Hello,
I've recently finished calibrating my X4S for photogrammetry work. I've put a few weeks of research into this process, reading academic papers on image optimization for photogrammetry as well as ensuring that every pixel was pampered all the way into the photogrammetry process.
My setup and parameter settings are below:
Equipment:
Inspire 2
Loki PPK Unit for
Inspire 2
Lab Calibrated X4S including Interior Orientation file
GPS Base Station/Rover for base and GCPs
Picture Profile Settings (measurements made in a lab in a controlled environment):
WB: 5200k
Sharpness: +3
Contrast: -2
Saturation: +3
Color: Art
Mechanical Shutter active: maximum shutter speed up to 1/2000th
ISO: 100 or as low as possible
After 2x days of meticulous testing of every color/contrast/exposure/saturation combination, I found that D_Log, as well as D_Cinelike, is much too noisy to be useful. It was the Art profile that had the best balance between Dynamic Range and Noise.
You have to ensure that your image is white balanced correctly before each flight. Just use a gray card and set camera in DJI Go 4 to AWB and fill the screen with the card. Watch the reading. Then switch back to custom WB and set to the Kelvin number.
Also, I roll on a color chart (DataColor Spydercheckr 48) before each flight. Ensure to roll on the color chart AFTER you have white balanced.
Make sure to calculate your exterior orientation before photogrammetry as well using some sort of bundle adjustment software/plugin.
The higher the resolution of your images, the more accurate your photogrammetry. This means flying at lower altitudes (but not too low as to skew building tops!)
Don't over or underexpose the image. The more pixels you overexpose, the fewer pixels the photogrammetry has to match/tie points. Same with underexposure. Noise is the enemy of matching points.
It has been proven that less contrast is actually better for matching/tying points, however, you have to find the optimal balance between dynamic range and noise floor.
You may reference my post
here with regards to SNR per each ISO selection.
Since we are currently limited to JPEG recording (as of this posting date) you have to ensure the image is as close to balanced as possible before it's encoded to 8-bit JPEG format.
Don't transcode or work from JPEG copies of your original images. JPEG compression degrades the geometry of the image files used for photogrammetry.
The ultimate goal is to pack as much information/pixel into each image as possible and ensuring the integrity of the geometry.
Hope this helps.
Z