Posting

2 pengikut Ikuti
0
Avatar

Discrepancies in number of images taken and estimate

Hi Map Pilot Team,

I'm trying to troubleshoot a bad Pix4D MapperPro result from a flight the other day. I execute my flights in mountainous terrain with no cell service, and your flight software has been critical to making this project possible. I have successfully processed a couple of maps so far using Pix4D and images taken from flights planned with your software. In processing last week's flight data, I found that a large contiguous chunk of the photos were "not registered" meaning that Pix4D couldn't find the tie points between them.

I suspect the issue has to do with the amount of overlap across the photos, as I'm trying to map forested area (40ha at a time). I was surprised at how few photos were taken on last week's mission (about 1100 photos), because the estimated number of photos was ~1500 and a successfully processed mission from last October generated ~1800 photos (though Map Pilot only estimated about 1250 photos would be taken for that mission). I actually increased the amount of overlap for this latest mission (which is why the estimated number of images is greater for the most recent mission over the same total area), so am doubly surprised at why so few photos were taken compared to the October mission.

I did notice that photos aren't being taken along the flight path as the UAV travels to the starting point or abandonment point (since these are multi-battery missions), but I don't think that accounts for 700 more images for the October mission.

Any thoughts on why there is such a difference between the estimated number of photos and the actual number taken? Ideally, I'd just take more photos without having to fly any slower (since it seemed like it was possible before).

I'm flying approximately square 40ha areas of forest 100m agl in Connectionless Mode and with Terrain Awareness. In October, I flew with Map Pilot Business v2.6.4 with 85% front overlap and 90% side overlap. Last week, I flew with Map Pilot Business v2.7.4 with 89% front overlap and 90% side overlap.

 

 

Michael Koontz

Komentar resmi

Avatar

The image estimate is not something we put a huge amount of time in to since there are so many variables involved. It gets especially hard in terrain aware flight. It depends on how your aircraft does it's connectionless flight, what firmware version is on it and what camera is being used. DJI may have updated something that made it change from a timed trigger to a distance based trigger or something like that. 

The speed of the flight will be adjust according the available lighting conditions if that setting is enabled (it is by default). Manual exposure settings could play a role here since it plays a role in the ground smear that is measure and what speed it gets set to so as not to fly too fast for the lighting.

If you made your SD card write speed too fast the camera may not be able to keep up and may be dropping images which would account for the lower number.

Zane
Tindakan komentar Permalink

Harap masuk untuk memberikan komentar.

9 komentar

0
Avatar

A little more digging into the .csv flight logs...

 

The October flight (85% front overlap/90% side overlap; 1800 photos) looks like it was taking images every 2.03 seconds (the mean time difference between rows in the .csv file when the App_Tip column records a "DJI_XXX.JPG" image captured). The mean speed when those images were captured was 7.2 m/x.

Last week's flight (89% front overlap/90% side overlap; 1100 photos) looks like it was taking images much slower-- every 3.03 seconds and at only 6.7 m/s on average.

I do think I adjusted the SD card speed to the faster setting for last week's flight (1 image every 2 seconds instead of 1 image every 2.5 seconds like October's flight), which I would have thought would make it take pictures faster.

What else might result in photos being taken less frequently? Both days were bright and clear. I assume the shutter speed and ISO settings were the same, but I will check that next...

 

EDIT:

It looks like ISO was 100 for each flight, but the exposure time was faster for the October flight (1/402 on average versus 1/214 on average last week). Would this be enough to slow down the frequency of camera capture?

Michael Koontz 0 suara
Tindakan komentar Permalink
0
Avatar

Okay, so the number of images should be treated as an estimate. Fair enough!

That must have changed since your comment a year ago: "The number of images taken are carefully calculated based on the overlap , altitude, and speeds that are defined by the user. They are pretty accurate. We don't know what other apps do but our number is quite solid." [https://support.dronesmadeeasy.com/hc/en-us/community/posts/115000313283-Number-of-Images]

The flight logs suggest the aircraft firmware was the same for each flight: 1.3.1.00

The camera and aircraft were the same for each flight-- a Zenmuse X3 mounted on a Matrice 100.

The SD card is a SanDisk UHS Speed Class 3 (U3), so it should be fine to write 1 image per 2 seconds instead of 1 image per 2.5 seconds. And it wrote that fast in October (1 image every 2.02 seconds on average).

The speed of the flight doesn't seem to be the issue-- it was pretty similar both times. It seems to be in the triggering of the images themselves.

I learned from here [https://support.dronesmadeeasy.com/hc/en-us/articles/217006946-Connectionless-Mapping] that Distance Triggering isn't possible with my aircraft, so it must be doing Timed Triggering.

The camera was on Auto Exposure both times (and exposure was a bit longer when fewer photos were taken, but just fractions of a second longer. I wouldn't think this would cause a 1 second delay in triggering every image).

So it seems like it's something with Timed Triggering. A photo every 3 seconds versus one every 2 seconds is pretty significant. What can I do to change this timing?

Michael Koontz 0 suara
Tindakan komentar Permalink
0
Avatar

Ha! Straight from the horse's mouth...

Terrain may affect things but timed triggering for connectionless mode will definitely affect things. 

If you have the log files you can email them to us for a look.

Zane 0 suara
Tindakan komentar Permalink
0
Avatar

Thanks! An extra set of eyes on the flight logs would be super helpful. What is the best email to send them to? You just want the .csv files?

Michael Koontz 0 suara
Tindakan komentar Permalink
0
Avatar

Some more thoughts... 

Please use AirData to visualize what is going on here in a better format. Have you checked the actual number of images that get written to the SD card and compared that with what is showing up in the log file. The lines in the log file that say what images got written are not getting written sequentially, there are gaps. The DJI_XXXX.JPG entries in the log file are not guaranteed to get there. For some reason they don't all show up but since it is operating in connectionless mode the triggering is happening on the aircraft and the files seem to be getting written. If they are not, that is a different issue. 

Look for images on the SD card that are not listed in the Log File. 

It also looks like aircraft that are being triggered by distance on the aircraft treat passing through a waypoint differently. Some reset the distance measurement and others don't. The result here is longer gaps after passing through a waypoint and if you have a super complicated linear mission it will have more gaps in it as it goes through all the waypoints. 

Zane 0 suara
Tindakan komentar Permalink
0
Avatar

Wow! Was this patched in 2.8.1?? Is that what the "Fixed the Connectionless interval imaging on older cameras" refers to?

You all are amazing.

Michael Koontz 0 suara
Tindakan komentar Permalink
1
Avatar

We hope so. Since the camera intervalometer can only be set to integer seconds (which is extraordinarily dumb) we adjust the aircraft speed downward to get the overlap to work with the quantized timer settings.

Let us know how it works for you.

 

Zane 1 suara
Tindakan komentar Permalink
0
Avatar

YES! Excellent work and thank you for making this patch available. By slightly adjusting the flight speed in the way Map Pilot 2.8.1 calculates necessary speed for a given front overlap,  I now get image capture every 2 seconds instead of every 3 seconds. I was getting about 1300 images before, and now get about 50% more (about 1900). In Map Pilot 2.7.4, 90% front overlap with the X3 camera used to set flight speed at about 6.7 m/s, and now 90% front overlap sets flight speed to about 6.3 m/s.

Missions are now negligibly longer, and way more photos are taken. Perfect.

 

Michael Koontz 0 suara
Tindakan komentar Permalink