"m d 
wh = ®© 
Lo 
Uw =< we UW 
  
  
Chapman, David 
  
This paper documents an exercise that attempts to assess the utility of a simple, lightweight, robust range-imaging 
sensor that was felt to be suitable for low-precision survey in the complex and congested environments found typical of 
offshore oil-platforms. 
2 RANGE IMAGING SYSTEMS 
Range sensors can generate complete data of visible surfaces that are rather featureless to the human eye or a video 
camera. Furthermore, they do not require operator-assisted algorithms to generate the 3D coordinates. The 
disadvantages are that unreliable results may take place on highly reflective surfaces and where sharp range 
discontinuities exist (El-Hakim and Beraldin, 1994). Furthermore, these 3D active range sensors are usually configured 
for a specific volume of measurement or range and are not re-configurable. Those designed for large volumes are 
usually costly and bulky. 
2.1 Triangulation-Based Sensors 
For applications requiring less than 10-meter range, short and medium range sensors, which are usually based on 
structured light and triangulation, are often employed. Short range sensors (e.g. ShapeGrabber, http://www.vitana.com) 
are designed for mainly small objects placed at less than 0.5 m, while medium range (e.g. Cyberware, 
http://www.cyberware.com) can cover ranges from 0.5 m to 2 m depending on the configuration. They have been used 
successfully in model creation applied to cultural heritage and space (Beraldin et al, 1998 and 1999). For ranges 
between 2 m and 10 m, there are a limited number of triangulation-based sensors available. However, some systems 
have been built with baselines in excess of 1000 mm to cover distances up to 100 m. A sensor with such a baseline has 
limited usefulness in an application such as ours for lack of compactness. However, a more compact sensor known as 
the Random Access Camera (RAC) is based on the auto-synchronized scanning (Rioux, 1984) and has a baseline of 
only 90-mm. It can cover ranges from 0.5m to 10 m. Expected accuracy at the closest range is 0.03 mm but it is about 
20 mm at the 10-m range. A short-range sensor will require a much higher number of 3D images; thus the high 
accuracy of the individual images may be offset by larger error propagation due to registration. Medium range sensors 
require a lower number of images but the main disadvantage is that the accuracy deteriorates rapidly at longer ranges as 
expected with triangulation-based sensors. 
  
2.2 Time Of Flight — Time Delay Methods 
Long-range sensors, which are capable of covering a large volume of the scene in a single image or perhaps a few 
images at ranges exceeding 10 meters, are usually based on the time-of-flight (TOF) technology. Data from this type of 
range sensor scanners, plus texture, has been used for creating VR model of the interior of large structures (Johnson et 
al, 1997 and Miyatsuka et al, 1998). Although the large volume coverage is very convenient, since the whole scene can 
be covered with a small number of 3D images, the accuracy is usually in the 10-100 mm range. These scanners also 
tend to be more costly than shorter-range sensors. For a given site, the choice is now between using a long range sensor 
that produces less accurate data but with less registration error, or a short to medium-range sensor that produce more 
accurate data but requires larger number of images resulting in greater registration error. Sensor cost and size are also 
important factors to consider. 
2.3 The ‘Long-Range’ Biris — A Short-Baseline Active Triangulation System 
The Biris range imaging system was initially developed by the National Research Council of Canada (NRC) (Blais et 
al, 1992) and has been adapted for use as a low-cost, highly portable measurement device suited to close-range 
applications from stand-off distances of the order of 0.3m (Beraldin et.al. 1998). 
This experiment sought to evaluate a modified Biris device that had been developed to work over ranges of up to 3m 
(El-Hakim et al, 1997). The sensor is mounted on a Directed Perception pan/tilt device that enables full panoramic 
coverage from an image station through the projection and detection of a laser light stripe (figure 1). The pan-tilt unit 
can be used to scan the 3-D laser profile around a 360? pan angle and a 110? tilt angle. The scanning parameters as well 
as the image resolution are computer controlled and therefore fully programmable. Different modes of low resolution 
and high resolution images can be preprogrammed to completely cover the surroundings. A novel iris arrangement in 
the detector unit coupled with real-time sub-pixel (0.2 pixel) detection of the reflected light stripe supports real-time 
range determination with a relatively low noise level. The very short baseline employed in the sensor (150mm) enables 
a robust and compact construction highly suited to this application at the expense of a relatively low precision (table 1). 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 123