Post

3 followers Follow
0
Avatar

usage questions about the Mavic 3 Multispectral

Hi all, I sent an email also, but thought others might benefit from your response too. Some of these questions are about the equipment itself, and many are about how it works with Map Pilot. I hope it's okay to include both topics!

 

1) Are the RGB and multispectral images captured simultaneously, or do I need to select one mode or another during each flight?
2) Is the field of view of the RGB and multispectral sensor the same? That is, will the same flight pattern result in the same degree of image overlap (I understand that the images from the two sensors would have different ground sampling distance if this were true, owing to the different pixel dimensions of their imagery)
3) What is the actual image capture rate for both the RGB and multispectral sensors? I see the RGB sensor is rated at 0.7 seconds per image. In my Map Pilot Pro app (v5.5.3 build 062720232) I'm seeing that there is still an upper speed limit of 2 seconds per image for the default camera possibilities. If I have a fast writing SD card, can I go faster than this limit? In your experience, what is the fastest capture rate in real world settings?
4) How do I fly the Mavic 3 Enterprise series within Map Pilot? I don't see any options for the Mavic 3 series (enterprise or consumer), let alone the Mavic 3 Multispectral. Would I set up a custom camera? What would the settings be for X resolution (pixels), Y resolution (pixels), Pixel Pitch (micrometers), Focal Length (mm), Max Image Rate (seconds per frame)? I assume I'd need to use different settings for both the RGB and the multispectral sensor?
5) I understand that the kit includes the RTK module, but I think that the full RTK functionality is only possible with the DJI base station? I would have guessed that would be this equipment but its description doesn't list the Mavic 3 Enterprise (or any Mavic enterprise level drone) as being compatible.
6) Can I use my own base station to get the centimeter-level accuracy, or does it have to be a DJI base station?
Michael Koontz

Please sign in to leave a comment.

9 comments

0
Avatar

The M3M is well supported using Map Pilot Pro running on the remote that comes with the aircraft. 

  1. Yes. MPP will collect all the images at the same time. 
  2. No, close but not identical. They have different resolutions anyway so it doesn't really match up. 
  3. It can go up to .7 seconds per image but we don't like to run it any faster than 1 second or so. This lets it fly really fast. 
  4. Use the Enterprise version of the APK that you can download from our site. It is a built in camera. 
  5. Talk with sales about that. 
  6. Sales. 

 

Zane 0 votes
Comment actions Permalink
0
Avatar

I should ask: can you tell me what the exact camera specs are for the RGB and multispectral sensor? We'll have to report actual overlap and other parameters when we publish our work. I'm specifically interested in:

vertical and horizontal fields of view for each sensor

vertical and horizontal dimensions in pixels for the images coming from each sensor

vertical and horizontal dimensions in millimeters for each sensor

 

Thank you so much!

Michael Koontz 0 votes
Comment actions Permalink
0
Avatar

Great! Thanks so much for sharing those test data. I'm seeing:

RGB horizontal field of view: 71.1°


RGB vertical field of view: 56.3°


RGB horizontal image dimension: 5280 pixels


RGB vertical image dimension: 3956 pixels


RGB focal length: 12.3 mm


Multispectral horizontal field of view: 62.2°


Multispectral vertical field of view: 48.7°


Multispectral horizontal pixel dimension: 2592 pixels


Multispectral vertical pixel dimension: 1944 pixels


Multispectral focal length: 4.3 mm


Interestingly, the horizontal and vertical fields of view reported in an actual test image are a little different than those reported on the DJI website (https://ag.dji.com/mavic-3-m/specs) which says the multispectral sensor has a 61.2° x 48.10° FOV.


Using the FOV and focal lengths from the TerrainDrop example images we get a 17.58 x 13.16 mm image sensor for the RGB and a 5.18 x 3.89 mm image sensor for the multispectral sensor.


[showing the work: tan(FOV / 2) * focal length * 2 = image sensor dimension; make sure your tangent function knows whether it is getting FOV/2 in degrees or radians!]


I'll confirm all of this when I purchase.

TerrainDrop looks like an amazing tool. I saw y'all's testimonial on their product sheet. I'm going to reach out to them regarding some cyberinfrastructure our group is building-- maybe an opportunity for collaboration!

Michael Koontz 0 votes
Comment actions Permalink
0
Avatar

Teeing up another question here...

Using the equations on Page 17 from here (https://doi.org/10.1002/ecs2.4206), and an assumed flight altitude of 120 m (~400 feet), I get:
 
RGB horizontal ground footprint: 171.5 m

RGB vertical ground footprint: 128.4 m

Assuming I want 90% forward and 90% side overlap, and assuming I set the image capture rate to the recommended 1 image per second, I get:
  
Transect spacing = 17.2 m

Flight speed = 12.8 m/s

Assuming that the overlap settings when planning a mission correspond to the RGB sensor and not to the multispectral sensor, we can calculate the specs for the multispectral images too (using the other equations on page 17 in that linked paper):
    
Multispectral horizontal ground footprint: 144.6 m

Multispectral vertical ground footprint: 108.6 m

Actual side overlap of multispectral images: 88.1%

Actual forward overlap of multispectral images: 88.2%

So my questions:

1) How does Map Pilot plan the overlap for the M3M? Using the RGB camera specifications (leading to a slightly lower overlap for the multispectral imagery compared to the plan) or using the multispectral camera specs (leading to a slightly higher overlap than planned for the RGB imagery)?

2) I'm noticing some artifacts in the corners of the RGB imagery in the sample images you linked to-- as though the shutter was partially closed when the images were captured. Is that related to Map Pilot at all? Or is that something fundamental to the sensor on the drone? Or just something wonky about that specific drone that was flown?

Michael Koontz 0 votes
Comment actions Permalink
0
Avatar

1. You can pick which camera you want it to do the overlap for: multispectral or visual. This is done in the Settings. 

2. The dark corners are vignetting caused by the direct light to the sensor through the lens. DJI has a setting in their app to "Dewarp" the image. This basically means they do some math on it to make it look prettier. Map Pilot Pro does not touch that setting so set it up how you would like it in the DJI app and images taken while using Map Pilot Pro will honor that. 

Zane 0 votes
Comment actions Permalink
0
Avatar

Hello -- I have a M3M from DME and found the multispectral flight planning in Map Pilot didn't work as I expected based on the info here. After choosing the M3M camera under Settings, the maximum flight speed allowed by Map Pilot for a 120 m AGL flight with 90/90 overlap was 5.4 m/s, and the image capture rate was about 2.5 seconds/capture. Changing the image capture rate in Settings did not affect this -- it appears to ignore the user-specified capture rate when the camera is specified as M3M. This is the same max flight speed that DJI Pilot 2 uses for a mission with the same altitude and overlap -- perhaps this is a SDK limitation? I was disappointed that my Map Pilot flights were so much slower than I expected based on the info here.

I did stumble upon an alternative that allowed for a faster multispectral flight. In Map Pilot Settings, I chose M3M-Visual as the camera and planned the mission, but then in DJI Pilot 2 I enabled the multispectral camera, then quit DJI Pilot 2. Then the mission executed by Map Pilot recorded multispectral (and RGB) images, and it *did* use the image capture rate that I specified in Settings, allowing for a much faster flight. However I'm not sure if this is how it's intended to be done or if it was a fluke, if the multispectral would remain enabled after a battery swap, etc. (my testing time was limited).

Any tips on the best approach to achieve the recommended 1 second/capture with multispectral in Map Pilot? Is the approach I used the best approach? Anything I'm missing?

 

derek young 0 votes
Comment actions Permalink
0
Avatar

90/90 overlap is REALLY tight and might be the limiting factor. The M3M should collect images at 1 second per image (set) but a little slower than that is better for consistent results. The limitation you are seeing could also be related to the exposure time. To test that turn off the light based speed limitation in the settings.

Zane 0 votes
Comment actions Permalink