条件平差
平差方法
基本的线代基础
(
A
B
)
T
=
B
T
A
T
[
(
A
B
)
−
1
]
T
=
[
(
A
B
)
T
]
−
1
∂
A
B
T
∂
X
=
A
T
∂
B
∂
X
+
B
T
∂
A
∂
X
\begin{aligned} &(AB)^T=B^TA^T \\ &[(AB)^{-1}]^T=[(AB)^{T}]^{-1} \\ &\frac{\partial AB^T}{\partial X}=A^T\frac{\partial B}{\partial X}+B^T\frac{\partial A}{\partial X} \end{aligned}
(AB)T=BTAT[(AB)−1]T=[(AB)T]−1∂X∂ABT=AT∂X∂B+BT∂X∂A
条件平差的函数模型:
A
V
−
W
=
0
W
=
−
(
A
L
+
A
0
)
\begin{array}{l} AV-W=0\\ W=-(AL+A_0) \end{array}
AV−W=0W=−(AL+A0)
随机模型:
D
=
σ
0
2
Q
=
σ
0
2
P
−
1
D=\sigma_0^2Q=\sigma_0^2P^{-1}
D=σ02Q=σ02P−1
估计准则:
V
T
P
V
=
min
V^TPV=\min
VTPV=min
根据拉格朗日乘数法:
Φ
=
V
T
P
V
−
2
K
T
(
A
V
−
W
)
\Phi=V^TPV-2K^T(AV-W)
Φ=VTPV−2KT(AV−W)
对其求导:
∂
Φ
∂
V
=
2
V
T
P
−
2
K
T
A
=
0
\frac{\partial \Phi}{\partial V}=2V^TP-2K^TA=0
∂V∂Φ=2VTP−2KTA=0
即:
P
T
V
=
A
T
K
→
V
=
P
−
1
A
T
K
=
Q
A
T
K
P^TV=A^TK\rightarrow V=P^{-1}A^TK=QA^TK
PTV=ATK→V=P−1ATK=QATK
其中权阵
P
P
P的转置等于其自身
将式代入: A V − W = 0 AV-W=0 AV−W=0
则有:
A
Q
A
T
K
−
W
=
0
→
K
=
(
A
Q
A
T
)
−
1
W
=
N
−
1
W
AQA^TK-W=0 \rightarrow K=(AQA^T)^{-1}W=N^{-1}W
AQATK−W=0→K=(AQAT)−1W=N−1W
再将
K
K
K的解代回:
V
=
Q
A
T
K
V=QA^TK
V=QATK
则有:
V
=
Q
A
T
N
−
1
W
=
Q
A
T
(
A
Q
A
T
)
−
1
W
V=QA^TN^{-1}W=QA^T(AQA^T)^{-1}W
V=QATN−1W=QAT(AQAT)−1W
精度评定
按照单位权中误差的定义,有:
σ
^
0
=
V
T
P
V
n
−
t
=
V
T
P
V
r
\hat \sigma_0=\sqrt{\frac{V^TPV}{n-t}}=\sqrt{\frac{V^TPV}r}
σ^0=n−tVTPV=rVTPV
上述参数都可根据平差过程中求得,在此不再赘述。
按照协因数阵与方差阵之间的关系,有:
D
L
^
L
^
=
σ
^
0
2
Q
L
^
L
^
D_{\hat L\hat L}=\hat \sigma_0^2 Q_{\hat L\hat L}
DL^L^=σ^02QL^L^
令
Z
=
[
L
W
K
V
L
^
]
=
[
I
−
A
−
N
−
1
A
−
P
−
1
A
T
N
−
1
A
I
−
P
−
1
A
T
N
−
1
A
]
L
+
[
0
−
A
0
−
N
−
1
A
0
−
P
−
1
A
T
N
−
1
A
0
−
P
−
1
A
T
N
−
1
A
0
]
Z=\left[\begin{array}{c}L\\W\\K\\V\\\hat L\end{array}\right] =\left[\begin{array}{c}I\\-A\\-N^{-1}A\\-P^{-1}A^TN^{-1}A\\I-P^{-1}A^TN^{-1}A\end{array}\right]L+\left[\begin{array}{c}0\\-A_0\\-N^{-1}A_0\\-P^{-1}A^TN^{-1}A_0\\-P^{-1}A^TN^{-1}A_0\end{array}\right]
Z=⎣⎢⎢⎢⎢⎡LWKVL^⎦⎥⎥⎥⎥⎤=⎣⎢⎢⎢⎢⎡I−A−N−1A−P−1ATN−1AI−P−1ATN−1A⎦⎥⎥⎥⎥⎤L+⎣⎢⎢⎢⎢⎡0−A0−N−1A0−P−1ATN−1A0−P−1ATN−1A0⎦⎥⎥⎥⎥⎤
此处推导
L
^
\hat L
L^的计算过程:
L
^
=
L
+
V
=
L
+
Q
A
T
N
−
1
W
=
L
+
Q
A
T
N
−
1
[
−
(
A
L
+
A
0
)
]
=
(
I
−
Q
A
T
N
−
1
A
)
L
−
Q
A
T
N
−
1
A
0
\hat L=L+V=L+QA^TN^{-1}W=L+QA^TN^{-1}[-(AL+A_0)]=(I-QA^TN^{-1}A)L-QA^TN^{-1}A_0
L^=L+V=L+QATN−1W=L+QATN−1[−(AL+A0)]=(I−QATN−1A)L−QATN−1A0
则:
Q
L
^
L
^
=
(
I
−
P
−
1
A
T
N
−
1
A
)
Q
L
L
(
I
−
P
−
1
A
T
N
−
1
A
)
T
Q_{\hat L\hat L}=(I-P^{-1}A^TN^{-1}A)Q_{LL}(I-P^{-1}A^TN^{-1}A)^T
QL^L^=(I−P−1ATN−1A)QLL(I−P−1ATN−1A)T
接下来是变换的几个关键内容:
由于: N = A Q A T N=AQA^T N=AQAT为可逆矩阵,则 ( N − 1 ) T = ( N T ) − 1 = [ ( A Q A T ) T ] − 1 = ( A Q A T ) − 1 = N − 1 (N^{-1})^T=(N^T)^{-1}=[(AQA^T)^T]^{-1}=(AQA^T)^{-1}=N^{-1} (N−1)T=(NT)−1=[(AQAT)T]−1=(AQAT)−1=N−1
将上述关系式进行化简:
Q
L
^
L
^
=
(
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
)
(
I
−
Q
L
L
A
T
N
−
1
A
)
T
=
(
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
)
(
I
−
A
T
N
−
1
A
Q
L
L
)
=
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
+
Q
L
L
A
T
N
−
1
(
A
Q
L
L
A
T
)
N
−
1
A
Q
L
L
=
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
+
Q
L
L
A
T
N
−
1
N
N
−
1
A
Q
L
L
=
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
+
Q
L
L
A
T
N
−
1
A
Q
L
L
=
Q
L
L
−
Q
L
L
A
T
N
−
1
A
Q
L
L
\begin{aligned} Q_{\hat L\hat L}&=(Q_{LL}-Q_{LL}A^TN^{-1}AQ_{LL})(I-Q_{LL}A^TN^{-1}A)^T\\ &=(Q_{LL}-Q_{LL}A^TN^{-1}AQ_{LL})(I-A^TN^{-1}AQ_{LL})\\ &=Q_{LL}-Q_{LL}A^TN^{-1}AQ_{LL}-Q_{LL}A^TN^{-1}AQ_{LL}+Q_{LL}A^TN^{-1}(AQ_{LL}A^T)N^{-1}AQ_{LL}\\ &=Q_{LL}-Q_{LL}A^TN^{-1}AQ_{LL}-Q_{LL}A^TN^{-1}AQ_{LL}+Q_{LL}A^TN^{-1}NN^{-1}AQ_{LL}\\ &=Q_{LL}-Q_{LL}A^TN^{-1}AQ_{LL}-Q_{LL}A^TN^{-1}AQ_{LL}+Q_{LL}A^TN^{-1}AQ_{LL}\\ &=Q_{LL}-Q_{LL}A^TN^{-1}AQ_{LL} \end{aligned}
QL^L^=(QLL−QLLATN−1AQLL)(I−QLLATN−1A)T=(QLL−QLLATN−1AQLL)(I−ATN−1AQLL)=QLL−QLLATN−1AQLL−QLLATN−1AQLL+QLLATN−1(AQLLAT)N−1AQLL=QLL−QLLATN−1AQLL−QLLATN−1AQLL+QLLATN−1NN−1AQLL=QLL−QLLATN−1AQLL−QLLATN−1AQLL+QLLATN−1AQLL=QLL−QLLATN−1AQLL
其余结果可参考下表:
Q
Z
Z
=
[
Q
L
L
Q
L
W
Q
L
K
Q
L
V
Q
L
^
L
^
Q
W
L
Q
W
W
Q
W
K
Q
W
V
Q
W
L
˙
Q
K
L
Q
K
W
Q
K
K
Q
K
V
Q
K
L
˙
Q
V
L
Q
V
W
Q
V
K
Q
V
V
Q
V
L
^
Q
L
^
L
Q
L
^
W
Q
L
^
K
Q
L
^
V
Q
L
^
L
^
]
=
[
Q
−
Q
A
T
−
Q
A
T
N
−
1
−
Q
A
T
N
−
1
A
Q
Q
−
Q
A
T
N
−
1
A
Q
−
A
Q
N
I
A
Q
0
−
N
−
1
A
Q
I
N
−
1
N
−
1
A
Q
0
−
Q
A
T
N
−
1
A
Q
Q
A
T
Q
A
T
N
−
1
Q
A
T
N
−
1
A
Q
0
Q
−
Q
A
T
N
−
1
A
Q
0
0
0
Q
−
Q
A
T
N
−
1
A
Q
]
\begin{aligned} Q_{Z Z} &=\left[\begin{array}{ccccc} Q_{L L} & Q_{L W} & Q_{L K} & Q_{L V} & Q_{\hat{L} \hat{L}} \\ Q_{W L} & Q_{W W} & Q_{W K} & Q_{W V} & Q_{W \dot{L}} \\ Q_{K L} & Q_{K W} & Q_{K K} & Q_{K V} & Q_{K \dot{L}} \\ Q_{V L} & Q_{V W} & Q_{V K} & Q_{V V} & Q_{V \hat{L}} \\ Q_{\hat{L} L} & Q_{\hat{L} W} & Q_{\hat{L} K} & Q_{\hat{L} V} & Q_{\hat{L} \hat{L}} \end{array}\right]\\ &=\left[\begin{array}{ccccc} Q & -Q A^{T} & -Q A^{T} N^{-1} & -Q A^{T} N^{-1} A Q & Q-Q A^{T} N^{-1} A Q \\ -A Q & N & I & A Q & 0 \\ -N^{-1} A Q & I & N^{-1} & N^{-1} A Q & 0 \\ -Q A^{T} N^{-1} A Q & Q A^{T} & Q A^{T} N^{-1} & Q A^{T} N^{-1} A Q & 0 \\ Q-Q A^{T} N^{-1} A Q & 0 & 0 & 0 & Q-Q A^{T} N^{-1} A Q \end{array}\right] \end{aligned}
QZZ=⎣⎢⎢⎢⎢⎡QLLQWLQKLQVLQL^LQLWQWWQKWQVWQL^WQLKQWKQKKQVKQL^KQLVQWVQKVQVVQL^VQL^L^QWL˙QKL˙QVL^QL^L^⎦⎥⎥⎥⎥⎤=⎣⎢⎢⎢⎢⎡Q−AQ−N−1AQ−QATN−1AQQ−QATN−1AQ−QATNIQAT0−QATN−1IN−1QATN−10−QATN−1AQAQN−1AQQATN−1AQ0Q−QATN−1AQ000Q−QATN−1AQ⎦⎥⎥⎥⎥⎤
条件平差应用
应用 | 必要起算数据个数d | 必要观测数据个数t | 测角网中条件方程的列立 |
---|---|---|---|
水准网 | 1 | ||
测角网 | 4 | 2 p − q − 4 2p-q-4 2p−q−4 | 图形条件:三角形内角和=
18
0
∘
180^\circ
180∘ 极条件:边长条件 圆周角条件:圆周角为 36 0 ∘ 360^\circ 360∘ |
侧边网 | 3 | 2 p − q − 3 2p-q-3 2p−q−3 | 图形条件:三角形内角和=
18
0
∘
180^\circ
180∘ 极条件:边长条件 圆周角条件:圆周角为 36 0 ∘ 360^\circ 360∘ |
边角网 | 3 | 2 p − q − 3 2p-q-3 2p−q−3 | 方位角条件:从一个已知方位角出发,推算到另一个已知方位角,方位角的推算值要与已知方位角相等。方位角条件的个数等于已知方位角的个数-1。 纵横坐标条件 : 从一个已知点的纵横坐标出发,推算到不同已知点组中另一个已知点时,纵横坐标的推算值要与已知值相等。纵横坐标的条件个数等于多余已知点的个数乘以2-2。 图形条件:三角形内角和= 18 0 ∘ 180^\circ 180∘ 边长条件:极条件,余弦条件 圆周角条件:圆周角为 36 0 ∘ 360^\circ 360∘ |
单一导线网 | 2 p ′ 2p' 2p′ | 方位角条件:从一个已知方位角出发,推算到另一个已知方位角,方位角的推算值要与已知方位角相等。方位角条件的个数等于已知方位角的个数-1。 纵横坐标条件 : 从一个已知点的纵横坐标出发,推算到不同已知点组中另一个已知点时,纵横坐标的推算值要与已知值相等。纵横坐标的条件个数等于多余已知点的个数乘以2-2。 | |
GPS基线向量网 | 3 p ′ 3p' 3p′ | 方位角条件 直角条件 距离条件 面积条件 |
其中t为必要观测数,p为所有三角点的个数,q为多余起算数据的个数,d为基准个数
一弧度等于多少秒:
ρ
′
′
=
(
180
∗
3600
)
′
′
π
\rho ''=\frac{(180*3600)''}{\pi}
ρ′′=π(180∗3600)′′