Terrestrial laser scanners are very capable measuring instruments. They could well be the cause of the next revolution in surveying. But like a lot of recent technology, they are black-boxes, in as much as all their important functions are inaccessible to the user.
Black-boxes are dangerous for two reasons: on the one hand, they are easy to use, and on the other, because of their complexity, it is difficult for users to know either the accuracy or precision of the resulting measurements.
The argument offered in this paper is that laser scanners have more components which are subject to error than most surveying equipment, and therefore will require extensive calibration and checking.
Surveyors can, and should, turn the threat posed by non-surveyors - who will attempt to use these instruments - into an opportunity to secure an indispensable role in the use of laser scanners.
As a corollary of this argument, surveyors must make sure that they have the skills to estimate the effects of errors on any results derived from the measurements.
Although the commercially available scanners differ from one another in detail, they all have one thing in common. They obtain the shapes of objects by measuring vertical and horizontal angles and distances to many points at a very close spacing. The angles are measured by deflecting a laser beam with a mirror, which rotates about vertical and horizontal axes simultaneously. Distances of up to a few hundred metres can be measured without need of a reflector, with accuracies (depending on the particular instrument) down to millimetres. And they do it fast - at speeds of around 5000 points per second.
The Hoover Dam in the US as seen through the I-SiTE laser scanner: - Photo supplied by I-SiTE
Computer processing typically follows the scan session. This provides suitable results and output from the downloaded measurements. Thus scanners can provide all the essential measurements to describe an object as seen from one site. It is easy to come to the conclusion that the laser scanner provides almost everything the surveyor needs when measuring up appropriate object types.
Surveyors must make sure that they have skills to estimate the effects of errors ...
Because they are so capable, the instruments are quite complex. Current prices - of the order of $250,000 - reflect this. The operator's role is generally to do no more than set the instrument over a point, level it, orient it, input various settings and data requirements, activate it, and then download data for further processing (via the software) that will provide the sort of output the user wants. Presumably, laser scanners will get even better - more sophisticated, cheaper and smaller - and even easier for non-surveyors to use.
It can be argued that once laser scanners have the capability to recognise survey targets - perhaps with a little bit of prompting of the software by the operator, especially if used in conjunction with digital cameras - they could even replace total stations. Clearly, this would make measurement even easier for the non-surveyor.
The problem with any black-box is that it is fine until something goes wrong. This is true of domestic video machines and computer-controlled car engines as much as of electronic surveying equipment.
But surveyors are not used to this. In the case of surveying, having 'something going wrong' does not just mean breaking down; it might mean giving an incorrect survey result to a client, another surveyor, a supervisor, or a regulatory authority. Surveyors clearly need to be able to guarantee the measurements they generate; they need some way of controlling the errors even from a black-box.
The procedures involved in calibrating a laser scanner are quite complex, not obvious, and have not been widely investigated or publicised. They will not be easy for non-surveyors.
There are several possible sources of error in a scanner. The vertical rotation axis may not, in fact, be vertical. Presumably some idea of the error can be obtained by rotating the instrument about its vertical axis, if possible, and if we also note the bubble sensitivity. The horizontal rotation axis may not be horizontal. This is equivalent to trunnion axis tilt in a total station or theodolite.
Another source of error is the measurement of either the horizontal or vertical deflecting mirror. There may be an error in distance measure caused by the reflecting object or by the instrument itself. We assume that there is no source of cyclic distance error in the instrument because it times a pulse, rather than using phase measurement of a modulated wave.
Yet another possible error source is the alignment of the scanner to the ground co-ordinate system, or indeed, its position in the ground co-ordinate system.
The testing procedure to uncover errors will not be simple. It will be based mostly on having some recognisable targets (in known positions) among the points being measured. Presumably, having no background - making sure that all points other than the targets we place are beyond the range of the scanner, for example - will make analysis easier.
While checking of some photogrammetric systems can sometimes be carried out using objects of a regular shape, the problem with laser scanners is that they are suited to measuring very large objects, and it is very rare for such objects to be built precisely to standard geometric shapes.
We can assume that we need to carry out checking procedures (simply looking for errors) as well as calibration (determining the error due to a certain cause).
We should also recognise that the accuracy of the system is not the same as its precision. Two very close points on an object, at the same distance from the instrument, might be measured with an accuracy of 5 mm, but the two results could be the same, i.e. they have the same inaccuracy, but very good precision.
So the danger is that we could measure the shape of a wall that appears perfectly flat, and assume the result was accurate - when what it really is is precise. To determine measurement precision it will be necessary to take repeat measurements, but this may require repeated set-ups and the use of a number of targets in assorted positions in order to isolate all the different sources of error listed above. Positioning error will cause errors in the entire set of XYZ output, but this will be apparent only if all the targets have known co-ordinates.
There are four sources of angular error in a scanner, as shown here. To these need to be added the error in the distance measurement itself, which may depend on the target as much as on the scanner itself and on the precision with which the scanner itself is positioned: - Photo supplied by I-SiTE
However, we may need to be more concerned about software errors than hardware errors; it is harder to check the quantities produced by software. The merging of overlapping datasets, point triangulating, contouring, and finally volume calculation, can only be assessed by check surveys.
A single check survey will provide some evidence, but it will not provide statistically acceptable proof, so presumably numerous check surveys will be needed; this will make the software checking procedure very laborious and expensive.
There is a danger that many software functions (such as the ability to identify some geometrically shaped objects, such as pipes) will go unchecked. But regardless of the difficulties in carrying out this sort of checking, the possibility of errors in software means that surveyors will have to do something about checking for them.
My conclusion is that surveyors must know how to check both their scanner's measurements and the calculations derived from its software. But I have also argued here that surveyors should also be able to explain how any measurement or software errors affect their answers/results.
For example, it is not only necessary to know the accuracy and precision of measures of distance but also how errors in those areas affect any co-ordinates, areas or volumes which are calculated from the distances. This is, by the way, especially complex if the distances have been filtered or adjusted in other ways by the software.
I am not implying that there is any reason to doubt laser scanners' output. They are phenomenally accurate. Nor am I suggesting that surveyors should not use them; they enhance productivity like no other instrument. But they do not diminish the responsibility of surveyors to be able to defend their results.
These things are especially important because there are more things to be checked and calibrated in a scanner than in most other surveying instruments.
The results from scanners can be checked by other means, and surveyors should be able to do so. In fact, their ability to do so is one of the ways surveyors can differentiate themselves from non-surveyors.
Harvey Mitchell harvey.mitchell@newcastle.edu.au is in the School of Engineering at the University of Newcastle.
