Summary of Issue: I have a P4P V2 and I keep my camera size to 3:2. I did a mission last month for Drone Base.A couple of weeks after mission I got an e mail notifying me that the photos were acceptable but suboptimal. When I asked about suboptimal meant they told me that the photos size was at 4:3 and my camera would do 3:2. Ok so I checked my settings and its set to 3:2. I went out after that and did a mission shooting some photos individually with me snapping each shot. I then did a Drone Deploy mission same spot and the mission was successful. I then looked at the photos and all the photos I took were at 3:2 and all the photos taken by Drone Deploy are 4:3. are there settings in DD to allow me to choose the size.
Date Issue Began:
Verified 11.10.18 Drone Model:
Phantom P4P V2 Mobile Device Model and OS version:
I pad mini 4 IOS DroneDeploy App Version: latest
I don’t believe there is any way to shoot 3:2 in DroneDeploy, but it should not change your DJI Go settings… Last time I flew for DroneBase their requirement was 4:3 so it’s interesting they flagged that.
It doesn’t change my DJI Go4 settings, it remains at 3:2. Only the Drone Deploy app doesn’t shoot anything but 4:3. So sad I guess I’ll have to start using GSPro.
One advantage of 4:3 is that it better balances front and side overlaps. For the same side overlap, the angle to the far side extents with 3:2 is greater resulting in more photogrammetry error. So to approach the same quality as 4:3, it appears you would have to increase the side overlap to get the same angle over the 4:3’s field of view to the sides. So maybe 4:3 is an ok choice. 3:2 gives you more pixels on the sides so you could cover more area in the same time with the same % overlap but your new pixels on the side are not as high in quality. Flying 3:2 with the same speed and pattern as the 4:3 should give a slightly better result but at the cost of longer upload and processing time.
Interesting subject (4:3 vs 3:2 aspect ratio). I’m not yet convinced that shooting at 4:3 would provide superior results if the native resolution of the sensor is 3:2. Any real logic to back that up? I do know for a fact that if you throw away 10% of every image (which is what is happening using 4:3 on a P4P for example) you are going to need to collect a lot more images to cover the same area. The bigger the area, the more wasteful it becomes. Looking forward to good evidence that discarding 10% of every image is actually a good thing.
This talk has been had tens of times on these forums and has never made sense to me. Does using the full frame cause more trouble with distortion? I know that there is already some correction being done during processing, but does it get worse on the full frame? Regardless, the only flight parameter that would change is that our sidelaps would report higher. No, there wouldn’t be any need to fly lower or slower for frontlap because it’s the same frame.
See my updated post above. I checked the P4P camera specs and found only the pixels in the width changes. Hence my changed comments. Do my updated comments make sense?
Terry, is your conclusion that pixels recorded at a higher angle are inferior based upon SFM (Structure from motion) science or just anecdotal? If true, then wouldn’t a 1:1 crop provide the best quality? And if adding angle by virtue of a higher aspect ratio is bad, but requires less flying/images as a compromise, how do you qualify how much quality is lost by increasing the ratio? I don’t mean to put you on the spot and I’m not saying you’re wrong.
For the record, Map Pilot captures the full frame on the P4P and their processor doesn’t object.
Distortion and vignetting are things that happen when you push the extents of a lense by trying to use the full frame of the sensor. We could talk about this all day and totally disregard all the previous posts. @Jamespipe, is there any reason in particular DroneDeploy chooses 4:3 aspect ratio?
At higher angles the distance is greater and each pixel covers more area for close to flat surfaces so resolution is degraded. Make sense? The resolution loss is compounded by the greater photogrammetry error that comes with increased distance. The amount of quailty lost is best determined by experiment as I did not write the photogrammetey program and so do not have an accurate way to assess its contribution.
I will try to do some testing this week. The same property from the same elevation only adjusting the sidelaps. One flight with 3:2 and one at 4:3. @solarbarn, would you be willing to do some analysis?
I did find one of the previous post responses from DroneDeploy and the reported a higher rate of image capture failures. This does make some sense as the images would be bigger, but no one ever followed up on it.
It makes sense, but at what compromise? Maybe it’s negligible or maybe it’s dramatic. If it’s dramatic, then using that approach, maybe we should be using 1:1 aspect ratio, right? Or, maybe adjusting the capture altitude up 50’ would cancel out the loss? Lots of variables.
I’m not buying the “images are bigger (larger file size)”. The 4:3 images from the P4P are larger than those from the 4:3 P3P. I’m thinking that 4:3 is used because that’s what DD started with rather than any particular metric at this point.
Doesn’t seem like much of a compromise in my opinion. I will run some tests, but expect less than stellar improvements. If anything it would be a slight time savings. If there is more noticeable ANYTHING at the edges of the 3:2 that could cause negative results then I for one am fine with what we have. I have post-proceesed enough images (drone and DSLR) to not need another variable. If DroneDeploy did analytics and there was a notable issue that could potentially shed a poor light on their software or cause more harm than good then I trust them.
As for not buying into larger file sizes it seems pretty simple to me. If the images are 11.2MB instead of 10 then it is totally feasible that a 10% increase in file size could have adverse effects on the consistency of writes. I already see occasional issue with image writing/spacing being off, but within an acceptable range. Not to mention that we are already on the edge of the hardware capability. Not at the same rate, but sounds like the same reason we can’t shoot RAW.
Are you saying you think a few mb makes a difference in how fast the images can be written to the sd card? Or, it makes a difference in the processing? Remember we’re just talking .jpeg here.
“Consistency of writes” is what I referred to. It doesn’t matter what kind of file type it is, 10MB is 10MB. Add another megabyte or two to that every 2-3 seconds. All I’m saying is that it is fallable and totally dependent on how much detail is being captured, and I have seen it plenty of times. I average about 1,500/week and easily see 3-4 out of place images every flight, or about 1%. If it’s not fallable then every shot should be perfectly spaced every time.
I don’t know, Michael. I don’t really think that perfect spacing is critical for one thing. Getting a consistent overlap is obviously the goal, but if it varies slightly but doesn’t drop below a certain level (depending on type of ground cover) then it probably is acceptable. I doubt that the difference between a couple of MB in file size is what is causing any inconsistency anyway. I would look more at wind gusts, changing direction relative to the wind and things like that before the time it takes to write an image that is a couple more MBs. But I’m just guessing of course. When the conditions are such that you can clearly hear the uas, you can hear when it all of a sudden it adds power or makes a correction to stay on course at times. That’s where I would suspect the perfect image timing takes a hit.
Man @Dave, I know you like playing devil’s advocate but you’ve gone completely off the 3:2 topic at this point, lol. Everything that I mentioned stemmed from the fact that 3:2 images will be larger than 4:3 and in turn are a potential negative factor in a system that is already susceptible to small errors. I could care less whether or not 1% of the images is not in the correct place. Going all the way back to the fact that it might be a reason why DroneDeploy isn’t supporting 3:2. I just did a control set of 10 images in each aspect ratio and the 3:2 images are on average 25-30% larger.