Professional Documents
Culture Documents
Full download Adjustment Models in 3D Geomatics and Computational Geophysics: With MATLAB Examples Bashar Alsadik file pdf all chapter on 2024
Full download Adjustment Models in 3D Geomatics and Computational Geophysics: With MATLAB Examples Bashar Alsadik file pdf all chapter on 2024
https://ebookmass.com/product/basics-of-computational-geophysics-
pijush-samui/
https://ebookmass.com/product/modern-information-optics-with-
matlab-yaping-zhang/
https://ebookmass.com/product/3d-lung-models-for-regenerating-
lung-tissue-gunilla-westergren-thorsson/
Computational Frameworks. Systems, Models and
Applications 1st Edition Edition Mamadou Kaba Traore
(Eds.)
https://ebookmass.com/product/computational-frameworks-systems-
models-and-applications-1st-edition-edition-mamadou-kaba-traore-
eds/
https://ebookmass.com/product/computational-models-of-reading-a-
handbook-1st-edition-erik-d-reichle/
https://ebookmass.com/product/computational-geo-electromagnetics-
methods-models-and-forecasts-volume-5-1st-edition-viacheslav-v-
spichak/
https://ebookmass.com/product/distributed-acoustic-sensing-in-
geophysics-methods-and-applications-yingping-li/
https://ebookmass.com/product/fundamental-chemistry-with-
matlab-1st-edition-daniele-mazza/
ADJUSTMENT MODELS IN 3D GEOMATICS
AND COMPUTATIONAL GEOPHYSICS
Computational Geophysics
ADJUSTMENT
MODELS IN 3D
GEOMATICS AND
COMPUTATIONAL
GEOPHYSICS
With MATLAB Examples
VOLUME 4
BASHAR ALSADIK
Faculty member at Baghdad University – College of Engineering – Iraq (1999–2014)
Research assistant at Twente University – ITC faculty – The Netherlands (2010–2014)
Member of the International Society for Photogrammetry and Remote Sensing ISPRS
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any
liability for any injury and/or damage to persons or property as a matter of products liability, negligence or
otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the
material herein.
Library of Congress Cataloging-in-Publication Data
A catalog record for this book is available from the Library of Congress
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library
ISBN: 978-0-12-817588-0
vii
viii PREFACE
principle and its mathematical derivation is More advanced topics are presented in the
also introduced. Then, in Chapter 2, error second half of the book, such as the unified
propagation technique is presented and the approach of least squares adjustment of
variance-covariance matrix of observations Chapter 7. This is an advanced adjustment
and unknowns is described. Similarly, the technique where the unknowns have uncer-
preanalysis procedure is presented. In tainties and are then processed as observa-
Chapter 3, the adjustment using the condi- tions in the model.
tion equations and the observation equations Chapters 8 and 9 present more related ap-
is introduced and, at the end, the concept of plications in geomatics, namely the topic of
the homogeneous least squares method is fitting 3D geometric primitives and the 3D
presented. For a better understanding, dif- transformation computations respectively.
ferent adjustment models are shown in Then the book continues to give introduction
Chapter 4 such as intersection and resection to other advanced topics such as the Kalman
either in 2D or 3D using observed distances, filter in Chapter 10 and the nonlinear least
angles, azimuths, difference in heights, or squares using Levenberg-Marquardt in
computations from images. At the end of Chapter 11. Finally, in the Chapter 12 of
Chapter 4, an important computational geo- the book, the detection, identification, and
physics application is shown of earthquake adaptation (DIA) of the postadjustment
location determination. The general adjust- techniques is introduced. The chapter pre-
ment approach is presented in Chapter 5 sents the blunder detection methods of
when there is more than one observation in data snooping, robust estimation, and the
the mathematical model related to several un- random sample consensus (RANSAC). In
knowns. The important topic of adjustment the book’s Appendix, a MATLAB code is
with constraints is introduced in Chapter 6. given for the adjustment of horizontal
Three main topics are presented in this chap- geodetic networks.
ter: adjustment with constraints, adjustment
Bashar Alsadik
with additional parameters, and the adjust-
The Netherlands
ment with inner constraints (free nets).
C H A P T E R
1
Statistical Introduction
1.1 INTRODUCTION
Adjustment Models in 3D Geomatics and Computational Geophysics 1 # 2019 Elsevier Inc. All rights reserved.
https://doi.org/10.1016/B978-0-12-817588-0.00001-5
2 1. STATISTICAL INTRODUCTION
(A) (B)
FIG. 1.1 (A) Which end of the bridge is the correctly aligned one? (B) Which athlete is the gold medal winner?
All of these questions asked of Fig. 1.1 can be answered when we understand different con-
cepts and indices in adjustment of observations and theory of errors such as: accuracy, pre-
cision, reliability, calibration, standard deviation, weighted mean, residual error, most
probable value (MPV), redundancy, etc.
• Errors are the differences between observed values and their true values. An error is what
causes values to differ when a measurement is repeated, and none of the results can be
preferred over the others. Although it is not possible to entirely eliminate error in a
measurement, it can be controlled and characterized. We summarized the terminologies
related to errors as follows:
• Gross errors: These are large errors (blunders, mistakes, or outliers) that can be avoided in
the observations; however, they don0 t follow a mathematical or physical model and may
be large or small, positive or negative. With developments in instrumentation and
automated procedures, the main source of gross errors is human-related, for example,
recording and reading errors, or observing a different target than the intended one.
Careful reading and recording of the data can significantly reduce gross errors. It should be
noted that some references don0 t count mistakes as an error type [1].
• Systematic errors
Systematic errors (or bias when having many observations) occur when following
some physical models and therefore can be checked. Systematic errors are either positive or
negative; using proper measuring procedures can eliminate some of them, whereas
some are corrected by using mathematical methods. Systematic error sources are
recognizable and can be reduced to a great extent by careful designation of the observation
system and the selection of its components. In practice, the process of calibration of
instruments, such as cameras in photogrammetry, is to detect and remove systematic
errors.
• Fig. 1.2 illustrates a systematic error-free observation (red curve) biased in a certain
direction and amount (the blue curve).
1.2 STATISTICAL DEFINITIONS AND TERMINOLOGIES 3
• Random errors Random or accidental errors are unavoidable errors that represent residuals
after removing all other types of errors. This type of error occurs in observations because
of limitations in the measuring instruments or due to limitations related to the operator,
among other affecting undetermined factors. It should be noted that random errors
can be positive or negative, and they don0 t follow a physical model. Therefore, they are
processed statistically using the probability theorem because the majority follows normal
distribution. In repetitive observations, the average or mean can be used as the MPV.
As shown in Fig. 1.3, greater random errors cause a greater dispersion of normally
distributed observations around the mean. The dispersion is measured by the standard
deviations at a certain probability as will be shown in Eq. (1.6).
Mean
FIG. 1.3 Observations having a normal probability distribution. (A) With random errors. (B) Without random
errors.
• Uncertainty
All observations have uncertainties, which is a range of values in which the true obser-
vation could lie. An uncertainty estimate should address both systematic and random er-
rors, and therefore is considered the most proper measure of expressing accuracy.
However, in many geomatics problems, the systematic error is disregarded, and only ran-
dom error is included in the uncertainty observation. When only random error is counted
in the uncertainty evaluation, it is an expression of the precision of the observation.
4 1. STATISTICAL INTRODUCTION
Accuracy
related to
systematic error
Probability
True value
Observed value
X
Precision
(related to random error)
FIG. 1.4 Accuracy and precision.
F D F D
B
B P
P
A A
C E C E
• Residual error
Residual error v means the difference between the true value and its observed value.
However, because the true value is impossible to reach, it is statistically compensated
by the MPV. Therefore:
It should be noted that the MPV value of repetitive observations is the arithmetic mean,
whereas for other observations, it is the optimal value that offers the minimal of the squared
residual errors. This concept will be explained in the least squares principle of Section 1.7. It is
important to note that the MPV in the concept of the least squares represents the adjusted
values of observations. For nonlinear geomatical problems, the adjusted values are computed
by running the solution with initial values. Accordingly:
• Arithmetic mean x: when observing a quantity x for n times under the same conditions, the
mean x is computed as follows:
Xn
xi
i¼1
x¼ ðn ! ∞ Þ (1.3)
n
• Variance σ 2: the theoretical mean of the squared residual errors that can be computed as
follows:
X n X
n
ð xi xÞ 2 ð vi Þ 2
i¼1 i¼1
σ2 ¼ ¼ (1.4)
n1 n1
The denominator n 1 is called the redundancy r or the degree of freedom. The minimum
observations we need to determine the variance is only 1 value, which is termed as no; accord-
ingly, the rest of the observations are redundant, and r can be formulated as:
r ¼ n no (1.5)
• Standard deviation σ: the square root of the variance that expresses by how much the
observations differ from the MPV or the mean value. Further, it is a measure of how spread
out the observations are. Therefore, standard deviation is an expression of precision
computed as:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX n uX n
u u
u ð xi xÞ 2 u ð vi Þ 2
t t
i¼1 i¼1
σ¼ ¼ (1.6)
n1 n1
When the reference values are available, the standard deviation is termed as the Root Mean
Squared Error (RMSE), which indicates by how much the observed or derived quantities de-
viate from the reference (true) values.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX n
u
u ðxi xREFRENCE Þ2
t
RMSE ¼ i¼1 (1.7)
n
• Standard error σx : represents the standard deviation of the mean, which is computed as:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX uX
u n u n
u ð x x Þ 2 u ð vi Þ 2
u i u
σ t i¼1 t i¼1
σ x ¼ pffiffiffi ¼ ¼ (1.8)
n nðn 1Þ nðn 1Þ
1.4 NORMAL DISTRIBUTION CURVE 7
EXAMPLE 1.1
Given
A base line AB is observed in meters 10 times as follows:
Required
Find the MPV, the standard deviation, and standard error of line AB to the nearest mm.
Solution
The MPV for repetitive observations is the mean:
X
n
xi
i¼1
x¼ ¼ 26:348 m
n
The standard deviation is computed as:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX n
u
u ðxi xÞ2
t
i¼1
σ¼ ¼ 3 mm
n1
The standard error is computed as:
σ
σ x ¼ pffiffiffi ¼ 1 mm
n
Then
AB ¼ 26:348 m 1 mm
The normal distribution curve or the gaussian curve can be calculated and then plotted
using the following Eq. (1.9).
1 v2
y ¼ pffiffiffiffiffi e 2σ 2 ¼ Keh v
2 2
(1.9)
σ 2π
where
y: the normal probability,
K ¼ σ pnIffiffiffiffi
2π
,
h ¼ 2σ2 called the accuracy index,
2 1
EXAMPLE 1.2
Given
An angle is observed 50 times in degrees as illustrated in Table 1.1:
40.3371976 40.3395228
Required
– Compute the MPV of the angle and its standard deviation.
– Compute and plot the normal probability distribution curve of residual errors.
Solution
The MPV of the angle is simply the arithmetic mean, therefore:
MPV ¼ mean of angles ¼ 40.3404387 degrees
1.4 NORMAL DISTRIBUTION CURVE 9
whereas the standards deviation is computed as:
sffiffiffiffiffiffiffiffiffiffiffi
X ffi
v2
σ¼ ¼ 9:83500
n1
To compute the probability y of the normal distribution curve, the following elements are
computed: nI
K ¼ pffiffiffiffiffi ¼ 4:553
σ 2π
1
h2 ¼ ¼ 0:0051688
2σ 2
The computed residuals are grouped between two maximum/minimum limits of 22.4500 . Then,
to have 20 regular intervals for the plotting issue, interval I is computed to be 2.24500 as illustrated in
the following MATLAB code.
Table 1.2. illustrates the details of the computations to relate the residual errors in seconds to the
computed probability y of the normal distribution curve.
Then the normal distribution curve and histogram can be plotted as shown in Fig. 1.7.
4.5
3.5
3
Probability
2.5
1.5
0.5
0
–25 –20 –15 –10 –5 0 5 10 15 20 25
Residuals sec.
FIG. 1.7 The normal distribution of residuals in seconds.
MATLAB code
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%% chapter 1 - example 1.2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clc,clear, close all
%%%%%%%%%%%%% create normally distributed observed angles %%%%%%%%%%%%%%%%%
ang=40.34+10*(randn(50,1)/3600); %% adding random normal noise of 10 sec.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% compute residuals
resid =3600*(ang-mean(ang));
Resid=unique(round(resid*100)/100);
m=max(abs(Resid)); %% make equal intervals
I= ((2*m)/20) % 20 intervals % can be changed by the user
v=-m:I:m; v=round(v’*100)/100; % sample the residuals equally in seconds
The CDF plot in Fig. 1.8B explains that 90% of the observed angles are less than or equal to
60 degrees. This property is easy to interpret even for a nonprofessional. Therefore, key values
such as minimum, maximum, median, percentiles, etc. can be directly read from the diagram.
Accordingly, CDF is an efficient description for random variable uncertainty and is a useful
tool for comparing the distribution of different sets of data.
Let’s assume that three observers A, B, and C measured the same angle 300 times each. To
compare the skill of the three observers in observing the angle, we can use the CDF plot as
shown in Fig. 1.9.
12 1. STATISTICAL INTRODUCTION
FIG. 1.8 (A) Normal distribution histogram, (B) The corresponding CDF plot.
FIG. 1.9 The CDF plots of the angle observed by three surveyors.
MATLAB code
clear;clc; close all
% angle biased randomly
A=randn(300,1)*10+50
hA=cdfplot(A);hold on
set( hA(:,1), ’Linewidth’, 3);
B=randn(300,1)*10+48
1.6 THE PROBABLE ERROR AND LEVELS OF REJECTION 13
hB=cdfplot(B);hold on
set( hB(:,1), ’Linewidth’, 3);
C=randn(300,1)*10+53
hC=cdfplot(C);hold on
set( hC(:,1), ’Linewidth’, 3);
lgd =legend(’observer A’, ’observer B’, ’observer C’,’fontsize’);
lgd.FontSize = 12;
The probable error Pe_50% is defined as the 50% or equal probability that a true error can
arise. In other words, 50% probability occurs when j vi j > Pe_50% and 50% when jvi j < Pe_50% as
shown in Fig. 1.10.
For a group of observations, the 50% probable error is the median of residual errors after
sorting them in ascending or descending order, and therefore it is sometimes termed as the
median absolute deviation (MAD) (Chapter 12). However, by using this approach, the probable
error is not sensitive to gross errors as shown in the following illustrations where the probable
error is 500 for both error sets despite the existence of gross error of 9400 .
1st set: 1″ 1″ 3″ 5″ 6″ 7″ 11″
2nd set: 1″ 1″ 3″ 5″ 6″ 7″ 94″
Statistically, to relate the probable error Pe_50% with the standard deviations, we can apply
the following derivation by the integration of the normal distribution curve equation as:
ð +Pe v2
1
Area ¼ pffiffiffiffiffiffiffiffi e 2σ 2 dv ¼ 0:5 (1.10)
2πσ Pe
68% of Area
In the same manner, the probable error of the intervals 2σ and 3σ are computed as
shown in Eq. (1.13) and Fig. 1.12.
Pð2σ ε 2σ Þ ¼ 0:955
(1.13)
Pð3σ ε 3σ Þ ¼ 0:997
FIG. 1.12 Different standard deviations limits of the normal distribution curve.
Another random document with
no related content on Scribd:
T he Chase I sing, hounds, and their various breed,
And no less various use. O thou, great Prince!
Whom Cambria’s towering hills proclaim their lord,
Deign thou to hear my bold, instructive song.
While grateful citizens, with pompous shew,
Rear the triumphal arch, rich with the exploits
Of thy illustrious house; while virgins pave
Thy way with flowers, and, as the royal youth
Passing they view, admire, and sigh in vain;
While crowded theatres, too fondly proud
Of their exotick minstrels, and shrill pipes,
The price of manhood, hail thee with a song,
And airs soft-warbling; my hoarse-sounding horn