ibul 2004 
QUANTITATIVE MEASURES FOR THE EVALUATION OF CAMERA STABILITY 
A. F. Habib, A. M. Pullivelli *, M. Morgan? 
a 258 : : : ; : : 
Department of Geomatics Engineering, University of Calgary 
2500 University Drive N.W, Calgary AB T2N 1N4 Canada 
Email - habib(@geomatics.ucalgary.ca, ampulliv(@ucalgary.ca, mfmorgan(aucalgary.ca 
c 
TS — PS: Working Group I/2 Sensor Calibration and Testing 
KEY WORDS: Photogrammetry, Camera, Parameters, Reliability, Comparison, Calibration, Analysis, Close Range. 
ABSTRACT: 
Increasing resolution and reducing costs of off-the-shelf digital cameras are giving rise to their utilization in traditional and new 
photogrammetric applications, and allowing amateur users to generate high-quality photogrammetric products. For most, if not all 
photogrammetric applications, the internal metric characteristics of such cameras need to be determined and analyzed. This is 
achieved by going through a camera calibration and stability analysis process using a specific test field configuration. In a traditional 
test field, precisely surveyed ground control points (GCPs) are used as control information. The proposed test field in this research 
involves the utilization of linear features. Two quantitative methods for testing camera stability are introduced, where the degree of 
similarity between reconstructed bundles from two sets of Interior Orientation Parameters (IOP) is evaluated. In addition, an 
illustration of the test field created for the experiments as well as a few technical details on each camera used in the calibrations are 
presented. Through experimentation, the stability of the estimated IOP of each camera over a period of eight months is quantified 
and analyzed. 
1. INTRODUCTION 
The primary objective of photogrammetry is to generate spatial 
and descriptive information from two-dimensional imagery. 
Since its inception, the use of film metric cameras has been the 
norm in photogrammetric projects. However, the role of digital 
cameras in such projects has been rising along with its rapid 
development, ease of use and availability. 
In order to generate reliable and accurate three-dimensional 
information using such cameras, their internal characteristics, 
which are customarily known as the Interior Orientation 
Parameters (IOP), have to be modelled and carefully estimated. 
To determine the IOP, camera calibration is the universally- 
employed technique. Camera calibration requires control 
information, which is usually available in the form of a test 
field. Traditional calibration test fields consist of distinct and 
specifically marked points or targets (Fryer, 1996). Establishing 
and maintaining a conventional test field, as well as carrying 
out the calibration procedure, require professional surveyors 
and photogrammetrists. Such requirements limit the potential 
use of high quality and low cost digital cameras, and hence, a 
calibration test field consisting of straight lines and tie points 
can be adopted as an alternative for representing control 
information. 
For calibration, images covering the test field are acquired and 
incorporated in a bundle adjustment with self-calibration 
procedure to simultaneously estimate the IOP of the 
implemented camera and the Exterior Orientation Parameters 
(EOP) of the exposure stations. The results from different 
calibration sessions are then used in an IOP comparison 
procedure to check the stability of the implemented camera. 
Statistical testing is a possible methodology that can be utilized 
to accept or reject the hypothesis that the estimated IOP from 
these calibration sessions are equivalent. However, this 
methodology makes a number of idealized assumptions and 
does not provide a meaningful measure to show the differences 
between bundles or other possible discrepancies in the object 
space that could arise from using different IOP. Therefore, the 
methodology used in this research is a bundle comparison 
procedure that quantifies the degree of similarity between 
reconstructed bundles from two sets of IOP. In this research, 
there are two methods of quantifying this similarity, which are 
an Image Space comparison and an Object Space comparison. 
The methodology behind these calibration and stability analysis 
procedures was proposed by Habib et al (2002-a) and Habib 
and Morgan (2004). 
A number of amateur and professional cameras ranging in price 
from $500 to $6000 USD are used in the calibration and 
stability analysis. For each camera, a number of calibration 
datasets are produced. Each calibration dataset provides a set of 
IOP that is used to reconstruct a bundle of light rays where one 
bundle from one set of IOP is compared to another bundle from 
another set of IOP. By quantifying the difference between the 
two sets, an inference can be made on how similar the two sets 
are. 
The paper is organized in the following manner: 
e Section 2 provides a concise description of the calibration 
math model as well as the advantages and various 
approaches for utilizing straight lines in the calibration 
procedure. 
e Section 3 outlines the methodology for stability analysis 
using statistical testing, as well as the two proposed 
methodologies where the degree of similarity is evaluated 
between reconstructed bundles using two sets of IOP. 
e Section 4 provides a description of the test field and the 
cameras employed in the experiments. 
e Section 5 primarily focuses on the experimentation results 
including an analysis of the results. 
e Section 6 concludes with a brief summary and 
recommendations for future work.