5
$\begingroup$

I was given two photos flown with the X7 and P1. The purpose of the project is to classify the green vegetation through pixel classification. So pixel value consistency is vital. Especially if I am using the same model. Below I have attached an X7 photo and P1 photo. To me, it is easier to identify the green vegetation with the X7 than it is with the P1. I was informed that the P1 provides higher resolution imagery. So why is it harder to identify the green grass, or for that matter any object? I believe the X7 was using the 24mm lenses and the P1 was with the 35mm. I don't hold much knowledge with camera specs, more in charge of the image processing. Could this be a camera issue or a preprocessing/calibration issue? I've found about zero information online except for the basic specs that Zenmuse has released.

I tossed these into QGIS and picked a few points on the yellow line and found the following results:

X7 R:210-220

X7 G:190-200

X7 B:145-155

P1 R:165-175

P1 G:145-155

P1 B:115-125

X7

enter image description here

P1

enter image description here

Update

P1 meta data:

Dimensions: 8192 x 5460
Horizontal res: 72 dpi
vertical res: 72 dpi
Bit depth: 24
Res unit: 2
Color representation: sRGB

F-stop: f/7.1
Exposure time: 1/1000 sec.
ISO speed: ISO-400
Exposure bias: 0 step
Focal length: 35mm
Max aperture: 2.97
Metering mode: Average
Flash mode: No flash
Contrast: Normal
Light Source: Fluorescent
Exposure program: Normal
Saturation: Normal
Sharpness: Normal
White balance: Auto
Digital zoom: 1
EXIF version 0230
Item type: JPG File

X7 meta data:

**Dimensions: 5248 x 3936
Horizontal res: 72 dpi
vertical res: 72 dpi
Bit depth: 24
Res unit: 2
Color representation: sRGB
***Compressed bits/pixel: 3.50474

**F-stop: f/10
**Exposure time: 1/400 sec.
ISO speed: ISO-400
Exposure bias: 0 step
**Focal length: 24mm
Max aperture: 2.97
Metering mode: Average
***Subject distance: 0 mm
**Flash mode: No flash function
Contrast: Normal
Light Source: Fluorescent
Exposure program: Normal
Saturation: Normal
**Sharpness: Hard
White balance: Auto
Digital zoom: 1
EXIF version 0230
Item type: JPG File

** Items that are different

*** Extra items that contain data

$\endgroup$
5
  • $\begingroup$ Do you have any aperature / shutter speed in the exif? Could be helpful to include that. At a glance I'd say your P1 is at least 2/3 of a stop less exposed, which could translate as a lower shutter speed or ISO, either of which could enhance overall picture quality, provided there's no underexposure. Also, while it's not explicit in your post, I assume that the two photos were taken at the same time (no difference in cloud coverage for instance)? $\endgroup$
    – anonymous2
    Commented Aug 5, 2021 at 1:51
  • $\begingroup$ I'll see if I can get my hands on the photos and update my post. These are currently snap shots of an orthomosaic. I can't say they were taken at the exact same time, but they were at least within a couple hours of each other. I would bank on there being no difference in cloud coverage or other atmospheric differences. Especially since it is like this for every ortho I have seen with the P1. $\endgroup$
    – Binx
    Commented Aug 5, 2021 at 14:43
  • $\begingroup$ I don't necessary know what all of the meta data means such as F-stop, but looks like the exposure for the X7 is longer causing the image to be brighter. I wonder if I manually changed everything to be exactly the same (except for dimensions), if that would fix the problem? $\endgroup$
    – Binx
    Commented Aug 5, 2021 at 16:41
  • $\begingroup$ how have you got on with this query? $\endgroup$
    – dhrot
    Commented Aug 11, 2022 at 3:34
  • $\begingroup$ @dhrot There as been little progression. My data acquisition team has been unresponsive with trying/solving/troubleshooting this, so unfortunately I still have no new information. Someone mentioned to me looking into something called LUTZ. I believe it is a method that highlights specific pixel properties that a user chooses, but I am not sure on that. With the lack of effort my team has put into this, I doubt this question will ever be answered. Hoping for the best though! $\endgroup$
    – Binx
    Commented Aug 18, 2022 at 15:04

3 Answers 3

3
$\begingroup$

Rather than continuing in comments, I feel confident enough about the information we have to give an answer.

The differences between the theoretically adjustable settings (shutter speed, aperature) are fairly negligible. Though it is theoretically possible that 1/400th is not sufficient to completely preclude the possibility of drone vibrations reducing image clarity, zooming in on the picture I was unable to detect any motion blur. Furthurmore, that really shouldn't affect the colours all that much.

The aperatures (f-stops) are fairly similar; in a controlled environment I would personally prefer 7.1 to 10 simply because depending on the lens you can start losing focusing capacity as you lower aperature size (raise f-stop), and 7.1 is plenty high enough to avoid depth of field issues.

The big difference that I'm seeing is not only image size, but also sensor size: the P1 is actually full frame, with a 35.9x24mm sensor whereas the X7 is a much smaller sensor, with 23.5x15.7mm. So not only is your P1 taking a higher resolution of picture, the image inside the camera is spread across a larger sensor, which increases the optical precision.

I'm also assuming that the pictures that we're seeing are shrunken versions of originals, since neither is anywhere close to the dimensions that the drones take. Since the editing software starts with nearly double the pixelage for the P1 photo, it has way more detail to play with.

Bottom line: I doubt you can get the X7 to perform as well as the P1; you can try playing with settings, but you simply don't have as much resolution or as big a sensor.

Also, feel free to drop comments asking for clarification; I'm aware I may have thrown in some terms you may or may not be familiar with, but probably easier to clear up in comments. :D

$\endgroup$
3
  • $\begingroup$ I'll have to read this over a few times before I start asking questions :) but I am actually trying to get the P1 to more or less line up with the X7. I've found it easier to recognize objects in the images with the X7 than the P1, which is hard to believe since the P1 is about 1.5x the resolution. I messed around with the saturation, brightness, and contrast values in QGIS. Really ramped them up on the P1 to "appear" more like the X7, really helps with the visual aspect, but I am not sure how that will affect pixel or object identification. $\endgroup$
    – Binx
    Commented Aug 5, 2021 at 20:47
  • $\begingroup$ I am going to upvote your answer, but leave it open for a bit so that I have some time to mess around with the settings. $\endgroup$
    – Binx
    Commented Aug 5, 2021 at 20:48
  • 1
    $\begingroup$ Interesting... My guess is that you're on to something; that the visual identification is easier with the X7 because of lower exposure compensation; it compresses the colours more. I guess I automatically attributed that to a higher bit depth, but looking back over the exif they are actually the same, which indicates that your P1 photo is slightly underexposed. In that case, you may be able to increase colour vibrancy by decreaseing your f-stop (increasing aperature size), decreasing shutter speed, and, if possible, increasing ISO. $\endgroup$
    – anonymous2
    Commented Aug 5, 2021 at 21:44
1
$\begingroup$

I am having a very similar experience with my P1, cant get the greens I need to run weed detection over my Orthocollages, with my photo's coming back washed out an somewhat saturated. I tried ISO and Aperture, but with limited success. Compared to my Mavic's hasselblad I am a tad underwhelmed. Would love to find out what you some up with @Binks. I did find outputting with RAW rather than JPEG helped a little, coloration if not definition. Am also concerned my old trusty microSD from my Hasselblad is not up to the data transfer rate required by the P1 (or the adapter is causing data congestion, is that a thing?) so have fitted a new SDXC Sandisk Extreme Pro 170MB/s card for tomorrow's mission. lets see if that makes any differenceenter image description here

$\endgroup$
0
$\begingroup$

Wasn't sure if I should add this as an update or answer, but I wanted to add a potential answer (though not the solution - problem still being worked on).

To piggy back off of @anonymous2 comments and from something my co-worker thought of, the cameras may be using the same compression transformation. The problem with this is that perhaps the P1 (or X7) could have a bi-model histogram (pixel value range) that when compressed into a jpeg (DJI raw = .DNG), those values are not being transformed "correctly". We believe that it would be best to tweak the ISO, aperture, or shutter and not mess around with the transformation until we need to.

I'll Post another update when I receive some more information.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.