Please select Into the mobile phone version | Continue to access the computer ver.
Need P4 camera specs for photogrammetric calculations
9170 17 2017-2-17
Uploading and Loding Picture ...(0/1)
o(^-^)o
JBinFla
lvl.1
United States
Offline

Hello,
I'm a Geomatics student working on a paper for my senior project.  I purchased the Phantom 4 to use in my evaluation of the horizontal precision in orthorectified geoTIFFs using consumer hardware and software.  To do this will require a lot of computations on the photos that I get.  Most of these computations need accurate data for items like sensor size (both physical and pixel count), focal length, lens parameters, etc.  Much is posted on the website but some things I'm a bit confused with.  

For instance, photo size is 4000x3000 pixels.  That makes 12mp, but it's listed at effective 12.4mp. Are the actual pixel dimensions more thant 4000x3000?  Of course if they are more, then all my calculations will be inherently flawed.

Additionally, the focal length is listed at 20mm (35mm equivalent). Can anyone explain this? Is anyone in the know enough to know what value I should use?  The difference is large enough that I should be able to calculate which one is correct, but again having accurate specifications is paramount to accurate results.

Thanks for any info you may have.

- Joe



2017-2-17
Use props
Nigelbrinkmann
First Officer
Flight distance : 586391 ft
Australia
Offline

Hi, mmm..well you lost me when you said 'orthorectified geoTIFFs'!
Is this to do with digital elevation?  
Some one a lot smarter than me might be able to answer this for you.
2017-2-17
Use props
Jasonstarbird
lvl.3
Flight distance : 116998 ft
United States
Offline

Google can help you...
2017-2-17
Use props
WindSoul
lvl.4
Flight distance : 16 ft
Canada
Offline

you need to make a plan and ask questions one at a time. at this moment you mix everything in. ultimately i doubt you need specs: all pics are made in the same resolution and have the same dimensions. theoretically the gimbal maintains the projection angle, you have the gps coordinates of both camera and the land feature.
the equivalency of a 20mm focal on a 36mm sensor means the 1/3" sensor of the camera (8mm) gets the same angle of view and focus.
2017-2-17
Use props
Labroides
Captain
Flight distance : 9991457 ft
  • >>>
Australia
Offline

Seems that you haven't got a useful answer yet.
I'll try to help.
DJI list the specs here:  http://www.dji.com/phantom-4/info#specs
The maximum actual pixel size of images is Image Size        4000×3000.
They don't give an actual focal length for the lens and just say that the 94° (diagonal) field of view (FOV) is the same as a 20 mm lens used on a 35 mm full frame camera.
Someone's done some calculations to derive the lens specs here:  http://projects.vizfx.com/dji/DJI_P4_Lens_convert_camera.pdf

2017-2-18
Use props
JBinFla
lvl.1
United States
Offline

Thanks to those who replied.

Many of the calculations require the sensor size and exact pixel size.  This is because, using trigonometric functions, you can calculate the location of the pixel on the sensor, then using the focal length, lens parameters (distortion parameters will be least squares calculated and adjusted, this isn't a metric camera after all!).  Using common points in multiple photos, and using some trigonometry and a little calculus you can start to build a 3D world.  Much software does this automatically, I'm looking to do more of an academic approach.

I thought I was very clear, but to reiterate:

The pixel size is confusing.  They list 12.4mp, but the image is 4000x3000 which is 12.0mp.  This will matter in my calculations

The focal length (that is the lens dimension, 20mm, etc.) maybe you are right, it's a 20mm equivalent.  Unfortunately, 20mm on a 24x36mm negative doesn't tell me what size it actually is for the 1/2.3" sensor.  

I couldn't find this info, but clearly others have it.  I'm hoping some of them (or the developers) can chime in on the specifics.  Thanks!
2017-2-18
Use props
fans687d3f39
lvl.1

Italy
Offline

http://forum.dji.com/thread-47336-1-1.html i think is sony sensor  http://www.sony-semicon.co.jp/pr ... roducts/imx377.html, 12.35 effective MPixel
2017-2-18
Use props
fans687d3f39
lvl.1

Italy
Offline

data sheet http://www.sony-semicon.co.jp/pr ... y_v1.5_20150414.pdf
2017-2-18
Use props
JBinFla
lvl.1
United States
Offline

Thank you!!!!  Page 2 of the data sheet has much of the info I was looking for.  The sensor itself is large than 4000x3000, but is actually 4104x3062 with edge pixels being used for calculation but not image processing.

Again thanks, this info is very helpful.  Does anyone know who makes the lens/frame for this?  The sensor is now a "known" as far as I'm concerned.  Thankfully, with accurate sensor info, I can somewhat reverse engineer the focal length, i'm just hoping it's either 20mm or 35mm, as if I get 29.2mm in my calculations I'll be scratching my head

- Joe
2017-2-18
Use props
FlyaDrone
lvl.1
Flight distance : 188750 ft
United States
Offline

Joe - you probably have by now figured all of this out - however here are links to answer all that you asked and then some.
2018-9-21
Use props
FlyaDrone
lvl.1
Flight distance : 188750 ft
United States
Offline

@JBinFla Joe - you probably have by now figured all of this out - however here are links to answer all that you asked and then some.
https://support.pix4d.com/hc/en- ... the-Camera-Settings

And here is a link to their GSD calculator
https://support.pix4d.com/hc/en-us/articles/202560249

Be sure to look at the recommended links at the bottom of the pages as this is a series of articles explaining exactly what type of information and calculations you are looking for using.
Let me know if this info is too late for you to use.
Regards - Michael
2018-9-21
Use props
Labroides
Captain
Flight distance : 9991457 ft
  • >>>
Australia
Offline

FlyaDrone Posted at 2018-9-21 03:18
Joe - you probably have by now figured all of this out - however here are links to answer all that you asked and then some.

Probably?
He was asking almost two years ago

Links?  What links?
2018-9-21
Use props
FlyaDrone
lvl.1
Flight distance : 188750 ft
United States
Offline

@Labroides There are 2 hyperlinks (https://support.pix4d.......) to Pix4D tutorials in the posting that I made just above your reply. When you bring your mouse over them they turn blue which indicates that they will open in a new tab if clicked on. Feel free to PM me if they are not active for your interface.
Question - what are the units of measurement in your Flight Distance : 3,045,396?
2018-9-21
Use props
Labroides
Captain
Flight distance : 9991457 ft
  • >>>
Australia
Offline

FlyaDrone Posted at 2018-9-21 04:23
@Labroides There are 2 hyperlinks (https://support.pix4d.......) to Pix4D tutorials in the posting that I made just above your reply. When you bring your mouse over them they turn blue which indicates that they will open in a new tab if clicked on. Feel free to PM me if they are not active for your interface.
Question - what are the units of measurement in your Flight Distance : 3,045,396?

The links are there now... but when I replied, your post was a one-liner.
Units ?   Same as everyone else's - metres.

2018-9-21
Use props
Geebax
Captain
Australia
Offline

Actually I don't see the point of all this accuracy as DJI do some pretty major image manipulation in the camera to remove lens distortion and vignetting.
2018-9-21
Use props
JBinFla
lvl.1
United States
Offline

Geebax Posted at 2018-9-21 15:21
Actually I don't see the point of all this accuracy as DJI do some pretty major image manipulation in the camera to remove lens distortion and vignetting.

Lol, although it was 2 years ago anyone looking to do what I did will be interested in that.  To Geebax, the accuracy is critical.  We're flying large sites and looking to get an orthophoto, often 100's of MB in size with a GSD (pixel size) of 1", over areas large as 100+ acres.  AND, we need the resultant image to be accurate to +/- a pixel or two.  This is possible with computer vision (Pix4D, Photoscan, etc.) but you need to have accurate lens and sensor.  In the commercial sector, with $100k-$1m+ cameras flying on airplanes they are what they call "metric", and have been calibrated to often microns.  That's how you get accurate data.

If you're looking for a picture of your kid playing ball, you don't need this.  If you're looking to do actual work and produce accurate maps (for design, construction, etc) then you need to go this far.  
2018-9-21
Use props
Geebax
Captain
Australia
Offline

JBinFla Posted at 2018-9-21 15:28
Lol, although it was 2 years ago anyone looking to do what I did will be interested in that.  To Geebax, the accuracy is critical.  We're flying large sites and looking to get an orthophoto, often 100's of MB in size with a GSD (pixel size) of 1", over areas large as 100+ acres.  AND, we need the resultant image to be accurate to +/- a pixel or two.  This is possible with computer vision (Pix4D, Photoscan, etc.) but you need to have accurate lens and sensor.  In the commercial sector, with $100k-$1m+ cameras flying on airplanes they are what they call "metric", and have been calibrated to often microns.  That's how you get accurate data.

If you're looking for a picture of your kid playing ball, you don't need this.  If you're looking to do actual work and produce accurate maps (for design, construction, etc) then you need to go this far.

That's all and OK, but DJI crop the image output and apply fish-eye correction among other distortions, so it would not be aan accurate image anyway.
2018-9-21
Use props
JBinFla
lvl.1
United States
Offline

Geebax Posted at 2018-9-21 15:33
That's all and OK, but DJI crop the image output and apply fish-eye correction among other distortions, so it would not be aan accurate image anyway.

Not true, the photogrammetric workflow has corrections for lens distortion and what you talk about.  These corrections are calculated based on a least squares type adjustment.  I've actually done this, not just read about it, and of course spent time in school covering the rather extensive math involved.  Although difficult, for a computer and multiple images getting the correction factors for tilt, rotation, perspective, lens distortion, centering, etc. is a trivial task.  Granted, sometimes it takes a while (hours) to process a set of data and produce the resultant orthophoto, but I've done this many times, and have compared it against high-accuracy surveyed control and proven reliably that with sound procedures getting this level of accuracy is not rocket science.  I've done this many times.  This is only possible with accurate lens parameters.

I'm not the only one doing it by the way.   I've been surveying 20+ years and although I'm new to photogrammetry, I'm certainly qualified to measure to the 1/1000th of a meter and have done so and checked the resultant photos.  I'm continually amazed at what the computer vision / photogrammetric data can do with enough photos and plenty overlap.
2018-9-21
Use props
Advanced
You need to log in before you can reply Login | Register now

Credit Rules