A 869Kb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE
We provide (mostly through our AirGon subsidiary) a lot of technology and services into the drone mapping market. Our focus is primarily on site types where drone mapping is a more economical alternative than manned airborne or traditional surveying. This range of application includes quarries (open pit mines), industrial sites and bare earth construction projects. At the moment, we use only dense image matching technology to create 3D models of these sites. I have discussed the strengths and weaknesses of dense image matching (DIM) in several Random Points columns but the net is that DIM is inexpensive (the plus) but cannot penetrate vegetation (the minus).
We sometimes are asked to do mapping of or provide technology for sites that have vegetation cover such as that of Figure 1. The magenta points are topo shots collected with an RTK rover. Note in the profile view that these shots appear far below (over 6 feet) the surface derived from the DIM model. In fact, due to miscorrelation by the DIM creation software, the points are distorted in the vertical by more than twice the height of the vegetation. This should dispel any notion of simply subtracting an average vegetation height from the DIM to arrive at an approximate model; you simply cannot use DIM modeling in these types of situations. We need a LIDAR!
I thought it would be worthwhile to discuss the attributes of a good drone-based LIDAR for small area mapping where the focus is on high vertical accuracy (at least as good as DIM) and low noise (e.g. high precision). This sets a pretty high bar because we routinely achieve better than 2 cm of vertical accuracy (relative to the base station) and less than 2 cm of noise using DIM in bare earth areas. I won’t talk about size, mass and power since these are givens for a drone system payload.
A gating criterion is cost. You may be the lucky service provider who has found the customer who will "pay what it takes" to get a LIDAR solution but that is the rare customer indeed! Most jobs have a limited mapping budget. If the new technology will not support that budget, then a lower cost method such as an RTK "pogo" survey will be deployed. The trouble with drone-based LIDAR technology right now is the rapid pace of change. You cannot purchase a system today and expect it to be competitive for very long. Thus you have to amortize the cost of the system over no more than perhaps 24 months. A second primary consideration with respect to cost is the system risk. You can expect a total system failure (e.g. a crash) at a rate of about 1%. I am more than a bit reluctant to put an expensive system into such a high risk environment. You might try to reduce risk by equipping your drone with a ballistic parachute but there are triggering problems with these; a subject for another article.
A second consideration is range. Low cost LIDAR systems suitable for drones today simply do not have the range needed for mapping sites with deep pits (most mine sites!). We often map sites with vertical ranges of 150m or more. If we add a reasonable flying height of say 30m to this, we need a system with a range of at least 180m. Low cost (under 15K for the sensor head) providers advertise ranges on the order of 100m but actual performance is more like 35m. The solution promoted by the LIDAR manufacturers is to use terrain adaptive flying. Obviously they have never flown a mine pit!
The third consideration is precision. I talked about this in the last issue of Random Points. Low cost LIDARs today have terrible precision (on the order of 5 to 10 cm). This is much worse than the precision that can be achieved over bare earth areas by dense image matching. While I would love to see precision better than 1 cm, I would settle for 2 cm.
Reliable multiple returns are mandatory. Without multiple returns, you cannot distinguish true ground points from those reflecting from overhead structure such as vegetation and low height equipment. The minimum is two returns but the preferred is four. If we are going to ask for a more perfect system, why not the ability to distinguish returns with a spacing as small as 10 cm?
The technology does exist today in the Riegl mini VUX-1UAV. This is without a doubt the most capable LIDAR scanner for drone deployment on the market today. However, at far north of 100K, it is cost prohibitive for all but the most specialized (meaning you can charge a lot of money for the service!) of drone-deployed applications. Scanners at the low end of the market fail most of the technical parameters outlined above but to be fair, these systems were designed for collision avoidance applications, not mapping.
It will be interesting to see this market evolve. Emerging from the shadows are focal plane array linear sensors. Will some unheard of emergent company bring an entirely new FPA LIDAR to market and start us down a new path of drone mapping? Of one thing we can be certain; this terrain will be considerably different one year from today!
Lewis Graham is the President and CTO of GeoCue Corporation. GeoCue is North America’s largest supplier of LIDAR production and workflow tools and consulting services for airborne and mobile laser scanning.
A 869Kb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE