Follow

Flight Recommendations for Better Reconstruction

Flight Recommendations

Outputs are highly dependent on the raw dataset that was uploaded by a user.  If there are holes, noises, and other unwanted features appearing in the reconstruction, there is a chance that there are issues with the photos, especially images capturing terrain not suited for feature detection.

In this article, we'll discuss basic flight recommendations to help you get better 3D reconstruction of your site!

 

Front and Side Overlap

As a general rule, we recommend flying with

  • Front overlap of 80%
  • Side overlap of 70%

Going slightly below these overlap values may decrease flight time and still produce good results, but introduces small risk of inconsistency.

Low overlap values are at high risk of bad reconstruction.  Please avoid sacrificing output quality for better flight times, as this might lead your team to doing a refly.

 

Altitude and Speed

Generally, recommended settings for mission planning would be 3 - 8 meters/second ground speed at an altitude of 60 - 120 meters. However, sometimes it's beneficial to fly the mission at a higher altitude so each photo has less distortion and contains a larger area:

  • Agricultural fields with repeating patterns
  • large solar panel installations

For sites with significant changes in terrain elevation, flying at a fixed altitude above the ground will produce the most consistent output.  When that's not possible, multiple flight paths should not differ by more than 10 meters.

 

Images to Filter out/Avoid 

In general:

  • Avoid any blurry images (due to fog, dust particles, heavy cross winds), and remove photos captured on takeoff/landing and traveling to/from the flight area. 

Ticket_Teck_GHO_-_BRK_Blurry_images_due_to____Skycatch___Zendesk___2022-12-29_at_7.22.06_p.m..png

  • If oblique imagery is included, it must have at least 80% overlap on every side (top, bottom, left, right) and the object of interest should be centered in each photo.
  • Avoid capturing the horizon or sky as that will affect your photogrammetry results and time of delivery significantly.

 

skyview.png

  • Avoid capturing areas with a lot of featureless surfaces like water, reflective or featureless structures like roofs (repetitive metal sheets), and dense snow areas.

Ticket_Holes_in_point-cloud_due_to_water____Skycatch___Zendesk___2022-12-29_at_7.29.54_p.m..pngTicket_Holes_in_roof_due_to_featureless____Skycatch___Zendesk___2022-12-29_at_7.31.27_p.m..png

Ticket_TDRSA_from_Feb._23__2022_Outpus_with____Skycatch___Zendesk_2022-12-29_at_7.53.07_p.m..png

  • Avoid capturing areas with dense vegetation.

Ticket_Clayco_-_10103603_-_Inland_Woods____Skycatch___Zendesk_2022-12-29_at_8.01.24_p.m..png

 

In cases where you need to capture tricky terrain with dense vegetation, water body, featureless areas, or snow, please follow the recommendations below:

Dense Vegetation

  • Increase the overlap and Increase the flight altitude (the higher, the better chances you have for our engine to match visual similarities between overlapping images in vegetation areas )

Water Bodies

  • Increase the flight altitude (to cover more area per image) and Increase the overlap.
  • At least 30% of the land area captured by the images should be visible.

Featureless Areas (Roofs, Metals, Buildings)

  • Increase the flight altitude (the higher you fly, the better chances you have for our engine to match visual similarities between overlapping images in featureless structures) and, if the reconstruction is still undesirable, add images with different perspectives (Ex. Oblique imagery at high altitude that focuses on the target of interest, avoid using images that show the sky or the horizon)

Dense Snow

  • Increase the overlap and Increase the flight altitude (the higher you fly, the better chances you have for our engine to match visual similarities between overlapping images in highly vegetated areas). Delete blurry images.
  • Images should contain at least 30% of the bare ground or any other clearly distinct features.

Note: Images with 100% area covered in dense featureless snow, will most likely not be able to generate key points for a correct reconstruction.

 

 

 

Was this article helpful?
2 out of 2 found this helpful
Have more questions? Submit a request

Comments

Powered by Zendesk