The way to make the perfect 360 sequence

[Eesger]

In my opinion there are several things to consider:

1. the hardware (camera & GPS)
2. method of traversing the sequence
3. processing the data

1) the hardware
I think the Insta360 pro (2) is the best quality camera that is more or less affordable. I have had a love/hate relationship with my YI360 for a while now, also I have bought…

I now have a GNS2000 connected to my iphone. The GNS2000 is more accurate then the standard GPS sensor in the iphone (and all other consumer smart phones…).

Does anyone have experience with other (better?) location positioning hardware/software?

2) method of traversing the sequence
I have discovered that it is best to get the best data before processing it. The QooCam can level out by itself, the Insta360 Pro also, but that takes away part of the quality or has still buggy software… to get the best bases… I think one needs a gimbal.
Mapillary has SfM (Structure from Motion) this is also is a sort of post proccssing horizon leveler. but again, when you have a good gimbal setup, that quality is much better. And I want to remove Mapillary out of my equation.

Since I have spoken with a Insta360 pro user that to get a good sequence with it, one also needs a gimbal I decided not to buy an Insta360 pro and keep my YI360 for now (then the only remaining real reason would be the increase in pixels… but quality of the sequence is more than just pixels…) So I bought a Moza guru 360 air gimbal (bought for the future, it should be able to handle the Insta Pro also).

3) processing the data
I now can simply dump my GPX file and the MP4 file into a folder and activate my script, that does it all… create seperate images, detect the image rate, match the GPX file, match sub second locations, stabalize GPS data (the point earlier and after, averaging on the distance of the requested position), calculate the orientation… the works

Setting up the gimbal was quite a bit of figuring out… but I think i have a working setup… gonna find out if the real world matches with my theories

[dgreenwood-trekview]

@Eesger

3) processing the data
I now can simply dump my GPX file and the MP4 file into a folder and activate my script, that does it all… create seperate images, detect the image rate, match the GPX file, match sub second locations, stabalize GPS data (the point earlier and after, averaging on the distance of the requested position), calculate the orientation… the works

Is this available to look at anywhere @Eesger?

Also what was your decision making process in choosing video over time lapse?

[Eesger]

@dgreenwood-trekview

what was your decision making process in choosing video over time lapse?

The first very simple reason is that the YI360 has no time lapse function
Second, even if it did have one, the interval is moderate (about every 2 seconds). Just good enough for walking, but for on a bicycle it’s too slow. The YI has very nice settings for frame rate, ranging from 0.5 seconds to 60 seconds, I mostly use 1s. I am working on a setup where I can use it with my motorcycle, then I think I will try standard video and see if that is workable.

I do know that seperate images have better quality then images from video… but the software simply does not support it… And the difference in quality is not then huge to spend a lot of time in to build something for it…

@dgreenwood-trekview

Is this [the script] available to look at anywhere?

I started my company when PHP barely was used and Python did not exist. So I started my first “site automation” programming in Perl CGI. Nowadays I find that making command line scripts works well for my to do in Perl… so this is a Perl script… I could make a GitHub of that, but I now only have experience as a user (and a helper of sorts) of GitHub… I haven’t spent time figuring it out as a “project owner”…

[chelseabrian]

To answer your question about better hardware for GPS, I am looking into the Arrow 100 which boasts Real-Time Submeter GNSS Accuracy. Its a bit pricey at ~$2000 but it look like the perfect device for SV capture.

The Arrow 100 submeter GPS receiver was designed specifically with GIS and mapping users in mind. This receiver squeezes more accuracy (submeter GNSS out of the box) from GNSS constellations and free SBAS corrections than any other receiver in the world. Thanks to its patented technology, you can use the Arrow 100 submeter GPS receiver under trees, around buildings, and in rugged terrain — where other receivers will fail to deliver. And thanks to real-time corrections processing in the field, you no longer have to spend time post-processing your data.

[Eesger]

Thanks for that info, I hadn’t found that one. But $ 2000 is a bit much… I would like to get better resolution, but for a better price also

[Eesger]

I read/hear here:

An Introduction to GPS

A timely post, Have been looking at the GPS issue from a different angle this week - improve the hardware in the camera for better accuracy even in harsh environments removing the need for any post capture additional work. Watch this demonstration of one of the latest chips. Centimetre accuracy. Jump to 3:45 for the start of the demo. Enjoy

That such a unit is under $ 100 ? That is quite a bit less then the $ 2000… is that the “hobby in action” Or might there be cheaper sollutions out there with (much) better resolution then my GNS2000

If someone would be able to create a unit that:

• is USB powered
• uses the ZED-F9P
• uses accurate time registration
• can record location & altitude at at least once per second (twice or even four times a second would be even better)
• stores data on a USB stick (or memory card) in GPX file format
• add a start/stop button

If you can get the price below or around $ 250 I’dd buy one!!

update:
searched the net for a bit:

Building a simple u-blox F9P data logger with a Sparkfun OpenLog board

Interesting…

[update]
Interesting read:
https://www.agsgis.com/Bluetooth-GPSGNSS-For-Mobile-GIS-Field-Tested-and-Compared_b_38.html

[Eesger]

@Eesger

I do not have the capability to level out images nicely with software, not automated Anyway. Thus I bought a moza guru 360 to get a better bases…

For that ‘perfect 360 sequence’ what do you think… go for ‘the perfect base data’ in using a gimbal to get a good level bases… or later on fix it with software?

Adding to that… try and get a ‘perfect gps location’ or enhance it later on with software…

Or somewhere in between… what’s better…?

[OmniSynThesis360VR]

Arrow 100 press release in 2014, that looks very legacy to me @chelseabrian

[OmniSynThesis360VR]

Already thought that the original post is Jim’s
I think that I prefer the F9K.
We are talking to Labpano to have such in their next gen camera next year.
Needing to make the case that geospatial imagery market segment is worthwhile enough for them to invest more in it.

https://www.u-blox.com/en/docs/UBX-19047326

For the time being, happy with my Labpano Pilot Era in most situations. Check GSV and Mapillary on Curaçao.

[Eesger]

The labpano 8k was unknown to me until you mentioned it. It is nicely compact! If they could squeeze some extra megapixels in there and that F9K for their next version (preferably for the same price ) I’ll buy one!

(It can do 7fps at the highest resolution, nice. Is there image quality improvement in photo mode? How fast can it do timelapse images?)

[OmniSynThesis360VR]

I do not want higher resolution, I consider 8K the sweet spot for several reasons. I would like better dynamic range, however. You are referring to the Pilot One. Better check their own website Labpano.com. That F9K is not likely for the same price.

No better image in photo mode. Currently, it doesn’t do photo interval, but @JimGayes360 and I are trying to talk them into that. Would be helpfull if you express your interest by writing to support@labpano.com

It can do timelapse video, but not at a useful rate for us.

[smathermather]

Depending on where you are doing your image capture, even the best GPS will require position correction in places like urban canyons and real canyons. I did some work on this a few years back using proprietary tools:

FOSS4G Korea 2014, poor GPS photos, and mapillary (part 2 of n)

But it should be easily implementable in OpenDroneMap (full disclosure, I am the co-founder of that ecosystem of free and open source tools).

[Eesger]

@smathermather: I like!

I can’t completely fathom your ODM project… (like the first letter a lot ) Is there a part what I can use where it can “read” a set of images and “spit out” the OpenSfM corrected lat/lon? (is it linux based?)

[smathermather]

Is there a part what I can use where it can “read” a set of images and “spit out” the OpenSfM corrected lat/lon?

Yes.

It is linux-based, although there is a Windows installation process here for running it on Windows via docker:
https://docs.opendronemap.org/installation.html#windows
and here for Mac OS:
docs.opendronemap.org/installation.html#macos

You can also get free credits to try it without installing by signing up at webodm.net (I do not get compensated in any way by webodm.net, but it is run by my co-founder, and he uses it to fund development on the project).

The process

Feed in images, get a bunch of stuff out. Among the products it creates, it generates a file called shots.geojson which contains the corrected coordinates.

I am currently re-running some of my 360 image captures through to see what we get from images.geojson, as this is a newer feature which precedes my 360 camera tests.

[chelseabrian]

This gps “normalization” is very interesting to me. Wow, I have so much to learn :).

I took a quick try at running some of my panos through webodm but the process failed with the following error message - The program could not process this dataset using the current settings. Check that the images have enough overlap, that there are enough recognizable features and that the images are in focus. You could also try to increase the --min-num-features parameter.The program will now exit.

These panos are spaced approximately 3-4 meters apart. Maybe this is too far between images? I Captured with Ricoh Theta Z1. I will revisit the features parameter settings to see if that helps.

[Eesger]

@smathermather

You can also get free credits to try it without installing by signing up

I have started work on an other project “OpenWanderer” (there is that word “open” again )
Maybe we can work together? If payment is required then i/we will need to figure things out on our own (would much rather work togehter though)

[smathermather]

Ah, sorry: set your camera-lens to spherical. 3-4 meters should be ok, depending on where you are at and what features are available to match.

Also, I have had the best luck two sequences side-by-side. A single sequence should be possible, but I haven’t enough testing to recommend good settings.

[smathermather]

For most small projects the way to go with OpenDroneMap is to just install on a PC. It’s a fully FOSS project. If you need more compute power, WebODM.net is a very affordable option (it’s often more affordable than hosting on your own cloud instance).

In short, I am happy to help however I can to get you up and running, but I don’t control credits on webodm.net .

I think for most 360 camera projects, getting it running on a modest desktop machine is more than adequate.

[Eesger]

… when I find the time I’ll start some testing (I will have to be very cautious though… one sequence is easily 500 images, so for testing I’ll need to limit it to about 10 or so…)