Rekon Drone Photogrammetry Performance
In this write up we will walk through a mapping job performed using photogrammetry and see what the accuracy and deliverables look like. But first, what is photogrammetry? Wikipedia defines photogrammetry as "the art, science and technology of obtaining reliable information about physical objects and the environment through the process of recording, measuring and interpreting photographic images."
How do we at Rekon apply that technology? First we have to plan a flight mission that will cause the RPAS (remotely piloted aircraft) to fly uniform parallel lines, equally spaced, and at a specific height covering the mapping area. During the flight, multiple overlapping pictures of the area are taken at an equal intervals. All of that comes together to produce the photographic images needed by photogrammetry software. The Software then performs computations that takes the 2D images, meta-data in each image, survey data that we input, and produces a colored 3 dimensional point cloud of the mapped area.
For this job and most of the photogrammetry jobs that Rekon does we use the DJI Matrice 210 RTK V1.0 (M210RTK) RPAS (remotely piloted aircraft system) with the Zenmuse X4S 20MP camera. The benefits of the DJI system for mapping are that; it has a redundant battery system, it can handle winds up to 43 kph, it is rain proof, and it is capable of operating in temperatures down to -20C. In addition, due to the popularity of the DJI systems there are many options from 3rd parties for flight management applications specialized in producing automated flight paths for photogrammetry.
The mapping area for this project was defined using property lines provided by the client’s operations team. The team required monthly monitoring of the stockpiles that would provide volume reports of all stockpiles with an accuracy of +/- 2% or better. Most of the terrain was exposed gravel surfaces with a very small percent being either vegetation or aggregate processing equipment. Overall, the area was 28 Hectares in size and at the longest dimension was 750 meters in length. Even though the size of this area was large, a single launch site was able to be employed to cover the terrain safely and to maintain a visual line of sight. Ground control points (GCPs) were positioned in 9 locations to ensure the location and scale of the model produced through photogrammetry would be accurate.
After the ﬂight, the survey data was post-processed using GPS-H and CSRS-PPP. GPS·H is a Natural Resources Canada online service that allows the conversion of GNSS ellipsoidal heights h, that are in NAD83(CSRS) to orthometric heights H (heights above mean sea level). CSRS-PPP is another online application by Natural Resources Canada for GNSS data post-processing allowing users to compute positions with a varying degree of accuracy from their raw observation data, the longer the obeveration, the more accurate the position. By using these services, we were able to resolve accurate locations for GCPs as well as shift the elevation data to the correct orthometric height required by client.
The photograph portion of the data set consisted of 1,192 individual pictures. Each of these pictures contained latitude, longitude, altitude and orientation data that described the location and orientation of the camera at every point a picture was taken. The M210RTK system uses an on-board RTK (real-time Kinematic) GPS system to precisely navigate way-points, follow headings and trigger shutter events in the camera. This leads to more accuracy in the models that are produced with this system than are seen in systems that do not employ as precise a flight navigation system.
After the GCP data processing, the raw picture data and GCP data were input into the photogrammetry software. Rekon used Pix4D and processed this project on the Pix4D cloud server. Processing 1.192 pictures took 6 hours and 30 minutes. The resulting data was then moved to a local server where volume reports, DEMs(digital elevation model), Ortho-photos, contour lines and point clouds were exported into the required file formats for the client.
To evaluate the accuracy of the 3D point cloud produced using photogrammetry, 28 ground proofing points were recorded using terrestrial GPS survey. The points collected were then compared to the elevation data of the 3D point cloud.
The error distribution shown in the adjacent histogram illustrates the results. An RMSE (root mean square error) of 0.031m and a Standard deviation of 0.025m were obtained on this project. While setting up and executing a job like this would take approximately 5 hours the first time, subsequent visits to monitor the site would take less than 2 hours. This would result in very economical monitoring of volumes and site topography. In addition, the results also demonstrate how well Rekon's photogrammetry mapping practices can provide accurate topographic and volumetric data in short periods of time.
If you have any further questions about Drone Photogrammetry or if you would like to speak to one of our experts and Rekon please feel free to Contact Us, we would be more that happy to help you with your next project.