Before we begin, I must once again remind you that my companies GeoCue Corporation and QCoherent Software provide consulting services and software to the LIDAR industry. In addition, GeoCue is the North American distributor for Terrasolid. This does, of course, introduce some level of bias into my articles.
In this next series of articles, I will move through some best practices for LIDAR data planning, collection and processing. While many of the techniques apply to both airborne and mobile, I will (at the request of the LIDAR News Editor) focus on airborne. The article is also very North American-centric with respect to positioning technology but it is easy to localize these processes (e.g. replace NGS with OSGB).
Through my GeoCue work, I have the privilege of working with many North American LIDAR acquisition and production companies (airborne and mobile) at a fairly intimate level. This provides a good window in to what works well, what needs improvement and things that are really bad! Everyone, without exception, is very interested in creating very high quality products at a reasonable profit. This, of course, leads to a big interest in best practices. The question is, of all of the promoted best practices, what works and what does not?
Lets assume you are involved with all aspects of LIDAR data acquisition, geometric correction/validation and data processing. If this is not the case for you, simple skip the areas that are outside your responsibility. The most important thing, by far, is that you develop a written process that you rigorously follow. The process should be cast in clay, not in stone. As you gain experience, evolve the process. Equally critical to creating this written process is to ensure that all team members rigidly adhere to the process. You would not believe how many times I hear every LIDAR job is different! Yes, it certainly is if you are not developing and following processes!
Plan, plan, plan!! It is often terribly expensive to discover data defects that require reflights. Which is more expensive; adding an additional 5% overlap or risking a reflight? 3D mission planning is critical for areas of undulating terrain since one of the specifications you will be required to meet is data density. In general, it is always better to over-collect and then, if necessary, thin the data; storage is cheap. It is indeed the rare job that will not require GNSS base station(s). In additional, you absolutely must always have sufficient control in a project to do not only accuracy testing but also geometric correction.
You should do an OPUS (NOAAs Online Positions User Service) solution of your base station(s) before remobilizing since an unrecoverable error here will mean a reflight (or redrive for mobile mapping). You can do this in the field or FTP base station data back to your processing office (we prefer the back-office approach since this tends to be a much more controllable environment plus it prevents home office-field battles down the road). For high precision/accuracy projects (e.g. geomatics level mobile mapping) you will be forced to use Horizontal Time Dependent Positioning (HTDP) and this type of correction must be performed via a very rigorous, documented process.
You will definitely want to do field checking of both trajectory and base LIDAR data each day in the field. These checks include ensuring you can converge on a stable forward-backward trajectory solution, a LIDAR data coverage check (any holes in the data?) and gross geometry checks (for example, Multi-Pulse in the Air systems are notorious for finding phantom surfaces hundreds of meters in the air!).
I very strongly recommend that you not use LIDAR hardware vendor-supplied algorithms for transforming coordinate reference systems (CRS) or performing geoid/tidal shifts. Do all these operations in one central location (such as your GeoCue workflow system) and document/test the process. We have noted that some vendors are using open source transformation code that was designed for GIS rather than geomatics processing. There is little more embarrassing than delivering data with a vertical shift due to incorrect CRS application so make sure you are applying at least two completely independent tests.
Unfortunately, there is a ton of misinformation out there regarding system calibration versus project geometric correction. Folks initially involved in photogrammetry understand this distinction but it is often not conveyed to LIDAR-only shops. Ideally, your systems require calibration on a periodic basis whereas geometric correction is generally required on every job. It is very important to perform these processes separately. Calibration deals with sensor parameters such as mounting angles and lever arms. For a well engineered system, calibration should hold over fairly long time periods.
Project geometric correction, on the other hand, is primarily correcting for inaccuracies in the locations originating from the Position and Orientation System (POS); these vary from sortie to sortie and indeed, even within sorties. Some of the POS issues may be due to the quality of the GPS/IMU hardware that you have purchased (this is not the place to try to save money!) whereas the remainder are due to the variable nature of determining kinematic position (satellite configurations, bank angles, urban/canopy canyons and so forth).
You may be able to get away with performing calibration using heuristic methods (e.g. visually inspecting differences in flight lines) but such techniques are extremely primitive (see Figure 1 where the overlapping profiles of two flight lines are displayed). A much more scientific (and repeatable) approach is to apply a rigorous mathematical solution such as that provided by Terrasolids TerraMatch system.
Figure 1: Heuristic Geometric Correction
In part two, I will continue the discussion of geometric correction and move into LIDAR project setup.