Achieving Faster First-Step Processing with eMotion 3.11.0 and Pix4Dmapper
Fixed-wing drones have made the collection of high-resolution mapping more efficient. This is especially true of medium-to-long range mapping missions that cover large areas and collect vast amounts of data. Once collected, the data must then be processed using photogrammetry software, which can be time-consuming, depending on the quantity of data captured.
To help bridge that gap, our automation team added new functionality to eMotion 3.11.0, specifically to the sensor firmware of Aeria X and senseFly S.O.D.A. 3D, which accelerates the initial calibration stage in Pix4Dmapper 4.3, resulting in a time savings of up to 35-40% if you choose Pix4Dmapper’s accurate geolocation and orientation calibration method.
We have been doing this type of “direct in-flight geotagging” for a while now, but the results were not as robust as desired. While using high-precision GNSS to precisely geotag the location of each image taken during flight is effective; the IMU angles in previous eMotion versions were at a lower precision. With the update to eMotion 3.11.0, and by using the senseFly S.O.D.A. 3D and Aeria X sensors on an RTK/PPK activated eBee X, we’re able to use the same IMU data as before, but with an improved algorithm that better calculates the absolute angle of the camera when each picture is taken.
This upgrade to the sensor firmware increases the absolute angle accuracy given by the high precision IMU in the senseFly S.O.D.A. 3D and Aeria X, resulting in angular accuracies below two degrees in pitch/roll, and four degrees in yaw, which are within the requirements of the “accurate geolocation and orientation” calibration method in Pix4Dmapper 4.3 or greater.
Below we want to share with you the results of the test flights done by our team, which sought to validate the new features and demonstrate achievable results.
For this project, the following criteria was analyzed:
- Processing time in calibration (step 1 only)
- Calibration robustness
- Dataset accuracy (accurate geolocation)
A site of approximately 250 hectares/618 acres was selected to map. This example was used as it provided good vegetation aspects, several buildings, road infrastructure and farmland, which is typical of your average topographic survey.
We took turns processing the captured data and comparing the processing speed between the standard calibration and the accurate geolocation and orientation calibration methods in Pix4Dmapper. Ground truthing of the datasets were controlled using check points.
We used a senseFly eBee X drone with RTK/PPK activation, Aeria X (APSC-C) 24-megapixel camera and senseFly S.O.D.A. 3D camera. A senseFly Geobase to provide RTK corrections and terrestrial GNSS with enabled VRS for capturing checkpoints to verify the project’s geolocational accuracy were also used. And we used Pix4Dmapper to process all the data.Flight tests
For flight preparation, we used senseFly’s eMotion flight planning software version 3.11.0 to create a horizontal and 3D mission block.
Ground sampling distance (GSD) was set to 2.0 cm for the senseFly S.O.D.A 3D and 2.5 cm for the Aeria X, resulting in an altitude above elevation data (AED) between 88 and 118 meters, respectively.
To ensure optimal flight settings, lateral overlap was set to 60% while longitudinal overlap was between 60% for the Aeria X and 85% for the senseFly S.O.D.A. 3D with the tilting angle set to a default 30 degrees.
A terrestrial rover in RTK/VRS mode was used to capture a set of distributed check points. Each check point was given a 45-60sec observation period, which achieved an overall RMS of 3–5mm in X, Y, Z. The datum coordinate system used was WGS 84 UTM Zone 32N with no vertical offset. The flights conducted averaged between 45 and 65 minutes. All flights were flown in standalone mode, and the image geotags were corrected to absolute accuracy using the PPK workflow in eMotion’s Flight Data Manager (FDM).
Processing and results
Once our field work was complete, we evaluated the calibration robustness and processing times. All the data captured was processed in Pix4Dmapper and Pix4D cloud.
Initial results were positive. Once we processed the flight data, we were able to confirm that the standard calibration method took longer than when we used the accurate geolocation calibration method. However, it’s important to remember that the results reflect the first-step processing stage in Pix4Dmapper.
By using Pix4Dmapper’s accurate geolocation and orientation calibration method, we are providing Pix4Dmapper with precise IMU angles and orientation for each photo taken, which results in faster calibration of the image data, since no optimization is needed anymore.
For example, after capturing 1000 images using the senseFly Aeria X camera, we observed a first-step calibration time savings of 35% using the above method.
We also observed similar results using the senseFly S.O.D.A. 3D. After capturing 1500 images, we found that the standard calibration took roughly two hours and 12 minutes to process, whereas employing the accurate geolocation and orientation calibration method took roughly one hour and 23 minutes—a time savings of 37%.
Faster processing speeds were not the only benefits witnessed. We also saw an increase in median image matching and more images calibrated resulted in greater robustness of the processed data.Evaluation of dataset accuracy
After the processing of initial matching and calibration, checkpoints were added—with each point manually pinpointed to ensure positional accuracy.
Additional steps were also taken to determine the positional accuracy of the in-flight geotagging process. Check points were manually computed and the project re-optimized, which resulted in a mean RMS value of the following:
- Error X- 0.005m
- Error Y- 0.013m
- Error Z- 0.009m
All processed flights were found to have an error to check points initial position to be within centimeter-level accuracy threshold (1-2 pixels GSD in X, Y axis and 1-3 pixels GSD in the Z axis).JTNDZGl2JTIwY2xhc3MlM0QlMjJyYXRpby0xNi05JTIyJTNFJTNDaWZyYW1lJTIwY2xhc3MlM0QlMjJyYXRpby1pbm5lciUyMiUyMHRpdGxlJTNEJTIyQSUyMDNEJTIwbW9kZWwlMjIlMjBzcmMlM0QlMjJodHRwcyUzQSUyRiUyRnNrZXRjaGZhYi5jb20lMkZtb2RlbHMlMkZlM2VkNTA0YzgxZjY0ZTk5OWQ0NGMwZDJiMzk3N2E5YyUyRmVtYmVkJTIyJTIwd2lkdGglM0QlMjIxMTcwJTIyJTIwaGVpZ2h0JTNEJTIyNTQwJTIyJTIwZnJhbWVib3JkZXIlM0QlMjIwJTIyJTNFJTNDJTJGaWZyYW1lJTNFJTNDJTJGZGl2JTNFWe also observed no discernible difference in absolute accuracy when using either the standard or accurate geolocation or orientation calibration method. This is good, as we’re able to maintain consistently accurate results while benefiting from faster first-step processing speeds.
The validation project demonstrates positive results and confirms that the eBee X and eMotion 3.11.0, combined with Pix4D mapper 4.3’s accurate geolocation and orientation calibration method, offers significantly faster initial processing, resulting in first-step calibration time savings of up to 37% without any effects on the dataset accuracy.
By using our direct in-flight geotagging algorithm in conjunction with Pix4Dmapper’s accurate geolocation orientation calibration method, we can process large datasets faster. What’s more, we found no compromise to accuracy.
Additionally, using the outlined method enhanced the median of matching and produced better point cloud quality and sharper orthomosaics. We also found that overall data quality was more robust when mapping challenging and repetitive textures, such as dense vegetation.
Georeferencing of the project was also excellent, with an error to check point initial position found to be well within the 2-3 GSD per pixel tolerance.
Click here to learn more about senseFly drones.
Latest blog posts
Talking Drone Training & senseFly’s e-Learning Platform with Andrea Blindenbacher
With the launch of senseFly’s new e-learning platform and dedicated Certified senseFly Operator Program, Waypoint recently sat down with senseFly Global Head of Training Andrea Blindenbacher to learn more about how the new platform and certification course works, where to access it and how senseFly users (and even non-users) can benefit from the various self-guided tutorials...
Expand Your Surveying World
Land departments now accept drone data for cadastre. More GIS users employ drone data in place of satellite imagery. Frequent quarry surveys are now possible, golf course modeling is common, and the list goes on. As operators around the world are learning, a drone not only complements existing survey...
Advanced Drone Operations: 3 Need to Know Benefits for Commercial Pilots
The term ‘advanced drone operations’ may not yet be widely used, but these operations have the potential to bring real, significant commercial value to a wide range of sectors – from agriculture and construction to mining and energy. With unmanned aerial vehicles (UAVs) becoming increasingly accessible for today’s commercial...