What Drives Geometric Accuracy, a Provider's Perspective--Boring?
Written by Jamie Young   
Saturday, 25 March 2017

A 1.645Mb PDF of this article as it appeared in the magazine—complete with images—is available by clicking HERE

There have been many articles over the years regarding LiDAR geometric accuracy and particularly vertical accuracy. It continues to be a relevant topic of discussion based on the continued changing LiDAR technology. It tends to be a boring topic but an important topic to insure that end users receive accurate LiDAR data. ASPRS continues to hold workshops on geometric accuracy and how to check for it. There is a lot of discussion about how good one vendor's data is versus another. At every conference there is discussion about a particular project or a particular provider and their accuracy. It is important to take any talk about a project or provider regarding accuracy at face value. If your company provides perfect data on every project, every time without any mistakes ever, where do I sign up? Transversely, you, the people talking smack about other companies, get the facts and then draw your conclusions because hearsay is just that.

Who should have concerns about accuracy and what do providers have to deal with to try and get good geometrically accurate data? Who should have concerns about LiDAR data accuracy? The data providers are more concerned about meeting the required accuracies as they relate to a given specification prior to delivery. This is on most every LiDAR provider's mind. Some more than others and there are several reasons why. The LiDAR market is extremely competitive and especially in the larger area collection market such as for USGS. Furthermore USGS is very clear on what is expected. Additionally, since the market is so competitive there is little margin for error as mistakes and rework drive the cost up on a project. Hopefully, this speaks for most providers but you can bet we want to get the data right the first time. Contrary to a few peoples believe there are very few LiDAR professionals that do not have pride in their work and want to do a good job. The LiDAR data providers have to deal with LiDAR systems and software that does not preform the way we would like it too. In the past 10 to 15 years there has been numerous projects delivered with accuracies that far exceed the manufactures published sensor accuracies. This should speak volumes for the professionals doing the work or maybe the sensor manufactures just like to underestimate their sensor's capabilities. It is probably a little of both. Additionally, the manufactures are driving to be the best, so sensors are pushed to market and new challenges with the complex sensors arise in a production environment that where not realized in a R&D environment. The manufactures are very thorough in trying to test their sensors in all types of conditions but It's guaranteed that providers figure out ways to use the systems in configurations that the manufacture never thought of.

The client/end users obviously are concerned about getting accurate data. Users want what they paid for and users are very good about making sure their data meets the requirements. Additionally, a lot of projects have social, political and economic implications that if not right will impact several people's lives. One of the first main uses of LiDAR was FEMA flood Plain mapping which if not correct has serious social and economic ramifications. If it was possible to assess the positive social and economic impact of LiDAR for this application, it would be staggering over the last 20 years. FEMA used to rely on maps accurate to 4 foot vertically to assess the flood zone verses the now QL2 accuracy which is 10cm RMSEz.

LiDAR providers struggle with bidding on projects as a result of lowest bidder prevails which in turn potentially compromises the accuracy LiDAR. There are many well established LiDAR providers that refuse to bid on projects because the winning price per square mile could potentially compromise the quality of the product including accuracy. It is not to say that those providers that don't bid on a job couldn't do it for the lowest bid but we just decide the risk is not worth the return and the potential to compromise our product or reputation. There is enough specialized work out there that can meet our revenue needs. There has been several articles about collecting and processing quality LiDAR for a given price and yeah it is possible but not desirable. The bottom line is making money and doing a good job. That being said these are the components of providing a quality LiDAR product, factors that affect a quality and accurate LiDAR product and what we as providers think about when generating our products for the end user.

It is clear that as a result of lower prices, there is a rise of poor data quality and degraded accuracy as a trade-off to push bidding price down. Data providers vary the procedure, frequency and extent of their LiDAR calibration. When efficiencies and technologies warrant, we improve the process to keep the cost down. This is not to say we can't do things cheaper. It is do we want to? A lot of data providers use automated boresight and calibration tools which could potentially have negative outcomes. The manufactures of software have made vast improvements on the automated boresight and calibration software but some manual assessment is more often than not required to guarantee an accurate product. The important component of calibration and the required accuracy is understanding the sensor limitation and the associated error budget of the project specifications. As previously indicated the providers typically get the most out of the sensors as they relate to the stated accuracies. The one positive for automated boresight and calibration is that there is a lower skill level required to execute the process. The problem with that is that if there is a problem will it be realized immediately and will it be addressed in an efficient manner. The problem with automated boresight and calibration is that it does not always find and flag calibration issues, acquisition issues, sensor malfunctions or human mistakes. Additionally, unfortunately some cheat to get the proper calibration and accuracy. This is very difficult as it relates to certain specifications such as the USGS NGP LBS specification which helps keep providers honest. Some of the ways to cheat are chipping off or reclassifying edge lap. This gets rid of the least accurate data. Also, mathematically a custom "geoid" model could be created from vertical error between flight lines and ground control. Some providers can hide error through other creative techniques especially if they discover problems after the plane has left the project area.

Calibration or boresight is still key in getting required geometric accuracy. For purposes of this article this process will be referred to as calibration. Typically, several steps of calibration are required to make sure that a sensor will yield the results required. These steps are calibrated after sensor installation, scheduled calibrations, and project calibrations. Additionally, control points will be required for calibration. These provide a reference to the absolute accuracy. The control points are independent to the check points used to check the data to make sure the data meets the project specification. The installation calibration is done when the sensor is installed in the platform. This insures that the sensor is calibrated prior to mobilizing to a project. It also insures that the sensor properly functions before arriving on a project. Scheduled calibrations are done when the sensor remains in the aircraft for an extended period of time to insure that nothing changes over time as a result of aircraft vibration, human interaction and sensor wear. The project calibration is typically cross flights or some configuration of calibration over an airport. This is done as a result of changing measurements of the sensor between flights.

The planning is very intentional as it relates to achieving geometric accuracy. Typically, the project is planned at higher point densities. Flight lines are constrained to certain distances based on sensor functionality. These distances can vary based on sensor and provider. Although the technology has improved dramatically over the last 20 years, the flight lines are intended to fly in a certain pattern to make sure that the IMU continues to find itself. One important change is that we now don't have to solely relay on base stations for the flight reference. CORS and VRS and other control networks can be used effectively or a combination of any of the above can be used. Once the project is collected the key as said before to getting good geometric accuracy is calibration. This involves solving for the errors or minimizing the error is a better way to describe this. This distributes the systematic errors equally throughout the data set. There are several ways of doing this and it will vary by provider.

There are several quality assurance steps most providers take during the calibration and processing of a data set. These steps are ingrained into the process and are vital in making sure that a data set meets the geometric accuracy requirements of a project. It should be noted that steps are taken to insure most importantly the vertical accuracy of the data but now the horizontal accuracy is becoming a key component to checking accuracy. Horizontal accuracy was assumed to be correct as a result of good vertical accuracy. USGS has spent the last couple of years developing ways to check horizontal accuracy and they have made the process available to the providers, which is helping this process along at a much faster rate if done otherwise. Many providers have developed ways to check horizontal accuracy as well.

Although, geometric accuracy can be boring it is a very important part of providing a quality product. End users can be assured that most providers take geometric accuracy very serious and try to provide a product that meets or exceeds the project requirements. The drivers in cost continue to drive the perceived geometric accuracy or lack thereof but end users can be assured that the providers continue to improve the process as a result of competitive pricing. Bottom line is that the end users will continue to see geometric accuracy improve because we continue to do things better, faster and cheaper.

James Wilder Young (Jamie) CP, CMS-L, GISP is currently a Senior Geomatics Technologist for Merrick & Co. located in Greenwood Village, Colorado. He is currently supporting all aspects of LiDAR technology development and is lead for all HDS and Drone functions. 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 1.645Mb PDF of this article as it appeared in the magazine—complete with images—is available by clicking HERE