Point of Beginning

Decoding the Cloud

June 1, 2007
Figure 1. Above is a true color scan of 6” diameter pipes and valves.

As surveyors, we take great care in understanding sources of errors in measurements, and strive to find ways to decrease or eliminate the effects of errors. With each instrument we use, we need to have an understanding of the sources of errors that can come from it. Most people know the sources of errors when using GPS, total stations or levels. In this article we will look at the types of errors that can be found in a relatively new technology: laser scanning.

There are four common error sources that can appear in laser scanner data (point clouds): mixed pixels, detector saturation and blooming, multipath and incidence angle.1 These errors are systematic and appear as noise or as bad measurements. Often, these error sources can be detected when you know what your subject point cloud is supposed to look like and you can weed out the bad information. It is beneficial to know what types of errors to be aware of when it is time to process your data.

Figure 1, Part 2. Above is a cross section of two pipes showing the mixed pixels at the edges of the pipes.

Mixed Pixels

Mixed pixel errors occur when you are scanning a surface that has another surface closely behind it. When the edges get scanned, the laser beam gets split and part of it hits the front surface and part of it hits the rear surface. The energy that is bounced back to the scanner comes from two different distances; the scanner interprets them together and the result is a point in space that is not on either surface. This error is more pronounced with systems that have wide laser spot sizes. Some software makers have routines that will trim the edges of point clouds, which helps to remove points that trail away from the edge. If you are aware of this error, you can create more accurate models by ignoring the mixed pixels when fitting an object to a point cloud. Figure 1 and figure 1, part 2 show a group of pipes that were scanned and the cross section of two of them that have points trailing off from the edge of the pipe.

Figure 2. A scan is of a 1” square reflective target.

Detector Saturation and Blooming

Detector saturation occurs when something is scanned that is highly reflective and the returned energy exceeds the dynamic range of the detector.² A range error occurs and the data points either come nearer or farther to the scanner than where they should be. I have seen this happen especially when reflective targets for total stations get scanned. There is a stream of points that come out at the scanner as seen in Figure 2 where a 1" reflective target was scanned. A simplified way of picturing how this happens is to imagine the laser bouncing back really fast to its source, which results in a shorter range measurement when a time-of-flight scanner is used. Another instance where this often occurs is on reflective lane lines on roads. The lines can sometimes slightly rise off the road surface. Sometimes there are surface reflectivity issues that cause an increase in the range measurement. I experienced this when I was scanning a 1-ft square white plate that had an orange cross painted on it. The orange appeared to be inside the plate, even though it was really on the surface (see Figure 3 and Figure 3, Part 2).

Blooming is similar to detector saturation. This error happens when very reflective objects are scanned and energy from the surrounding area of the laser beam is returned. Often this is seen when a retro-reflective target is scanned. This means that while prisms are ideal for use with a total station, they do not work well with a laser scanner.

Although there is no way to prevent detector saturation or blooming, surveyors should avoid using retro-reflectors and be aware of these errors when processing the data.

Figure 2, Part 2. A cross section view of the 1” target shows points streaming out of the face of the target.


Surveyors usually associate multipath errors with GPS measurements. They make an effort to reduce the effect of multipath by choosing the proper antennas and avoiding setups with lots of multipath reflectors. Similar range errors happen in laser scanning when the laser bounces off highly reflective surfaces and creates backscatter as it is reflected off more than one surface. This range error can be up to the maximum range of the laser scanner and is more easily detected when the error is large. The most common occurrence of this error is when a mirror on a wall is scanned. There will be points on the other side of the wall that will be a corresponding mirror image of whatever is in the room. Figure 4 shows an example of a large mirror on a wall that created extra data that had to be deleted before the data could be processed. If possible, cover mirrors before scanning, or apply a coating to a highly reflective surface to minimize multipath errors.

Figure 3. A scan shows a 1’ square white plate (the green points) with an orange cross painted on it (the blue points).

Incidence Angle

Incidence angle errors are familiar to many who use a reflectorless total station. Range errors due to non-normal incidence errors need to be considered, especially when performing high accuracy work. One of the main factors contributing to this error source is beam width. The narrower the laser beam, the smaller the effect will be from a non-normal incidence angle. One way to visualize this is to picture a flashlight being shone on a wall at an angle. The spot of the light will be an ellipse and not a circle. Different scanners interpret the return from an ellipse in different ways; you can calculate the range error at a given range if you know the beam width and incidence angle.3 For example, if you had a 3 mrad (milliradians) beam width at 100 m with an incidence angle of 45o, the range error would be 0.15 m (see reference 4 for the equation). To avoid having incidence errors, it is best to scan perpendicularly to surfaces. If you are scanning a road, that might mean using a tall tripod.

Figure 3, Part 2. A cross section view of the 1’ plate shows how the points did not fall on the same plane because of the color difference.

Aiming for Accuracy

With any measuring device, an understanding of the systematic errors will ensure accurate data. Along with new technology come new sources of these errors. Some of the sources mentioned can be avoided with proper field techniques; those that cannot be avoided have to be addressed when it is time to process the data. Although these errors are mostly unavoidable, knowing about them is the best way to deal with them when the data is processed.

Figure 4. This point cloud shows multipath errors from scanning a mirror.