Efficient Processing of UAV Projects Dr. Philippe Simard President SimActive Inc. IMAGE
About SimActive Leading developer of photogrammetry software since 2003 Thousands of users in 50+ countries: military & government organizations, blue-chip companies, SMEs Based on the same technology as for large/medium format camera systems & satellite data, Correlator3D processes UAV datasets at unparalleled speeds
Introduction As part of our pre/post sales activities, we have seen thousands of UAV datasets within the last couple of years Quality of input data is key in generating good mapping products This presentation will provide recommendations for efficient processing of UAV data DSM / Point cloud DTM Mosaic
Choosing a UAV Platform Main criteria: Flight planning tools Endurance & stability Operational efficiency Payload (camera, GPS) Fixed-wing vs. multi-rotor Price
Recommendations UAV Larger platforms & stabilized gimbals Camera Avoid fish-eye lenses Favor large sensor sizes GPS Correctly synchronized with camera & INS-aided
Data Requirements Regular flight plan Traditional photogrammetry: 60% 30% UAVs: Minimum of 70% 50% Quick sanity check through images to avoid: Out-of-focus pictures Corrupted files Landing / take-off photos
Effective Overlap Definition: common part between two overlapping images that can be used for photogrammetry With UAVs, effective overlap is often lower than actual overlap because: Cameras used on UAVs may show severe distortion Low-altitude flights may lead to occlusions & other effects Even if overlap appears to be sufficient, effective overlap may not be
Using GCPs Whether GCPs are used or not, relative accuracy remains similar Absolute accuracy usually requires GCPs Main recommendation: good spatial distribution Choice and tagging of GCPs is key Bad GCP Good GCP
Importance of Aerial Triangulation Aerial triangulation is the most critical step in processing Often underestimated, users try to directly generate products instead Detailed quality reports for each image are required Effects of bad AT on subsequent data products: Noisy DSM / point cloud Misalignments in mosaics
DSM vs. Point Cloud For mapping applications, camera is nadir Point cloud will bring limited additional information Point cloud requires larger storage space Manipulation & viewing much slower Whether you choose to deliver point clouds vs. DSMs will depend on the final application
DTM Orthos vs. True Orthos DTM Orthos True Orthos Advantage Visually appealing Advantage Geometrically correct Drawback May see side of objects Drawback May lead to artifacts
Adapting Flights to Mission Specifics Mission plan may also need to be modified depending on the scene to be mapped Autocorrelation for AT / DSM is much harder for fields and forests Such harder missions may require: Larger overlap / sidelap Higher flying altitude Better camera
Common Manual Editing Initial clean-up (e.g. project boundaries) DSM adjustments for reflective surfaces, mainly waterbodies DTM touch-ups Modify seamlines for complex features (e.g. large buildings, powerlines, fences) Adjusting overall mosaic colors
Five Case Studies To show common problems and highlight recommendations, different datasets will be shown: 1. Shoreline 2. Urban area 3. Rural zone 4. Agricultural field 5. Archaeological site
Case Study 1: Shoreline Project Specifics Observations Camera Frame Size Image GSD Canon PS SX260 12 MP 1 cm Missing altitude data Incorrect flight path Image saturation issues
Case Study 2: Urban area Project Specifics Camera Frame Size Image GSD Canon PS S110 12 MP 4.5 cm Observations Unstable UAV platform Poor / inconsistent overlap Low altitude flight
Case Study 3: Rural Zone Project Specifics Camera Frame Size Image GSD Pentax Richo 16 MP 5 cm Observations Proper flight path Sufficient overlap
Case Study 4: Agricultural Field Project Specifics Camera Frame Size Image GSD Canon PS S110 12 MP 6 cm Observations High overlap Acquisition adapted to mission specifics (harder project)
Case Study 5: Archaeological Site Project Specifics Camera Frame Size Image GSD Olympus E-PL3 12 MP 0.7 cm Observations High-quality GPS data Good GCPs True orthos generated
Conclusion Collecting proper data is critical for efficient processing Garbage in, garbage out Choice of proper UAV, payload and mission planning tools is essential Investing in higher-end equipment pays off in the long-run Mapping using UAVs is far from trivial
Thank You! Dr. Philippe Simard President SimActive Inc. Tel.: (514) 288-2666 ext. 21 psimard@simactive.com www.simactive.com