“When I tell people what I do, when I tell them I’m an engineer, nobody really cares,” quips Ted Knaak, president of Certainty3D. “But when you tell them you do mobile LiDAR, you take 3D images down the road at 50 miles per hour and you can extract all this CAD work, even your wife’s impressed sometimes.”
Speaking to his company’s TopoDot Users Conference, Knaak wasn’t just going for a laugh. In some ways, the gaming world and Hollywood movies have lent an air of excitement to at least a portion of what land surveyors and geospatial professionals do every day. With 3D imaging and virtual reality becoming buzz words, flashing a point cloud on a computer screen can get some attention. If it is married to photo images and has a look of reality, they’re riveted. Getting there isn’t easy or cheap, and it may not deliver what a client needs. But, done right, and done in conjunction with other technologies, mobile LiDAR is powerful.
“This is the next wave in technology,” Knaak continues. “Someone could say, well, it’s been around for a long time, but we’re talking about it becoming mainstream. It’s mainstream now.” Referring to his work with the Florida Department of Transportation, Knaak notes, “You can’t get a continuing land survey contract without mobile mapping capability. That’s not like playing around with it once in a while. You have to have it.”
Harnessing the Power of Data
We can now acquire survey data at highway speeds, Knaak says. It’s good data, and there’s a methodology that’s faster, better, cheaper that will get it all the way down to survey quality, he points out. “The bottom line is, that being the case, that changes everything. What happens is, now the DOTs realize a lot of the problem is how do you use it?”
A mobile mapping job is typically done for a particular purpose. The data collected for one project is used for that project, but then what? A DOT might be planning 700 miles of construction in a year, but it has another 5,000 miles to cover for GIS. “The data kind of hangs there,” Knaak says. While one project is gathering survey-grade data on a stretch of road, there are also requirements for GIS data collection, which has a much lower accuracy requirement. The DOT wants to be able to use this data for multiple purposes, but the GIS data may be at accuracies of plus or minus a foot. On a construction project, the surveyor will say, “We can’t use any of this data.” Knaak describes the surveyor’s reaction saying, “One road’s here, the next level’s here. It’s all over the place. The pictures don’t line up because their immediate spec is plus/minus a foot, plus/minus 3 feet, or plus/minus 6 feet.” Based on the specification in the bid, that’s how they collected the data.
“Accuracy costs money,” Knaak comments. “So, the question is, ‘What’s the right balance between spending a little more on a lot of data and preparing it so you can improve it later?’”
He points out, “You don’t want to spend a lot of money doing the higher accuracy for 5,000 miles. You’ll just kill yourself because you need control, you need all kinds of stuff. You’re not going to do it. However, you can spend a little more, just in time, you don’t need control out there, which is real expensive, if the guys do relative accuracy. In other words, do it correctly where the relative accuracy is right. And if they can do the relative accuracy right, they typically have the capability of later, with control, taking that and making it tighter so you can actually go back and pick and choose what you need. And that’s the key. How do you set up and organize and how do you make that happen?”
For Knaak, one of the important points is what is done up front in the quality assurance and quality control (QA/QC) areas. “All these engineers can only request and say I need data for this. It has to go through someone who understands the provenance of that data, where it came from. Will it meet those objectives? Is it in the right accuracy? Can we get it to the right accuracy if it isn’t? And, by the way, how’s the best way to extract what this person needs?”
More Sensors: Even More Data
At this point, Knaak takes a step back. “The other thing that’s happening is, they want everything on the truck.” He’s talking sensors. In addition to the mobile LiDAR, he describes a virtual rolling platform for data collection technology. “If you’re going to do 5,000 miles, you might as well get it all. All the guys that are doing GIS need mobile LiDAR because it’s the only thing that can get them spatially tight enough and be adjusted because it’s the only thing that sees outside the lane they’re in.”
He explains that the typical GIS setup only sees one lane. “You’re never going to put control in one lane and move it anywhere. So, with that data, it’s looking like this lane is here and the next lane is above and higher, and you can’t line it up because you don’t have overlapping data. You’re just looking at that lane.”
Trajectory is critical to tying the different data collection tools and results together. The trajectory is like a missile trajectory, Knaak says, it guarantees where everything is and all the sensors are tied to that. “The question is, how tightly do you tie the sensors to this? Every sensor needs its own traceability lineage back to this trajectory and then, inherently, back to the control.”
Some sensors, like subsurface radar, won’t be as tight. Those sensors are just looking for holes and cavities. Elevation isn’t important and accuracy may be OK within 6 inches.
Cameras are a different matter, and cameras have to be tied extremely tightly to the trajectory and back to the point cloud. “You’re never going to get there if you didn’t do it right up front,” Knaak cautions. “So you drive here, you drive here, all that data fits together. They will have the ability, when you give them control later when they need it, they can take those trajectories and move it and reprocess all the data and put it where it should be. That’s the way they do it now. They’ll go through without control first. And then they put the control points later. And then they readjust everything.”
Control is an expensive thing, he repeats, but “the symmetry between imagery and point clouds is really very important. It lessens the requirements of your point cloud and it gives you much more information. But you have to be tightly calibrated to use it right. If you just color your point cloud, basically you’re throwing away all your pixels.”
Knaak takes the discussion out of the realm of computer science saying, “There’s a lot of signal processing. That’s really where it is, not computer science.”
Defining the right specifications up front is the job of the client. But, responding to the request for proposal and delivering on those specifications belongs to the geospatial professional. There’s a joint effort in quality control and quality assurance and in organizing how data will be used after it is collected. “It all comes back to control, the control survey is the only thing that’s a legal document. And that’s the only thing you’ve got to tie it back to,” Knaak says. He concludes, “Where mobile LiDAR is efficient, makes everything more efficient.”
More Applications for Mobile LiDAR
It’s no secret that mobile LiDAR is an important tool for infrastructure development and for mapping that infrastructure. The accuracy of both is coming even more into focus as the transportation world looks towards a future of autonomous vehicles.
While those who build and maintain infrastructure are looking at their own needs for very precise geospatial data, infrastructure users will depend even more on the precision of that data . . . or their vehicles will. The same technology that collects data for mapping and maintaining that infrastructure will be employed to navigate it. As POB’s sister publication, Autonomous Vehicle Technology, observed, autonomous vehicles will depend on mobile LiDAR, “because no matter how smart a self-driving car is, its performance is only as good as its knowledge of the world around it.”
Glen DeVos, chief technology officer of automotive supplier Delphi, told AVT, “It is LiDAR that is seen as crucial to creating the high-resolution 3D maps of road routes that will allow autonomous cars to locate themselves on the streetscape, a fundamental requirement. The HD (high-definition) maps would function something like your memory, so that the control processor has prior knowledge of every upcoming hill, dip, curve and valley. LiDAR is well-suited to this “big picture” imaging, and in its current state can provide the needed 360-degree view.
Mark Romano, senior product manager for Harris Corp., took it a step further for attendees at the Certainty3D TopoDot Users Conference. Romano observed, Princeton Lightwave is already developing a close-range system [Geiger-mode LiDAR] camera for the automotive industry. “They’re testing it out to 300-400 meters right now. Their intent is to build these down to be small enough to be surface-mount integrated into the body of a car. So, today we’re using radar collision avoidance. Now they can see people and physical objects out in front of them.”
Another company, AGERpoint, is in research mode, Romano says. What AGERpoint has developed “is not a camera, it is a single-point, high-speed scanner that’s photon sensitive.”
Romano continues, “The thing with these devices is they are going to scale. This is not a very big LiDAR. If you surface mount this with a small ranging laser, it’s going to be very tiny, very light weight and eventually a surface mount as opposed to today we’re putting big scanners on the tops of cars.
“Because this is now commercially available, there’s a ton of research and a ton of money being thrown at it by the big automakers and people like that. They want to see this built into our driving platform,” Romano says.
The prospects for LiDAR, and specifically photon-sensitive Geiger mode LiDAR, are significant in part because of the developments in autonomous vehicles. But Romano sticks close to the technology’s roots. “[Harris] doesn’t own this technology. What we own is a lot of experience and intellectual property and how to make a point cloud out of it. It’s different. It’s signal processing, not measuring a range and applying a point where I get a range and a range and another range, I have three points in space, I’m done, correct measurement. This is a big ball of fuzz, photon-sensitive fuzz. When you look at it you’ll see just this big stack of a cloud of points and somewhere inside that is the signal, which is this part that we have to get out of that. And that took years of evolution of developing a whole different type of set of algorithms to be able to pull the signal out of it. It’s kind of like waveform processing that’s basically signal processing. That’s what we’re doing here.”
It’s difficult to conclude that LiDAR will follow the path of satellite positioning and be integrated into every phone in nearly every pocket in the world, but it could certainly take its role in the developmental stages of autonomous vehicles and become a key component in most vehicles in the future. While we wait for that day, the R&D efforts to improve and miniaturize the technology and improve its performance and interface are bound to benefit the survey community in the meantime.