The Drone LiDAR world is exploding right now and everyone wants one. Does anyone know what they are doing? Is Drone LIDAR any more accurate then manned LiDAR? What contributes to the accuracy and why isn’t it better? What’s all that noise in the Drone LiDAR? How much control is needed? When should Drone LiDAR be used? How does Drone LiDAR compare to Drone Image auto-correlation? Drones are clearly the latest shiny object in the geospatial profession and to make it shine brighter, we decided to add LiDAR to it. The problem is that it is imperative to understand LiDAR regardless of platform to be able to make the object shine correctly.
A 2.477Mb PDF of this article as it appeared in the magazine is available HERE
It is clear that LiDAR works on a drone but there are limitations to the technology. How do we properly operate LiDAR on a drone? The change in Drone LiDAR technology is advancing at light speed, this is in part a result of the manned LiDAR evolution. In the process of moving from the manned LiDAR world to the drone LiDAR world some realizations have come to light that not all drone LiDAR operators know what they are doing but as always there are companies that operate drones with LiDAR on them and companies that operate LiDAR on Drones. It is imperative for a company to understand LiDAR and LiDAR sensors because there are a lot of people that can operate drones properly. A lot of the questions on operation of Drone LiDAR are being answered as we conduct successful and unsuccessful Drone LiDAR projects. The best company to have conduct your Drone LIDAR project is a company that operates LiDAR on Drones. Why you ask? The understanding of LiDAR and the limitation of the components integrated into the Drone LiDAR is paramount in successful data capture.
The positional (POS) component is a much less expensive version and uses lesser technology than that of the manned POS systems currently being used today. The POS system is the GPS and IMU part of the LiDAR. The POS data is only as accurate as the technology allows it to be and there is a clear correlation between cost and accuracy in most cases. This doesn’t mean that the data will be bad, just less accurate. The lasers on most drone LiDAR systems are less experience and less accurate as well. This begs the question, well if the Drone flies much lower to the ground than how accurate do I need these components to be? The obvious answer is that given the flying height, the components can be less expensive and less accurate. Also, the functionality of the laser will play into the equation of accuracy and the resulting data. An example of this is the Velodyne Puke is a very popular laser currently being used on Drone LiDARs. The Velodyne Puck was designed for autonomous vehicle mapping and has 16 lasers on it, operating in a 360 degree field of view (FOV) with 2 returns. This is an excellent inexpensive laser but it has its limitation and this becomes apparent in the resulting LiDAR data. The most concerning element of the Puke is the resulting intensity information from this laser. The Puke really wasn’t designed to output intensity information the way the mapping profession currently wants it to be. So be ready to be a little disappointed as it relates to what you traditionally see.
When drone and sensor technology first started to become available there really wasn’t drone LiDAR yet but everyone was quick to develop the technology. A lot of the providers offered auto-correlated points form imagery as an alternative and the accuracy of this information ranged depending on process at which the missions where flown and how the data was processed. Most of the provides limited the expected accuracies to between 11 and 15 centimeters. In ideal situations, the data was more accurate than that but the limitation to auto-correlation is the same as it always has been and the ability to get accurate information is a function of the amount of vegetation in the area of interest (AOI). The more vegetation the less likely the success of the digital elevation model accurately representing ground. Basically, if there is vegetation in the AOI, then there is no likelihood that the ground will be represented correctly, if not at all. Additionally, The Drone LiDAR with a 2-return sensor will have a difficult time getting to the ground in heavy vegetated areas but it does get to the ground depending on the vegetation characteristics. This is similar to the manned LiDAR sensors. The advantages of the Drone LiDAR is that the number of points collected per meter is significantly more so the likelihood of getting to the ground is much higher based on point density. The Drone LiDAR typically get between 100 and 200 points per meter (PPM) but it can be higher depending on repetition rate, number of returns, and flying height. The new Riegl Mini-VUX has the ability to collect up to 5 returns which is a step in the right direction to increasing the likelihood of getting to the ground in vegetated areas. There are going to be situations where it is very unlikely that any LIDAR will get to the ground just like the manned LIDARs but your vegetation definition will look excellent.
The noise in Drone LiDAR appears to currently be much greater than that of manned LiDAR depending on the type of manned lidar being used. Noise in this case is defined as point return repeatability. If the LIDAR was capable of pulsing the same location all the time the measurement of that point return for that location will vary to a degree. Some manned LiDAR sensors have much more noise than others. This is the same for the Drone LiDAR sensors. Like manned LiDAR each Drone LiDAR regardless of make and model will have different characteristic just like the exact same lasers will have different intensity aesthetics. Typically, the noise associated can range from between 10 to 15cm on some systems down to 5cm on other systems. The key to dealing with the noise is how the data is processed to remove the noise. When flying Drone LIDAR and getting between 100 to 200 ppm or more is it really necessary to have that many points to define the feature in that square meter. In most cases the answer in no it is not necessary so the processing can take out the noise based on the best way to define those features and put them in a different class.
The fact that most current Drone LIDARs have more noise than we would like facilitates the importance of ground control. The algorithms used to define the noise and detect what is ground and noise need to be checked to make sure they are working right and that ground was actually what was found. Additionally, like most metric mapping projects, clients like to know their data is accurate to the specifications agreed on. Additionally, because of the limitations of the less expensive POS systems used it is nice to have horizontal information to assert the horizontal accuracy of the Drone LiDAR. There are no specifications on what is required for control currently and it really depends on the project characteristics and size but having control increases the level of confidence in the technology significantly.
Drone LIDAR is typically less expensive then manned LiDAR in most cases based on the relative sizes of Drone LiDAR project as they relate to manned LiDAR projects. Typically, Drone LIDAR provides consider a 3-square mile project as a large project whereas the manned folks consider that to be an extremely small project. It is important to remember that Drone LIDAR like many other LIDAR technologies is another tool in the Geonerd tool box and typically doesn’t replace any other tool. The Yellowscan Surveyor typically flies between 40 and 60 meters above ground and other similar Drone LIDARs fly at that less attitudes. The Reigl MiniVUX can fly at this attitude but also can fly up to 150m AGL but you probably wouldn’t fly it at that attitude. It seems that reasonable attitude for this sensor is between 90 and 100 meters. Given the following Flight attitudes it could expected at 40meters that a square mile would take about a day to fly at 3 m/sec. This seems to be the best flight configuration for the Yellowscan Surveyor. The same area would take about half that with a Reigl Mini-VUX. Typically, these systems could roughly fly up to 15 linear miles in one day. It should be noted that exact numbers would be best given by the LIDAR provider chosen.
Drone LIDAR technology is evolving extremely quick and provides a unique tool for small project. The cost should be less than the same project using manned LiDAR in most cases. The data generated for this technology is extremely detailed and the amount of detailed information that can be extracted from this type of LIDAR is impressive. The technology has its limitations as does other LIDAR technologies but the benefits are impressive. It is imperative to understand the limitations of the technology so that it properly solves the problems and provides the solutions for intended applications. The accuracies depending on the type of Drone LIDAR can range between 2 to 3 cm up to 9 to 10 cm. It is always important to ask the questions about the technology and feel conformable with the answers you are getting before trusting a provider and the technology.
James Wilder Young (Jamie) CP, CMS-L, GISP is currently Director LiDAR Services for PrecisionHawk, headquartered in Raleigh, North Carolina, the leader in providing innovative information data using drones. He is currently supporting all components of LiDAR technology as it relates to drone technology. His experience includes all aspects of LiDAR including sensor development, applications development, data acquisition, data processing and project management. He graduated from The University of Colorado.
A 2.477Mb PDF of this article as it appeared in the magazine is available HERE