DEFORMATION MONITORING OF A SLOPE 
BY VISION METROLOGY 
S. Miura™*, S. Hattori ®, K. Akimoto, S. Nishiyama 
* Kajima Technical Research Institute, 2-1-1 Tobitakyu, Chofu-shi, Tokyo, Japan - miuras@kajima.com 
’ Fucluty of Eng., Fukuyama Unversity, Gakuencho 1, Fukuyama-shi, Hiroshima, Japan - hattori@fuip.fukuyama- 
u.ac.jp 
¢ Shikoku Polytechnic College, Gunge-cho3202, Marugame-shi, Kagawa, Japan - akimoto@shikoku-pc.ac.jp 
* Fucluty of Eng., Kyoto University, Yoshidahoncho, Sakyo, Kyoto, Japan - nisiyama@geotech.kuciv.kyoto-u.ac.jp 
Commission PS WG V/l 
KEY WORDS: Photogrammetry, Statistics, Design, Measurement, Monitoring, Simulation, Experiment, Three-dimensional 
ABSTRACT: 
This paper discusses monitoring of slope deformations by vision metrology with a CCD camera. Reflective targets are placed over a 
slope, and their object coordinates is measured by a photogrammetric technique. Precision and sensitivity of slope deformation 
measurement using vision metrology are investigated. Deformation of targets placed on a slope was detected by measurement at 
two time epochs using hypothesis testing, and a series of equations is derived for the detection. The strengths of the observation 
networks were evaluated from three view points, 1.e. precision of target object coordinates, sensitivity of observations and reliability 
of observation. Model experiments were carried out to verify the method's validity. A slope model of 1.1 m X 0.5 m in size was 
constructed. An reasonable exposure configuration is looked for, which is capable of detecting displacement of about 2 mm pro 30 
m. lt is thus clarified that sufficient precision, sensitivity and reliability are achievable for practical use by a total of 12 exposures: 
four for cach of three locations. 
1l. INTRODUCTION 
It is important to carry out periodic observations of slope 
deformations, both for disaster prevention during construction 
and for maintenance / management. Methods for detecting 
slope deformation include measurement of object coordinates of 
targets placed in danger locations using GPS observation 
networks, measurement with observation networks of high 
precision total station, and installation of strain sensors such as 
optical fibers. However, these methods have not been widely 
employed due to their long measurement time and/or high cost. 
This study proposes a method for measuring displacements with 
vision metrology using a digital camera (Fraser, 1984; Fraser, 
1985). 
In general it is hard to keep an ideal observation configuration 
for in-situ slope measurement, unlike for industrial 
measurements. Network design sceks observation conditions 
that can give satisfactory measurement results. However, it is 
difficult to obtain an analytical solution. An observation 
configuration is often pre-determined by geographical 
observation constraints and prior knowledge of displacements. 
In this study a deformed location is assumed to be predictable 
in advance. And our purpose is to obtain an appropriate 
observation configuration to detect whether deformation has 
occurred on the slope. From practical point of view, it is not 
assumed that any absolute control points are available, but 
assumed that a few of fixed points exist. 
2. DEFORMATION OBSERVATION MODELS AND 
DETECTION CAPABILITY CRITERIA 
Figure 1 shows a typical model of a slope and camera 
configuration. The X, Y and Z axes are defined as horizontal, 
  
Corresponding author. 
vertical and up-dipping directions against the sloping plane. 
Assuming that an unstable part of the slope is known, an 
environment for detecting whether Block B moved against the 
upper A region was considered. In reality, there are many sites 
that are continuously monitored to determine whether or not 
existing cracks have extended. It should be natural to place the 
targets in grid pattern both side of the boundary line as shown 
in Figure 1. Because there is a limited number of photo taking 
positions on the road slope, it is assumed that a photo is taken 
from below the road looking up. 
In general, the following four points were taken into account in 
the deformation detection (Kiamehr, 2003;Benzao, 1995). 
(1) Observation precision — This refers to the internal 
precision obtained from a variance-covariance matrix for 
the space coordinates. The space coordinates need to 
satisfy the given precision requirements. 
(2) Deformation detection sensitivity — When object 
coordinates are measured with two epochs of time, the 
probability of first order and second order errors needs to 
be sufficiently low for the lower limit of the deformation 
to be detected. 
(3) Gross error detection reliability — When gross errors are 
included, observation networks need have enough 
redundancy to be able to detect and delete them. Well 
known detection methods include the data snooping 
method, the balanced least square method and the robust 
estimate method (Koch, 1999a). 
(4) Observation cost — Although it is important, it is difficult 
to formulate so it is not take into account. 
    
   
  
   
    
   
   
  
  
   
    
   
    
    
    
    
   
     
    
    
   
   
   
    
   
   
    
   
    
    
    
    
      
   
    
   
   
   
   
   
    
   
  
    
Internat 
F 
3. 
Target 
P(X}. Y, 
adjustm: 
(hereina 
v= 4; 
where V 
(m, q) ¢ 
orientati 
is a disc 
the weig 
Observe 
indepen 
By appl 
eliminati 
obtained 
yz4 
en 
Because 
observat 
added s 
becomes 
coordina 
X pâreg