Is a UAV as accurate as RTK GNSS?
With technology quickly advancing, professional land surveyors are able to perform tasks much faster and more efficiently than ever before. The introduction of GPS to land surveying roughly 20 years ago has made a huge impact in the profession. When first introduced, surveyors were hesitant to use satellite location services. Today, almost every surveyor has a GNSS receiver and uses it daily. The same pattern can be seen with drones or unmanned aerial systems (UAS). Some surveyors are hesitant to start using drones in their surveys for some of the same reasons that were present with GPS—accuracy, time, and cost. This project is intended to compare the difference between RTK GNSS surveying and UAS (drone) surveying. It is intended to show the pros and cons of both RTK and UAS surveying. By analyzing the workflow, accuracy, timetable, applications, and efficiency of both RTK and UAS, this project will show the land surveyor what will be most valuable to them in their practice.
For this project, a typical topographic survey was performed on a parcel of land measuring 20 acres. It was surveyed once using an RTK GNSS receiver and once using a UAS. An RTK GNSS receiver was used in conjunction with the UAS only to set ground control points (GCPs). This is typical of a drone survey. Once the data was collected, the typical workflow was used to produce a deliverable for each piece of equipment. For the GNSS receiver topographic survey, AutoCad Civil3D was used to import points, create a surface, and produce a final map. For the UAS, Pix4DMapper was used to import GCPs and stitch the images together to create a DTM and orthomosaic. AutoCad Civil 3D was used to create a final map. Each process was carefully analyzed for time, workflow, accuracies and efficiency.
To begin this project, a suitable plot of land had to be selected to perform the survey on. To obtain results that would mimic a real world situation, the piece of land had to be big enough to be flown by drone, but reasonable enough to walk the entire parcel. The parcel also had to be free of overhead obstructions, mostly open, and with few people. West Mountain, near Payson, Utah was selected after reviewing a number of sites.
The main materials necessary for the project included an RTK GNSS receiver, a UAS, and aerial targets to record as ground control points (GCP). Other materials included spray paint (for painting check point targets) and lath (to visually mark check points on the ground), and a variety of materials on the office side.
The GNSS receiver used was a Trimble R10 along with the TSC3 controller. The UAS was the DJI Phantom 4 Pro. The aerial targets were custom 18-inch x 18-inch targets designed by Dan Perry from Utah Valley University. The drafting program was AutoCad Civil 3D and to process the images, Pix4D Mapper.
Utah has a great resource for geographic information called the AGRC (Automated Geographic Reference Center). Information for anything geographically related in Utah can be found on their site (www.gis.utah.gov). After downloading aerial imagery the site was georeferenced in AutoCad Civil 3D. Once the images were georeferenced in Civil 3D, a 20 acre boundary of the parcel was produced. This included features that would be used to test how well a UAS could pick up on features. There was a ravine in the middle of the parcel of land that would help to accomplish this goal.
Points were set at each of the corners and were imported into the data collector of the R10 to facilitate setting site corners at the location. Section corners of the surrounding section were also researched so they could be tied into as control for the project. The Utah County Surveyor’s website contained tie sheets to those section corners, including the position of the section monument.
Survey Workflow Using RTK
Once mission planning was complete, the topographic survey of the parcel of land could be completed. Part of this project was to determine the time it takes to perform each form of a topographic survey. Upon arrival at the location, time started being recorded. The first step for the RTK topo survey was to tie into the researched control—the section corners. Once the section corners were tied into and it was verified the GNSS receiver was working properly, the topographic survey began. First was setting the property corners to visually indicate where the desired property was.
A fifty foot grid was used to walk the parcel of land collecting data. Cross sections of the ravine were taken to later detail its topography. One continuing issue with using Network RTK was loss of connection. Stopping to recconnect slowed the job enough to force the use a smartphone hotspot. When the GNSS was connected to the phone, there was no connection loss. Unfortunately, this scenario happens on actual job sites and because it does, the process was included in the final workflow time.
From arrival at the site to the end of the process took approximately 10 hours, including checking control, the actual topo and detailing, handling network loss, and breaking down the equipment.
The other half of a topographic survey is the office side, which began with opening the .csv file that was created and cleaning it up. There were some bad shots that needed to be deleted and typos in description codes and point numbers that needed to be fixed. These are things that are common in the practice. With a clean .csv file, it is much easier to work with in Civil 3D. Including the control shots with all of the topo shots, the point file contained 368 points. These were used to generate 10 foot contours. A final topographic map was made using a title block and would be sufficient to give to a client. The office side of the topographic survey by RTK took approximately 11 hours to create a satisfactory deliverable.
Topographic Survey Workflow Using a UAS
To obtain precise results from a UAS, a GNSS receiver is required to shoot ground control points (GCPs). Without GCPs, the final image will be in an arbitrary coordinate system and can only be used for a nice picture.
The UAS topographic survey began by checking into the determined section corners to check control, then setting GCPs. To begin the aerial survey, GCPs must first be laid out and shot with the GNSS. An aerial survey usually requires a minimum of three to five GCPs. For this project, 20 GCPs were set. The targets were numbered so each GCP could be checked to test the accuracies of the UAS. 50 X’s were painted on the ground and labeled with a corresponding number, and a GNSS shot was taken on them to check them later in Pix4D.
After all the GCPs and check points were set out and occupied, the parcel was ready to be flown. A sidelap of 80 percent and a front lap of 70 percent were used. The flying height used was 150 feet. Once all the parameters are set, the UAS was ready to take off. The flight took approximately 30 minutes, which required a battery change in between.
With the setup of all the GCPs (not including check points since in a normal situation, they wouldn’t be necessary), data collection, and cleanup, it took approximately five hours in the field.
The beginning and ending of the office work of a UAS and GNSS are very similar. Bad shots and typos were removed from the .csv file. Images were checked to make sure all were clean and usable.
The first step in Pix4D, is to import the geotagged images from the UAS. In this case, there were 772 images. From this point, an initial process can be run or GCP coordinates can be imported to improve the location of the image. For the purpose of this experiment, the initial process ran without selecting any of the GCPs for which there were coordinates. Processing took 15 hours. After the initial processing is done, an image is shown of the project. GCPs were then imported to see the error. After selecting the GCPs, a consistent error of 40 feet horizontally was observed. In Pix4D, there is an option to “re-optimize” the photo using the GCPs. After running, “re-optimize” the computed GCPs were still about a foot off from the actual, GNSS recorded points.
The initial process was rerun with selected GCPs and check points. By rerunning the initial process, it yielded much better results. Out of 20 GCPs and 50 check points, only 14 GCPs and 49 check points were usable. Most of the differences between the actual, GNSS location of the GCPs and the Pix4D computed location of the GCPs were 0.1 foot or less horizontally. Vertical positions being over 0.1 foot was not a concern since GNSS isn’t great with vertical. The highest elevation difference wasn’t over 0.5 foot.
With the location of the project being precise, a DTM was processed (which included 10 foot contours) and an orthomosaic. With a ground sampling distance of 1.2307 cm/pixel, this produced a fairly precise orthomosaic and DTM.
The DTM and orthomosaic produced by Pix4D can be imported into Civil 3D. Using those files, a final topographic map was then created for the UAS survey. Typical work practices were used for drafting, including, linework, labeling, and title block. After all drafting was complete a final deliverable was created that would be suitable to give to a client. The time to complete all office time for the UAS survey was 26 hours (11 hours drafting and pre-processing, 15 hours processing).
Each method (GNSS and UAS) for performing a topographic survey will have their own pros and cons. The main objective of this project was to see what differences there were between them, which includes workflow, efficiency, time, and accuracy.
For RTK GNSS, workflow is one that many surveyors are familiar with and are very good at doing. It is smooth and straightforward. The surveyor goes out to the jobsite and collects the points, drafts, and then there is the deliverable. There is no waiting around for processing of images. The efficiency of a surveyor using an RTK GNSS could probably be debated. Walking a 20-acre piece of land on a 50-foot grid will take a full day. But with walking, comes the benefit of seeing everything that is on the ground. That means more detail for the survey. Time can go hand in hand with efficiency. For this project the office time for the RTK GNSS and the UAS were identical at 11 hours (subtracting processing). The time out in the field for the RTK GNSS was double the amount of time of the UAS—10 hours. The RTK GNSS process could be considered more “accurate” because, with the GNSS, more detail can be obtained. Man-made features on a project such as manholes, fences, building corners, and roads can all be tied down very precisely with a GNSS. Features like this can be missed when using UAS.
These are the conclusions gathered from using the UAS to perform the topographic survey. The workflow of the UAS took slightly longer than the GNSS. The UAS survey involves setting out GCPs and shooting points for those GCPs. Then there is the flight and cleanup. When the data is brought back to the office, the images have to be processed. To process the images correctly, the GCPs have to be manually selected. After everything is set up processing could take up to 15 hours. This all needs to be done before the actual drafting can start.
With UAS images, an orthomosaic can be produced. It will be the most current image of the site. Point clouds and DTMs can also be produced from the images and can be used to measure three dimensionally. These are things that cannot be done with the GNSS. The plus side of this workflow is that the processing can be done overnight so the next day, drafting can begin.
The field work of the UAS was very efficient compared to the GNSS. Once everything was set up, the actual flight of the UAS was very quick and efficient. The amount of time to fly this whole project was half the time of the GNSS.
The accuracy of the drone can be dependent on things such as camera specifications, side/front lap, or flying height. One of the most significant items in this project was tying down the GCPs correctly. When the GCPs were used and tied correctly, the accuracy differences between RTK GNSS and UAV were very minimal.
Drone vs. RTK: Is there a difference? Well, yes, there is a difference and there are similarities too. There isn’t a solid answer to which one is better, though. It really depends on what the project is and what kind of surveyor you are. The UAS can obtain similar accuracies to the RTK GNSS, but it may not always be the best tool for the job. In highly populated areas where great detail is involved, a GNSS would probably be a better tool to use. For large, open areas where there isn’t much detail, then a UAS might be the better option.
The option to choose one over the other also depends on the surveyor. Some surveyors wouldn’t mind hiking all day up and down hills and ravines, whereas others would dread that. If a UAS can fly it for them in half the time and half the amount of walking, they would prefer that. Even if a surveyor doesn’t want to use the UAS for a topo, they could use it to search for evidence of monuments or even just to get a current image of the jobsite.
I don’t believe a UAS will ever replace a GNSS. UAS are another tool to add in the toolbox of powerful tools that surveyors have available to them.