Terrestrial Laser Scanning

From HydraWiki
Jump to: navigation, search

Overview[edit]

These instructions provide a generic overview of the procedure of terrestrial laser scanning within a laboratory environment, in order to obtain digital elevation models which can be used to map the evolution of the experimental morphology. The exact process will differ for different scanner manufacturers.


Instrument Set-Up and Scanning Methodology[edit]

This section provides a recommended procedure for carrying out terrestrial laser scanning, there may be some variation depending on the exact equipment used, but the overall procedure will remain the same.

1. Fixed targets should be located around the whole of the experimental plot, ideally at varying heights. Whilst there is no limit to the number of targets required, it is recommended that a minimum of 4 targets are present within each scan envelope to allow for accurate registration of the scans. The more targets present, the higher the spatial location accuracy of the subsequent point clouds. There are two standard types of target that are used in the laboratory, spheres and checkerboards (Figure 1 ).

Figure 1: Left - Fixed spherical target. Right - Fixed checkerboard target

2. Once the experiments have been set up with initial conditions, the scanner is mounted. It is important to ensure that the scanner is completely level when mounted, and this can be done in two ways. The most common is to mount the scanner on a tripod (Figure 2), which can be adjusted to ensure it is completely level. The alternative is to use fixed mounting points around the experimental plot. If it is easily accessible, the ceiling is a good location as this will give an overview of the entire plot.

Figure 2: Terrestrial Laser Scanner mounted on a standard tripod, adjusted to ensure scanner is level.

When locating the scanner it is important to ensure it has a clear line of sight to at least 4 of the targets.

3. It is now time to begin the initial laser scans. Switch on the scanner and ensure that the settings on the scanner are appropriate for your requirements (i.e. scan area, point cloud density etc.). Depending on the scanner used, check the memory card is in place. Then start scanning, and move to a safe distance outside of the scan area. Each scan file will record millions of data records that include position, reflectance, and colour for single scan points.

4. Once the scan has been completed, then the scanner can be moved to another location. By carrying out multiple scans of the same experimental plot, issues such as laser shadowing or beam divergence will be mitigated.

5. The experiment can now be run. Whilst this is happening, the point cloud data can be downloaded from the memory card and stored in an appropriate location. Upon completion of the experiment, any water present should be drained from the plot, and the same process is repeated to obtain point clouds of the evolved experimental surfaces.

Post-Processing[edit]

Whilst most brands of scanner will have a proprietary software for processing the scan data, there are also a number of open-source programs such as CloudCompare [1] available. The below procedure is based on the use of a Faro Focus3D scanner and its corresponding proprietary software, Faro Scene, however it is anticipated that the procedure will be similar for different software programs.

1. For each iteration of the experiment, create a project within the software. This will produce all the necessary files and structures at a specified location for the post-processing to occur.

2. The raw scans for each iteration need to be imported into the project as a single cluster. This will ensure that when the scans are registered, all the information to build a single project point cloud is available. It is also useful here to import the exact locations of your targets. A table should be produced with each target assigned a name and co-ordinates in all three dimensions, x, y and z. This can be imported as a .txt file.

3.The scans are then processed. There are a number of options that can be included here:

a. Colourise Scans. If you scanner has also taken photos whilst scanning the surface, these can be applied to the point cloud to produce an accurate depiction of the experiment.

b. Filters. There are likely to be a number of filters which can help filter out erroneous data points at this early stage. For example, in experimental work the distance filter can be applied to remove points from outside of the experimental area.

c. Find Targets. Finding targets is important as this will help with the registration stage of the post-processing. Selecting Find Checkerboards and Find Spheres will identify the targets that you placed around the experiment during scanning. It is recommended not to rely on Find Planes, as this can erroneously identify the surface of the experiment as a plane which causes issues when comparing between iterations of scans.

4. Once the scans have been processed, the registration can take place. Although there are various methods for registration, based on previous experience in scanning of physical modelling it is recommended that the best method is target based. This will use the target co-ordinates you have inputted, and the targets identified by the software and register the scans in relation to them. It is important to verify targets, as the software can erroneously identify targets that were not present (Figure 3).

Figure 3: Scan identifying spherical target on ceiling of flume laboratory.

The registration will produce a single amalgamated point cloud of the entire scan area (Figure 4).

Figure 4: Single point cloud produced by combining four separate scans.


5. Once the registration has been completed, a registration report should be produced (Figure 5). This will give information about the error within the scans. In small scale experiments, it is particularly important that these error values are kept to a minimum.

Figure 5: Example of a registration report detailing error within the final point cloud.

It is also now possible to remove points outside of the area of interest, for example anything outside of the experimental plot (Figure 6).

Figure 6: Final registered point clouds showing only areas of interest.


6. This process should be repeated for each iteration of scans. This will produce a project point cloud which can be exported as a generic format for analysis in non-proprietary programs. By using the target locations for the scan registration, this will make comparison between different clouds easy. It is also possible to use the point clouds to produce data including polygonal meshes, cross-sections and high precision measurements, and multiple point clouds can be compared to get information about accretion and erosion within an experiment (Figure 7).

Figure 7: Digital Elevation Model of Difference (DOD) showing accretion (Red) and erosion (Blue) of sand at different times in an experimental regime.

Limitations[edit]

Whilst laser scanning can be used to produce detailed information about the evolution of a morphologically changing surface, it is also important to be aware of some of the limitations, and whether they are applicable to the particular scenario you are interested in.

In physical modelling there are three main issues that should be mitigated. Firstly, the colour or reflective index of the surface. If the surface is black or high reflective, this will affect the accuracy of the scan. Black will absorb the laser energy and reflective surfaces may scatter the laser causing the data returned to the scanner to be noisy. Reflective surfaces are likely to be an issue if any sediment is saturated during the scanning process, as the presence of water is likely to produce a reflection. This is why the experimental plot must be drained of water prior to scanning.

A second issue is beam divergence, which is caused by the expansion of the diameter of laser beam as it moved away from the scanner. This means that the greater distance from the scanner, the less reliable and accurate the data will be. Due to the scaled nature of physical modelling this is unlikely to be a major concern, except if very large scales experiments are taking place.

The final issue is laser shadowing. This is caused when the laser beam, acting as a light source causes objects to shadow one another. This can be quite a problem within physical modelling, as often other measurement equipment can be located around the experimental plot causing shadowing. Whilst this is not completely unavoidable, the issue can be minimised by carrying out scans from various point around an experimental plot.