How can you get the best from your Pointcloud data? In the article, ‘How is a Pointcloud made?‘ we discussed the need for the LiDAR and INS coordinate frames to be well calibrated, and how precise this needs to be. The obvious question then is how do we achieve this? There are many ways this can be done including trial and error, using a digitally printed mount or using a data-driven calibration technique. Trial and error is time-consuming and you must ensure that your setup does not change again to the slightest degree or you will have to repeat your trials. A printed mount is a convenient way to get precise angles and displacements built in to your setup. However, it is time-consuming and expensive. A third way is to use a data-driven calibration technique that analyses the Pointcloud data you have collected and calibrates the setup by comparing what is seen to what should be seen.
This latter method is what OxTS has developed. The user is required to simply manoeuvre their vehicle with their LiDAR and INS setup around, such that the LiDAR gets a good view of some targets from a range of distances and angles that are representative of your survey. The targets are very simply made of flat, square boards covered in highly reflective material. The manoeuvring takes only a few minutes and can be added onto any survey run. At the office the data can be uploaded into OxTS Georeferencer which will look for the angles that give the Pointcloud the best fit to reality. We have also developed a method for calibrating the displacements, and this feature is now available in beta mode in Georeferencer.
Georeferencing
OxTS Georeferencer can also, unsurprisingly, georeference LiDAR and navigation data to create Pointclouds. What is required are usually five files: The navigation data file, the LiDAR file (PCAP) and three configuration files that detail the angles and displacements between the LiDAR and INS, and also the INS to the vehicle. With these one can calibrate the configuration files as explained above or create a Pointcloud.
On top of this there are plenty of additional options the user has to get the most of their Pointcloud. Often time the LiDAR can be viewing parts of the vehicle. If this occurs then a part of the vehicle will be present in every frame that is recorded and in the Pointcloud it will appear as if something has been dragged around. This can simply be edited out by choosing a minimum distance away from the LiDAR that points must be. This allows users to be more flexible in their setup.
LiDAR usually record an intensity of reflection value for each point. This is one of the great features of LiDAR that give them so many unexpected uses. Users of OxTS Georeferencer are able to create Pointcloud at any reflectivity threshold. This is a great benefit particularly because creating Pointclouds at a reflectivity of 100 allows you to troubleshoot your setup while only being a few hundred kilobytes instead of the usual gigabytes of data.
Perhaps the most important feature of Georeferencer other than the calibration is that it will use a sophisticated formula to calculate the precisions that points can be known to using the diagnostics that our navigation data gives out. This allows the user to view which points are the most, or least, accurate and also to edit out these points. Users are able to choose the centimetre precision that all points in the Pointcloud must be under assuming the setup is calibrated so you can promise your client that you can deliver the right accuracy – with confidence.
