Drone LiDAR or Photogrammetry Debate
Which is better? Of course, the answer is, “It depends.” A good drone service provider knows when to use each to provide the best product at the right price. So, let’s look at a few categories side-by-side.
What is accuracy?
For 95% of all applications out there, this is what clients and service providers care about most. This is one of the most misunderstood criterial in the market place. You can quantify accuracy by either global (absolute) or relative accuracy. Global accuracy is simply how accurate the x,y,z points are in relations to the known points in the coordinate system being used. Relative accuracy is how accurate the point cloud is in-reference to each other. So is the relative model representative of what it is in real-life? With these definitions in mind, you cannot have good global accuracy without first having good relative accuracy. However, you can have good relative accuracy (1-2cm) and terrible global accuracy. Survey grade global accuracy is 2-5cm.
Here, LiDAR has an advantage over photogrammetry. LiDAR shoots out light and measures the return. It always knows the point of origin, the time, and the new position at return. This makes it very easy to calculate each pulse, since all points in the equation are measured values.
Photogrammetry uses cameras that are dependent on ambient light. Again we know the position of the drone at the time of collection but we do not know from one image how far away it was at that moment in time. As a result this requires a lot of computer post processing and ground referencing. Photos must be compared for similar features on the ground. However, today’s software can get an accuracy close to LiDAR after hours of computer post processing
Both methods are heavily dependent on ground control points. LiDAR usually needs one GPS base station with defined accuracy down to the millimeter. This usually takes 30 minutes and can be done while the crew is setting up. During flight, we practice differential GPS where the onboard GPS compares its position to the base station to remove global accuracy error.
Photogrammetry, on the other hand, needs targets called Ground Control points (GCP). Usually you need a minimum of 3 to ensure you properly correct your relative solutions. But the larger the area, such as a powerline corridor, may require many more. Each position requires 30 minutes to determine an accurate solution. For this reason, 75% of the time of a job is setting up for photogrammetry.
Compared to conventional surveying methods, both methods (with the right software integration) can provide useable products several times faster than surveying, and with very high-density data quality. This means traditional surveys can only shoot so many discrete points and interpolate the difference in between. With both LiDAR and photogrammetry there is no interpolation. All features seen are captured.
The difference is LiDAR can be fed into a CAD or GIS program shortly after landing. Photometry must first be processed with very powerful computers that can handle the intense number of images–the more photographs, the longer it takes to process.
Here is where LiDAR has an advantage over Photogrammetry. LiDAR scanners with multiple return capability are measuring more than just a single point when a laser is reflected. For example, these multiple reflections can represent the top of the tree, the middle branches, and then the ground. Processing and modeling software allows the user to analyze certain areas alone, or filtered with other data groups. For example, filtering the point cloud data to remove everything except for the ground returns, results in a highly accurate bare earth model. This is called Echo Digitalization.
Photogrammetry can only create points based on what the sensor of the camera can detect illuminated by ambient light. Users should understand contract, exposure, and filters to ensure they collect as much ambient light data as possible. Service providers cannot do this at night. This problem is compounded by commonly used small imaging sensors with poor sensitivity. Thus, flights over vegetated areas, with limited ground visibility and deep shadows, produce few ground points and only treetops are accurately represented.
Color may be one area Photogrammetry has over LiDAR. While pictures today can be imposed over a point cloud, it does not look the same. LiDAR only collects one wavelength of light. This allows photogrammetry to product stunning 3D images.
Until recently it was hard for many firms to justify the acquisition cost for LiDAR. A drone, a 4K camera, and software is only $8,000. But LiDAR was large and mounted on aircraft and helicopters, making it a luxury for the quality of data it presents. However, many manufactures have miniaturized the size of their LiDAR payloads so they fit nicely on a drone, and competition has reduced the price to make this much more attractive. Last year, high-end companies were offering packages from $150,000 to $200,000. The price for a full system is now down to $60,000 to $90,000.
Speed is money. In the past a firm would have to schedule a flight and wait on the data to be delivered. Now, using drones, they can be flown at any time. Drones not only allow any time access to data, but they can fly into areas previously inaccessible to aircraft. One case study in Portland, Oregon describes the use of a LiDAR system in deep canyons over a dam. This was too dangerous for surveyors and conventional aircraft, but very accessible for a drone integrated with a compact LiDAR system.
In the LiDAR vs Photogrammetry debate, as far as UAV acquisition goes, LiDAR is hard to beat. The added benefit of vegetation penetration, speed to market, and high accuracy makes it a no-brainer when the price is on par with lots of photogrammetrists. We see professional service providers opting to adopt this technology increasingly in the next year.