As of Aug 2020, none of the images collected with Phantom4ProV2 are from directly overhead when that should be what is required. One day the camera was collecting images as required, but the very next day all the images were from oblique angles (image attached).
.So it was a standard lawnmower/nadir flight plan and it went to obliques? If so I would have tried to shut everything down and try the same mission again. Sounds like a glitch to me whether it was DroneDeploy or the drone.
Not sure but when using crosshatch & perimeter all missions over several months from different sites all have oblique images. One mission during this period without these features used was top-down.
In order to get the image density obtained with crosshatch I might need to run two separate missions in different directions to get same effect.
Total area is only about 1 acre (60x80metres).
Logically, I would have assumed that only the perimeter images should be oblique, and the crosshatch should be topdown???
That’s it. Crosshatch is always oblique and cancels the nadir part of the mission. If I want a really nice model I will run one crosshatch and one nadir mission. You could run two nadir at perpendicular headings from each other and in the flight interface add the second mission so they run automatically in sequence. I have always thought and recommended that the camera angle be adjustable for crosshatch. As an alternative you can download the “Auto Flight Modes” app from the market and set the gimbal pitch yourself.
@gmcdonald Looks like you are intending to capture an Agriculture scene. What is your goal for generating a 3D model? Short story is that good 3D reconstruction requires oblique imagery. Thanks!
Thanks for that. Is a big help.
…of objects with vertical faces. I would not recommend obliques for terrain-only maps/models, especially if you use GCP’s.
Agriculture yes, but research. I am trying to get an accurate representation of canopy structure. Will never be as accurate as LiDAR but want to be able to get reliable model of crop canopy growth over time for areas as small as 1/8th of square metre (or less).
Primarily using imagery to map accumulation of ground cover (GSD <1cm) over time, but as crop grows I am wishing to extract more descriptive data from imagery (Volume accumulation, height distribution, etc).
You’ll just have to try different ways including processing.
Nadir only with Terrain processing
Nadir only with Structure processing
Obliques only with Structure
Obliques and nadirs with Terrain
Obliques and nadirs with Structure
Nadir only with Structure processing seems counterintuitive, but in my experience makes the object sharper vertically and very slightly improves the texture of the vertical faces. Just make sure to do it the same way every time in similar light conditions and have some type of visible monuments on the perimeter for your calculation areas so you get the most accurate comparisons. Not to muddy the topic, but this is something I would consider the sub-meter GCP’s on.
I don’t believe that distance or angle would have any meaningful, practical effect. What does matter is the sensor’s CFA.
That’s a very small piece of context to comment on, could you elaborate? Considering 90% of these drones have the standard Bayer RGB filter, which is not an option, I don’t understand how mentioning that applies to this scenario with known constraints.
