745 
HYBRID MEASUREMENT SCENARIOS IN AUTOMATED 
CLOSE-RANGE PHOTOGRAMMETRY 
Simon Cronk, Clive S. Fraser 
Department of Geomatics, University of Melbourne, Victoria 3010, Australia - (c.fraser, cronks)@unimelb.edu.au 
Commission III, WG HI/1 and TS 22 
KEY WORDS: close-range photogrammetry, automation, hybrid measurement, coloured retro-reflective targets, iWitnessPRO 
ABSTRACT: 
Automated close-range photogrammetric measurement has traditionally been associated with expensive specialist cameras and 
usually requires highly controlled image illumination conditions. In the past few years, however, there has been a significant amount 
of research into the development of automated routines and algorithms that support the employment of consumer-grade colour 
digital cameras for everyday photogrammetric measurement tasks. These include, but are not limited to, fully automatic network 
orientation, image point correspondence determination and camera calibration. This paper discusses the further requirement of a 
hybrid-measurement approach, where automatic orientation routines are supported, as well as follow-up manual and semi-automatic 
operations such as natural feature point extraction, manual point measurement and operator-assisted surface definition via image 
matching. Control over image illumination is not critical; however the use of retro-reflective targeting ensures a high level of process 
automation prior to the manual phase of hybrid measurement. Topics discussed include the exploitation of colour from imagery and 
targeting, issues regarding automatic network orientation and image point correspondence determination in normally exposed 
imagery, operator-assisted surface extraction, and a special condition placed on the bundle adjustment to support both automatic and 
manual image measurements. 
1. INTRODUCTION 
At the heart of the success of digital close-range 
photogrammetry for industrial measurement has been the ability 
of vision metrology systems to measure 3D coordinates of 
targeted object point arrays fully automatically. In order to 
achieve this, a number of developments have been required in 
digital camera technology, in photogrammetric processing 
algorithms, and in object point targeting. Most notable in the 
point signalisation area has been the continued use of retro- 
reflective targeting techniques, which generally demand a 
controlled lighting environment, if fully automatic image 
measurement is to be successful. For practical and accuracy 
reasons, panchromatic imagery has been exclusively employed 
to date in high-end vision metrology systems (e.g. Ganci and 
Handley, 1998). 
What is being witnessed today, however, is the increasing use 
of off-the-shelf, SLR-type digital cameras, which offer 
can significantly complicate the associated automatic network 
orientation stage. All phases of this process become more 
complex: (i) it is more difficult to robustly detect coded and 
other signalised points in the imagery; (ii) the subsequent image 
point correspondence problem is complicated by the much 
greater number of scanned ‘targets’, both real and erroneous; 
and (iii) there is a requirement for more robust blunder 
detection techniques within the network formation and bundle 
adjustment stages, because of the likely high number of 
plausible yet wrong image point matches, resulting from the 
correspondence determination stage. Aggravating these issues 
is the invariably less than favourable network geometry, which 
is often an unavoidable characteristic of measurement networks. 
Developments to be discussed in the paper include: the merits 
of colour for codes in normally exposed imagery; fully 
moderate to high metric performance at considerable cost 
savings over their dedicated photogrammetric counterparts. 
These sensors are exclusively colour, and so it has not been 
surprising to see the attributes of colour being adopted in close- 
range photogrammetric systems (Fraser et al., 2005; Cronk et 
al., 2006). More recently, colour retro-reflective targeting has 
found application in network configurations that incorporate 
both fully automatic and manual (unsignalised) point 
measurement (e.g. Fraser & Cronk, 2007). The attributes of 
colour imaging from consumer-grade cameras have greatly 
enhanced the prospects for productive ‘hybrid measurement’ in 
non-controlled environments, thus giving rise to some distinctly 
new application areas of close-range photogrammetry. 
One of the often under-appreciated problems associated with 
utilising lighting conditions conducive to normally exposed 
imagery that is required to support manual digitizing, is that it 
automatic network orientation; optional image point 
correspondence determination; and subsequent manual and 
semi-automatic hybrid measurement operations. iWitnessPRO, 
a newly-developed software package is also discussed 
(Photometrix, 2008). This paper also describes recent 
developments that have been undertaken to enhance the use of 
automated photogrammetric measurement in two new and quite 
distinct applications. These are the reverse engineering of often 
complex stairways to facilitate accurate design and installation 
of in-home stairlifts, and the deformation or so-called crush 
measurement of vehicles involved in traffic accidents to support 
accident reconstruction and analysis. A characteristic of both 
these systems is the requirement for fully automatic 
photogrammetric network orientation, coupled with subsequent 
manual 3D digitising of selected non-targeted features such as 
points and lines.