At DroneDeploy, we define a good drone map to have:
- 99%+ image coverage of the area of interest
- High data quality
- Fast delivery
But, on occasion, this cannot be achieved and you may see gaps (unstitched regions) in a map. This could be caused by a whole host of factors.
Stitching can be considered to be a ‘black box’ process; data is input but the mathematical process that drives the stitching is complex and difficult to predict — or even to precisely understand. In that sense, you could think of mapping with a drone as a bit of an art.
There are various causes of stitching failures
Below are some examples of IR (infrared) images (taken with an AgEagle using DroneDeploy) that show the comparison between a normal image and an image with an undesirable effect. The effects have been exaggerated for clarity.
(caused by fast-moving drones or vibration — less is better)
(camera out of focus — in focus is better)
Vignetting on images
(dark areas around the edges of the image — less is better)
Insufficient image overlap
The graphic above shows the progression of consecutive images taken from 0% up to 70% overlap. The higher the image overlap, the easier it is for image processing software to stitch together your images. This is because there are repeated areas from the images before it.
It’s similar to making a photo collage on your desk from actual photographs and trying to line them up so that they align perfectly. High overlap is best when melding all of the images together.
(e.g. photos taken during turns )
Non-nadir photos affect fixed-wing aircraft more than multirotor aircraft since during turns, fixed-wing aircraft have to bank at higher angles. A photo taken at an angle will be distorted in relation to the others.
This condition can be reduced by making sure that your flight plan is set in a way that allows the aircraft enough time to complete its turn before performing the next pass over the area to be mapped.
Photos taken at low altitude
If photos are taken at too low an altitude and at high ground speed they can be blurry, hard to stitch together, and you will also not be able to cover as much area in one flight as you would be if you flew at higher altitudes.
Remember to always obey your local/national altitude restriction regulations.
(e.g. water, full crop cover in fields)
Water is notoriously difficult to map by drone photogrammetry due to its constantly changing surface, refraction/reflection of light, and tendency to have hard-to-determine patterns.
DroneDeploy - Put your drone to work!
Map created with DroneDeploy: "Millennium Falcon"
And in the 3D model of the same map (note the jagged, sharp water of death that reaches 100 feet in the air):
The anomaly of homogeneous imagery affects all image processing software. It’s just that difficult for computers to process.
If you must map bodies of water, give the very expensive (and slightly impractical) aerial LiDAR methods a chance.