ⋅(Rt−Rref)+c24π⋅γ⋅(Rt−Rref)2−c4π⋅fc⋅(Rt−Rref)
证毕
2. 方位向傅里叶变换
原式为
s
(
X
a
,
Δ
K
R
)
=
{
A
⋅
exp
[
−
j
K
R
(
R
t
−
R
r
e
f
)
]
}
⊗
exp
[
−
j
Δ
K
R
2
2
b
]
A
(
X
a
,
Δ
K
R
)
=
C
⋅
rect
[
X
a
−
X
a
c
L
]
⋅
rec
[
2
Δ
K
R
b
c
T
p
]
\begin{array}{l} s\left(X_{a}, \Delta K_{R}\right)=\left\{A \cdot \exp \left[-j K_{R}\left(R_{t}-R_{r e f}\right)\right]\right\} \otimes \exp \left[-j \frac{\Delta K_{R}^{2}}{2 b}\right] \\ A\left(X_{a}, \Delta K_{R}\right)=C \cdot \operatorname{rect}\left[\frac{X_{a}-X_{a c}}{L}\right] \cdot \operatorname{rec}\left[\frac{2 \Delta K_{R}}{b c T_{p}}\right] \end{array}
s(Xa,ΔKR)={A⋅exp[−jKR(Rt−Rref)]}⊗exp[−j2bΔKR2]A(Xa,ΔKR)=C⋅rect[LXa−Xac]⋅rec[bcTp2ΔKR]
由于卷积符号后一项无方位向时间,所以不需要进行变换
S ( K X , Δ K R ) = F [ A exp [ − j K R ( R t − R r e f ) ] ] ⊗ exp [ − j Δ K R 2 2 b ] S\left( {{K_X},\Delta {K_R}} \right) = {\bf{F}}\left[ {A\exp \left[ { - j{K_R}\left( {{R_t} - {R_{ref}}} \right)} \right]} \right] \otimes \exp \left[ { - j\frac{{\Delta K_R^2}}{{2b}}} \right] S(KX,ΔKR)=F[Aexp[−jKR(Rt−Rref)]]⊗exp[−j2bΔKR2]
则
∫
A
exp
[
−
j
K
R
(
(
X
−
x
0
)
2
+
R
B
2
−
R
r
e
f
)
]
exp
(
−
j
K
X
X
)
d
X
\int A \exp \left[-j K_{R}\left(\sqrt{\left(X-x_{0}\right)^{2}+R_{B}^{2}}-R_{r e f}\right)\right] \exp \left(-j K_{X} X\right) d X
∫Aexp[−jKR((X−x0)2+RB2−Rref)]exp(−jKXX)dX
仍是利用驻定相位法,其相位为
θ
(
X
)
=
−
K
R
(
(
X
−
x
0
)
2
+
R
B
2
−
R
r
e
f
)
−
K
X
X
\theta \left( X \right) = - {K_R}\left( {\sqrt {{{\left( {X - {x_0}} \right)}^2} + {R_B}^2} - {R_{ref}}} \right) - {K_X}X
θ(X)=−KR((X−x0)2+RB2−Rref)−KXX
对相位求关于
X
X
X的导数求得驻相点为
X
=
−
K
X
R
B
K
R
2
−
K
X
2
+
x
0
X = - \frac{{{K_X}{R_B}}}{{\sqrt {{K_R}^2 - {K_X}^2} }} + {x_0}
X=−KR2−KX2KXRB+x0
则相位为
θ
=
−
R
B
K
R
2
−
K
X
2
+
K
R
R
r
e
f
−
K
X
x
0
\theta = - {R_B}\sqrt {{K_R}^2 - {K_X}^2} + {K_R}{R_{ref}} - {K_X}{x_0}
θ=−RBKR2−KX2+KRRref−KXx0
证毕
3.近似
S
(
K
X
,
Δ
K
R
)
=
{
A
⋅
exp
[
j
K
R
R
r
e
f
]
⋅
exp
[
−
j
K
R
2
−
K
x
2
R
B
−
j
K
X
x
0
]
}
⊗
exp
[
−
j
Δ
K
R
2
2
b
]
S\left( {{K_X},\Delta {K_R}} \right) = \left\{ {A \cdot \exp \left[ {j{K_R}{R_{ref}}} \right] \cdot \exp \left[ { - j\sqrt {{K_R}^2 - {K_x}^2} {R_B} - j{K_X}{x_0}} \right]} \right\} \otimes \exp \left[ { - j\frac{{\Delta K_R^2}}{{2b}}} \right]
S(KX,ΔKR)={A⋅exp[jKRRref]⋅exp[−jKR2−Kx2RB−jKXx0]}⊗exp[−j2bΔKR2]
对上式的近似主要是针对 K R 2 − K x 2 \sqrt {{K_R}^2 - {K_x}^2} KR2−Kx2
K R 2 − K x 2 = ( K R c + Δ K R ) 2 − K x 2 = K R c 2 + Δ K R 2 + 2 K R c Δ K R − K x 2 = K R c 1 + Δ K R 2 K R c 2 + 2 Δ K R K R c − K x 2 K R c 2 \begin{array}{c} \sqrt {{K_R}^2 - {K_x}^2} = \sqrt {{{\left( {{K_{Rc}} + \Delta {K_R}} \right)}^2} - {K_x}^2} \\ = \sqrt {{K_{Rc}}^2 + \Delta {K_R}^2 + 2{K_{Rc}}\Delta {K_R} - {K_x}^2} \\ = {K_{Rc}}\sqrt {1 + \frac{{\Delta {K_R}^2}}{{{K_{Rc}}^2}} + 2\frac{{\Delta {K_R}}}{{{K_{Rc}}}} - \frac{{{K_x}^2}}{{{K_{Rc}}^2}}} \end{array} KR2−Kx2=(KRc+ΔKR)2−Kx2=KRc2+ΔKR2+2KRcΔKR−Kx2=KRc1+KRc2ΔKR2+2KRcΔKR−KRc2Kx2
令 A X = 1 − K X 2 / K X 2 K R c 2 K R c 2 {A_X} = \sqrt {1 - {{{K_X}^2} \mathord{\left/ {\vphantom {{{K_X}^2} {{K_{Rc}}^2}}} \right.} {{K_{Rc}}^2}}} AX=1−KX2/KX2KRc2KRc2则
K R 2 − K x 2 = K R c A X 2 + Δ K R 2 K R c 2 + 2 Δ K R K R c = K R c A X 1 + Δ K R 2 K R c 2 + 2 Δ K R K R c A X 2 \begin{array}{c} \sqrt {{K_R}^2 - {K_x}^2} = {K_{Rc}}\sqrt {{A_X}^2 + \frac{{\Delta {K_R}^2}}{{{K_{Rc}}^2}} + 2\frac{{\Delta {K_R}}}{{{K_{Rc}}}}} \\ = {K_{Rc}}{A_X}\sqrt {1 + \frac{{\frac{{\Delta {K_R}^2}}{{{K_{Rc}}^2}} + 2\frac{{\Delta {K_R}}}{{{K_{Rc}}}}}}{{{A_X}^2}}} \end{array} KR2−Kx2=KRcAX2+KRc2ΔKR2+2KRcΔKR=KRcAX1+AX2KRc2ΔKR2+2KRcΔKR
参考文献1中公式8.70有
K
R
c
A
X
1
+
Δ
K
R
2
K
R
c
2
+
2
Δ
K
R
K
R
c
A
X
2
=
K
R
c
A
X
[
1
+
1
A
X
2
Δ
K
R
K
R
c
−
K
X
2
K
R
c
2
2
A
X
4
(
Δ
K
R
K
R
c
)
2
+
K
X
2
K
R
c
2
2
A
X
6
(
Δ
K
R
K
R
c
)
3
]
{K_{Rc}}{A_X}\sqrt {1 + \frac{{\frac{{\Delta {K_R}^2}}{{{K_{Rc}}^2}} + 2\frac{{\Delta {K_R}}}{{{K_{Rc}}}}}}{{{A_X}^2}}} {\rm{ = }}{K_{Rc}}{A_X}\left[ {1 + \frac{1}{{{A_X}^2}}\frac{{\Delta {K_R}}}{{{K_{Rc}}}} - \frac{{\frac{{{K_X}^2}}{{{K_{Rc}}^2}}}}{{2{A_X}^4}}{{\left( {\frac{{\Delta {K_R}}}{{{K_{Rc}}}}} \right)}^2} + \frac{{\frac{{{K_X}^2}}{{{K_{Rc}}^2}}}}{{2{A_X}^6}}{{\left( {\frac{{\Delta {K_R}}}{{{K_{Rc}}}}} \right)}^3}} \right]
KRcAX1+AX2KRc2ΔKR2+2KRcΔKR=KRcAX[1+AX21KRcΔKR−2AX4KRc2KX2(KRcΔKR)2+2AX6KRc2KX2(KRcΔKR)3]
证毕。
S ( K X , Δ K R ) = { A ⋅ exp [ j Δ K R R r e f ] ⋅ exp ( j K R c R r e f ) exp [ − j A X K R c R B − j K X x 0 ] exp [ − j Δ K R A X R B ] ⋅ exp [ j R B K X 2 2 K R c 3 A X 3 Δ K R 2 ] ⋅ exp [ − j R B K X 2 2 K R c 4 A X 5 Δ K R 3 ] } ⊗ exp [ − j Δ K R 2 2 b ] \begin{array}{c} S\left( {{K_X},\Delta {K_R}} \right) = \left\{ {A \cdot \exp \left[ {j\Delta {K_R}{R_{ref}}} \right] \cdot \exp \left( {j{K_{Rc}}{R_{ref}}} \right)\exp \left[ { - j{A_X}{K_{Rc}}{R_B} - j{K_X}{x_0}} \right]\exp \left[ { - j\frac{{\Delta {K_R}}}{{{A_X}}}{R_B}} \right]} \right.\\ \left. { \cdot \exp \left[ {j\frac{{{R_B}K_X^2}}{{2K_{Rc}^3A_X^3}}\Delta K_R^2} \right] \cdot \exp \left[ { - j\frac{{{R_B}K_X^2}}{{2K_{Rc}^4A_X^5}}\Delta K_R^3} \right]} \right\} \otimes \exp \left[ { - j\frac{{\Delta K_R^2}}{{2b}}} \right] \end{array} S(KX,ΔKR)={A⋅exp[jΔKRRref]⋅exp(jKRcRref)exp[−jAXKRcRB−jKXx0]exp[−jAXΔKRRB]⋅exp[j2KRc3AX3RBKX2ΔKR2]⋅exp[−j2KRc4AX5RBKX2ΔKR3]}⊗exp[−j2bΔKR2]
则根据卷积公式
H F S ( K X , Δ K R ) = exp [ j Δ K R 2 2 b ( 1 − α A X ) ] {H_{FS}}\left( {{K_X},\Delta {K_R}} \right) = \exp \left[ {j\frac{{\Delta K_R^2}}{{2b}}\left( {1 - \alpha {A_X}} \right)} \right] HFS(KX,ΔKR)=exp[j2bΔKR2(1−αAX)]
S ( K X , Δ K R ) = { A ( K X , Δ K R ) ⋅ exp [ j Δ K R R r e f ] ⋅ exp ( j K R c R r e f ) exp [ − j A X K R c R B − j K X x 0 ] exp [ − j Δ K R A X R B ] ⋅ exp [ j R B K X 2 2 K R c 3 A X 3 Δ K R 2 ] ⋅ exp [ − j R B K X 2 2 K R c 4 A X 5 Δ K R 3 ] } ⊗ exp [ − j Δ K R 2 2 b ] ⋅ exp [ j Δ K R 2 2 b ( 1 − α A X ) ] = exp ( j K R c R r e f ) exp [ − j A X K R c R B − j K X x 0 ] ∫ A ( K X , Δ L ) exp [ j Δ L R r e f ] exp [ − j Δ L A X R B ] ⋅ exp [ j R B K X 2 2 K R c 3 A X 3 Δ L 2 ] ⋅ exp [ − j R B K X 2 2 K R c 4 A X 5 Δ L 3 ] exp ( − j ( Δ K R − Δ L ) 2 2 b ) d Δ L ⋅ exp [ j Δ K R 2 2 b ( 1 − α A X ) ] \begin{array}{c} S\left( {{K_X},\Delta {K_R}} \right) = \left\{ {A\left( {{K_X},\Delta {K_R}} \right) \cdot \exp \left[ {j\Delta {K_R}{R_{ref}}} \right] \cdot \exp \left( {j{K_{Rc}}{R_{ref}}} \right)\exp \left[ { - j{A_X}{K_{Rc}}{R_B} - j{K_X}{x_0}} \right]\exp \left[ { - j\frac{{\Delta {K_R}}}{{{A_X}}}{R_B}} \right]} \right.\\ \left. { \cdot \exp \left[ {j\frac{{{R_B}K_X^2}}{{2K_{Rc}^3A_X^3}}\Delta K_R^2} \right] \cdot \exp \left[ { - j\frac{{{R_B}K_X^2}}{{2K_{Rc}^4A_X^5}}\Delta K_R^3} \right]} \right\} \otimes \exp \left[ { - j\frac{{\Delta K_R^2}}{{2b}}} \right]\\ \cdot \exp \left[ {j\frac{{\Delta K_R^2}}{{2b}}\left( {1 - \alpha {A_X}} \right)} \right]\\ = \exp \left( {j{K_{Rc}}{R_{ref}}} \right)\exp \left[ { - j{A_X}{K_{Rc}}{R_B} - j{K_X}{x_0}} \right]\int {A\left( {{K_X},\Delta L} \right)} \exp \left[ {j\Delta L{R_{ref}}} \right]\exp \left[ { - j\frac{{\Delta L}}{{{A_X}}}{R_B}} \right]\\ \cdot \exp \left[ {j\frac{{{R_B}K_X^2}}{{2K_{Rc}^3A_X^3}}\Delta {L^2}} \right] \cdot \exp \left[ { - j\frac{{{R_B}K_X^2}}{{2K_{Rc}^4A_X^5}}\Delta {L^3}} \right]\exp \left( { - j\frac{{{{\left( {\Delta {K_R} - \Delta L} \right)}^2}}}{{2b}}} \right)d\Delta L\\ \cdot \exp \left[ {j\frac{{\Delta K_R^2}}{{2b}}\left( {1 - \alpha {A_X}} \right)} \right] \end{array} S(KX,ΔKR)={A(KX,ΔKR)⋅exp[jΔKRRref]⋅exp(jKRcRref)exp[−jAXKRcRB−jKXx0]exp[−jAXΔKRRB]⋅exp[j2KRc3AX3RBKX2ΔKR2]⋅exp[−j2KRc4AX5RBKX2ΔKR3]}⊗exp[−j2bΔKR2]⋅exp[j2bΔKR2(1−αAX)]=exp(jKRcRref)exp[−jAXKRcRB−jKXx0]∫A(KX,ΔL)exp[jΔLRref]exp[−jAXΔLRB]⋅exp[j2KRc3AX3RBKX2ΔL2]⋅exp[−j2KRc4AX5RBKX2ΔL3]exp(−j2b(ΔKR−ΔL)2)dΔL⋅exp[j2bΔKR2(1−αAX)]
由于最后一项与 Δ L \Delta L ΔL无关,所以可以写进积分符号里面,对于积分符号里面的最后两项
exp ( − j ( Δ K R − Δ L ) 2 2 b ) exp [ j Δ K R 2 2 b ( 1 − α A X ) ] = exp ( − j Δ L 2 − 2 Δ K R Δ L + α A X Δ K R 2 2 b ) = exp ( − j α A X 2 b ( Δ L 2 α 2 A X 2 − 2 Δ K R Δ L α A X + Δ K R 2 + Δ L 2 α A X − Δ L 2 α 2 A X 2 ) ) = exp ( − j α A X 2 b ( Δ K R − Δ L α A X ) 2 ) exp ( j 1 − α A X 2 b α A X Δ L 2 ) \begin{array}{l} \exp \left( { - j\frac{{{{\left( {\Delta {K_R} - \Delta L} \right)}^2}}}{{2b}}} \right)\exp \left[ {j\frac{{\Delta K_R^2}}{{2b}}\left( {1 - \alpha {A_X}} \right)} \right] = \exp \left( { - j\frac{{\Delta {L^2} - 2\Delta {K_R}\Delta L + \alpha {A_X}\Delta K_R^2}}{{2b}}} \right)\\ = \exp \left( { - j\frac{{\alpha {A_X}}}{{2b}}\left( {\frac{{\Delta {L^2}}}{{{\alpha ^2}{A_X}^2}} - \frac{{2\Delta {K_R}\Delta L}}{{\alpha {A_X}}} + \Delta K_R^2 + \frac{{\Delta {L^2}}}{{\alpha {A_X}}} - \frac{{\Delta {L^2}}}{{{\alpha ^2}{A_X}^2}}} \right)} \right)\\ = \exp \left( { - j\frac{{\alpha {A_X}}}{{2b}}{{\left( {\Delta {K_R} - \frac{{\Delta L}}{{\alpha {A_X}}}} \right)}^2}} \right)\exp \left( {j\frac{{1 - \alpha {A_X}}}{{2b\alpha {A_X}}}\Delta {L^2}} \right) \end{array} exp(−j2b(ΔKR−ΔL)2)exp[j2bΔKR2(1−αAX)]=exp(−j2bΔL2−2ΔKRΔL+αAXΔKR2)=exp(−j2bαAX(α2AX2ΔL2−αAX2ΔKRΔL+ΔKR2+αAXΔL2−α2AX2ΔL2))=exp(−j2bαAX(ΔKR−αAXΔL)2)exp(j2bαAX1−αAXΔL2)
则
= exp ( j K R c R r e f ) exp [ − j A X K R c R B − j K X x 0 ] ∫ A ( K X , Δ L ) exp [ j Δ L R r e f ] exp [ − j Δ L A X R B ] ⋅ exp [ j R B K X 2 2 K R c 3 A X 3 Δ L 2 ] ⋅ exp [ − j R B K X 2 2 K R c 4 A X 5 Δ L 3 ] exp ( − j α A X 2 b ( Δ K R − Δ L α A X ) 2 ) exp ( j 1 − α A X 2 b α A X Δ L 2 ) d Δ L \begin{array}{c} = \exp \left( {j{K_{Rc}}{R_{ref}}} \right)\exp \left[ { - j{A_X}{K_{Rc}}{R_B} - j{K_X}{x_0}} \right]\int {A\left( {{K_X},\Delta L} \right)} \exp \left[ {j\Delta L{R_{ref}}} \right]\exp \left[ { - j\frac{{\Delta L}}{{{A_X}}}{R_B}} \right]\\ \cdot \exp \left[ {j\frac{{{R_B}K_X^2}}{{2K_{Rc}^3A_X^3}}\Delta {L^2}} \right] \cdot \exp \left[ { - j\frac{{{R_B}K_X^2}}{{2K_{Rc}^4A_X^5}}\Delta {L^3}} \right]\exp \left( { - j\frac{{\alpha {A_X}}}{{2b}}{{\left( {\Delta {K_R} - \frac{{\Delta L}}{{\alpha {A_X}}}} \right)}^2}} \right)\exp \left( {j\frac{{1 - \alpha {A_X}}}{{2b\alpha {A_X}}}\Delta {L^2}} \right)d\Delta L \end{array} =exp(jKRcRref)exp[−jAXKRcRB−jKXx0]∫A(KX,ΔL)exp[jΔLRref]exp[−jAXΔLRB]⋅exp[j2KRc3AX3RBKX2ΔL2]⋅exp[−j2KRc4AX5RBKX2ΔL3]exp(−j2bαAX(ΔKR−αAXΔL)2)exp(j2bαAX1−αAXΔL2)dΔL
做替换
= exp ( j K R c R r e f ) exp [ − j A X K R c R B − j K X x 0 ] ∫ A ( K X , α A X Δ L ) exp [ j α A X Δ L 1 R r e f ] exp [ − j α A X Δ L 1 A X R B ] exp [ j R B K X 2 2 K R c 3 A X 3 ( α A X Δ L ) 2 ] ⋅ exp [ − j R B K X 2 2 K R c 4 A X 5 ( α A X Δ L 1 ) 3 ] exp ( − j α A X 2 b ( Δ K R − α A X Δ L 1 α A X ) 2 ) exp ( j 1 − α A X 2 b α A X ( α A X Δ L 1 ) 2 ) α A X d Δ L 1 = A 1 exp ( j K R c R r e f ) exp [ − j A X K R c R B − j K X x 0 ] exp [ j ( α A X R r e f − α R B ) Δ K R ] exp [ j R B K X 2 2 K R c 3 A X 3 ( α A X Δ K R ) 2 ] ⋅ exp [ − j R B K X 2 2 K R c 4 A X 5 ( α A X Δ K R ) 3 ] exp ( j 1 − α A X 2 b α A X Δ K R 2 ) ⊗ exp ( − j α A X 2 b ( Δ K R ) 2 ) \begin{array}{l} = \exp \left( {j{K_{Rc}}{R_{ref}}} \right)\exp \left[ { - j{A_X}{K_{Rc}}{R_B} - j{K_X}{x_0}} \right]\int {A\left( {{K_X},\alpha {A_X}\Delta L} \right)} \exp \left[ {j\alpha {A_X}\Delta {L_1}{R_{ref}}} \right]\\ \exp \left[ { - j\frac{{\alpha {A_X}\Delta {L_1}}}{{{A_X}}}{R_B}} \right]\exp \left[ {j\frac{{{R_B}K_X^2}}{{2K_{Rc}^3A_X^3}}{{\left( {\alpha {A_X}\Delta L} \right)}^2}} \right] \cdot \exp \left[ { - j\frac{{{R_B}K_X^2}}{{2K_{Rc}^4A_X^5}}{{\left( {\alpha {A_X}\Delta {L_1}} \right)}^3}} \right]\\ \exp \left( { - j\frac{{\alpha {A_X}}}{{2b}}{{\left( {\Delta {K_R} - \frac{{\alpha {A_X}\Delta {L_1}}}{{\alpha {A_X}}}} \right)}^2}} \right)\exp \left( {j\frac{{1 - \alpha {A_X}}}{{2b\alpha {A_X}}}{{\left( {\alpha {A_X}\Delta {L_1}} \right)}^2}} \right)\alpha {A_X}d\Delta {L_1}\\ {\rm{ = }}{A_1}\exp \left( {j{K_{Rc}}{R_{ref}}} \right)\exp \left[ { - j{A_X}{K_{Rc}}{R_B} - j{K_X}{x_0}} \right]\exp \left[ {j\left( {\alpha {A_X}{R_{ref}} - \alpha {R_B}} \right)\Delta {K_R}} \right]\\ \exp \left[ {j\frac{{{R_B}K_X^2}}{{2K_{Rc}^3A_X^3}}{{\left( {\alpha {A_X}\Delta {K_R}} \right)}^2}} \right] \cdot \exp \left[ { - j\frac{{{R_B}K_X^2}}{{2K_{Rc}^4A_X^5}}{{\left( {\alpha {A_X}\Delta {K_R}} \right)}^3}} \right]\exp \left( {j\frac{{1 - \alpha {A_X}}}{{2b}}\alpha {A_X}\Delta {K_R}^2} \right)\\ \otimes \exp \left( { - j\frac{{\alpha {A_X}}}{{2b}}{{\left( {\Delta {K_R}} \right)}^2}} \right) \end{array} =exp(jKRcRref)exp[−jAXKRcRB−jKXx0]∫A(KX,αAXΔL)exp[jαAXΔL1Rref]exp[−jAXαAXΔL1RB]exp[j2KRc3AX3RBKX2(αAXΔL)2]⋅exp[−j2KRc4AX5RBKX2(αAXΔL1)3]exp(−j2bαAX(ΔKR−αAXαAXΔL1)2)exp(j2bαAX1−αAX(αAXΔL1)2)αAXdΔL1=A1exp(jKRcRref)exp[−jAXKRcRB−jKXx0]exp[j(αAXRref−αRB)ΔKR]exp[j2KRc3AX3RBKX2(αAXΔKR)2]⋅exp[−j2KRc4AX5RBKX2(αAXΔKR)3]exp(j2b1−αAXαAXΔKR2)⊗exp(−j2bαAX(ΔKR)2)
5.剩余视频相位的去除
剩余视频相位(RVP)校正指的是去斜处理,即解卷积,消除式中的置斜处理项,即消除卷积符号后的项。常规思路是将(23)进行距离傅里叶逆变换,然后去斜,再进行傅里叶变换得到二维频域的形式。由于傅里叶逆变换比较复杂,这里根据信号卷积的性质来进行去斜。
待证明的信号其实可以简化为
x
⊗
h
x \otimes h
x⊗h,现在欲去除
h
h
h的影响,只需要进行
x
⊗
h
⊗
h
1
x \otimes h \otimes {h_1}
x⊗h⊗h1,其中
h
1
h_1
h1可以将
h
h
h的效果抵消即可,所以这一步可以转换为设计
h
h
h的匹配滤波器
h
1
h_1
h1(这样表述有无问题?)
H ( Δ K R ) = exp ( − j α A X 2 b ( Δ K R ) 2 ) H\left( {\Delta {K_R}} \right) = \exp \left( { - j\frac{{\alpha {A_X}}}{{2b}}{{\left( {\Delta {K_R}} \right)}^2}} \right) H(ΔKR)=exp(−j2bαAX(ΔKR)2)
对 H ( Δ K R ) H\left( {\Delta {K_R}} \right) H(ΔKR)进行傅里叶逆变换
∫ exp ( − j α A X 2 b ( Δ K R ) 2 ) exp ( j Y Δ K R ) d Δ K R θ ( Δ K R ) = Y Δ K R − α A X 2 b ( Δ K R ) 2 \begin{array}{l} \int {\exp \left( { - j\frac{{\alpha {A_X}}}{{2b}}{{\left( {\Delta {K_R}} \right)}^2}} \right)} \exp \left( {jY\Delta {K_R}} \right)d\Delta {K_R}\\ \theta \left( {\Delta {K_R}} \right){\rm{ = }}Y\Delta {K_R} - \frac{{\alpha {A_X}}}{{2b}}{\left( {\Delta {K_R}} \right)^2} \end{array} ∫exp(−j2bαAX(ΔKR)2)exp(jYΔKR)dΔKRθ(ΔKR)=YΔKR−2bαAX(ΔKR)2
对相位求导得到驻相点为 Δ K R = b Y α A X \Delta {K_R}{\rm{ = }}\frac{{bY}}{{\alpha {A_X}}} ΔKR=αAXbY,所以逆傅里叶变换的结果为
h ( Y ) = exp ( j b Y 2 2 α A X ) h\left( Y \right) = \exp \left( {j\frac{{b{Y^2}}}{{2\alpha {A_X}}}} \right) h(Y)=exp(j2αAXbY2)
所以滤波器设置为上式的共轭即可:
H
R
V
P
C
(
Y
)
=
exp
(
−
j
b
Y
2
2
α
A
X
)
{H_{RVPC}}\left( Y \right) = \exp \left( { - j\frac{{b{Y^2}}}{{2\alpha {A_X}}}} \right)
HRVPC(Y)=exp(−j2αAXbY2)
参考书籍:
- 魏忠铨, 等. 合成孔径雷达卫星[M]. 北京: 科学出版社.
- 保铮 邢孟道 王彤. 雷达成像技术[M]. 电子工业出版社, 2005.
- Lan G.Cumming Frank H.Wong. 合成孔径雷达成像: 算法与实现[M]. 电子工业出版社, 2012.