Leica RTC360 Reality Capture Solution

[Felix_the_Cat]

[/quote]

There is also another small detail: with the HDR colors always have a hue ... pastel .. not really real ... but it's a detail

Piercarlo[/quote]

It is not a small detail! This is huge because when you combine files from 360 with other programs they jump out like clown cars on the screen.

I have even fooled with the image editor and it's weird clunky interface to no avail. This really needs to be fixed



Sent from my iPhone using Tapatalk

[jamesworrell]

Obviously with images the scan files increase considerably and the import times grows accordingly... and if you're scanning on medium or high, the file size grows accordingly.

Ideally would love to see most of our scans at low res with colour .. but when you have the massive HDR colour sets you start wanting to drop the number of stations - so we head back up to medium and get them further apart.

We want colour. If we didn't want colour I would have bought a P40 .. but I could give up the 4 additional sets of colour. I agree - it is a nice option to have .. but key word there is option ..

The other motivator for medium res is the distance to targets .. not sure I want to head down the faro route where we can window scan hi-res targets - but for really longer shots that might be a great tool to have (99% of our jobs will involve survey control) .. also gives me an excuse to keep the ipad - two links per station didn't really do it for me - I was hoping on defining more links to be honest .. closing loops .. defining decent bundles in rooms/areas etc .. two links is kind of a waste of time - VIS pretty much does that far you anyway. An operator in the field has a clear understanding of overlaps - so doing it in the field made more sense to me. Auto-cloud is always dangerous .. I have had plenty of disasters - so I tend not to do it .. so an office operator has to go through the dataset looking for overlap and defining the links. And on that point - Register 360 - defining links from the cloud viewer instead of going back to the sitemap? But we won't be in R360 any time soon - target heights?!?!

Not sure if this is happening behind the scenes or not - but perhaps instead of HDR-full res .. maybe HDR low-res to perhaps more closely match the scan resolution. If we are doing low-res scans 4 - 5m apart with 12mm "pixels" .. why have images with pixels that match the 3mm scans? So maybe sub-sampling the imagery to correspond to grid size plus a bit? I haven't even tried to work out the "GSD" of the images as it is - although to be honest they don't seem _that_ high ..

[pburrows145]

Have you changed the pano import option to low res in REG360? It's under General->Import

I believe this overrides all pano images imported to the lowest/medium/highest res however, rather than choosing L/M/H relative to the scan resolution: is this what you'd prefer?

If scanned in X resolution give me X resolution imagery
If scanned in Y resolution give me Y resolution imagery
If scanned in Z resolution give me Z resolution imagery

[jamesworrell]

The joys of dealing with users eh? No doubt everyone has a different opinion/idea .. to be fair just about every scan I do, I go through a "checklist" of what are the best settings/placements etc .. so not only is every job different (we have clients to satisfy too) .. nearly every scan could throw something up. I'm just chucking ideas around as we play with the new toy.

Scenarios to suit I guess:

1) longer range scans/hi-res .. you would want highest res imagery to match - you need to colour the points and no doubt the required "GSD" is pretty high .. looks like you have 432 megapixels .. vs 170 million points.
2) low res .. same 432 megapixels for only 10 million points .. chances are if you can't read something .. choose a closer scan? I don't know ..
3) medium for completeness .. 432 vs 42 million .. 10 times the pixel-to-points count.

At 700mb for the imagery .. say 4 pixels per point .. to get to low res - want 40 million pixels? 700mb -> 75mb? But what does a 40 megapixel spherical look like? Is it worth it? Certainly would seem enough to colour the points .. but maybe too far of a loss in terms of image quality?

I know a set of low-res scans I did the other day I was running at about 4-5m station-to-station spacing .. so it isn't like the imagery has to go too far in that scenario. The medium res job Tuesday - we were aiming for about 10-12m centres. I need to do some resolution versus distance to target tests - thought Mike Harvey had a youtube video somewhere - might have been BLK though. I guess the extreme example is you are doing a scan in a cupboard and you get 700mb of imagery for a 1m x 1m box - but if you were doing a network rack - you might want to read the labels .. you can't win.

How does the import setting relate to the 432mpix? 432mpix becomes ..?

For a different scenario, in the meshing world - a low res scan textured with hi-res imagery is probably more the go .. just for a spanner.

[gsisman]

Have you changed the pano import option to low res in REG360? It's under General->Import

I believe this overrides all pano images imported to the lowest/medium/highest res however, rather than choosing L/M/H relative to the scan resolution: is this what you'd prefer?

If scanned in X resolution give me X resolution imagery
If scanned in Y resolution give me Y resolution imagery
If scanned in Z resolution give me Z resolution imagery

This would be a great option Paul. I think that is what the guy in the previous comment was indicating. Then you can mix and match based on whether you need very high imagery for visualization or less imagery need and more scanning density for modeling, etc.
BTW , I'm sure they need the highest resolution HDR for the VIS technology which is why it must always be on before you are moving with VIS.
It's very possible that it is using the VERY HIGH Resolution 360 dome static pictures to be the control basis for the VIS going forward as you move-tracking pixels resection. If it could do this and internally and then give the option at Software import on how you want to deal with that imagery it would be better.
It could still use the Hi-Res even to do the VIS Poses analytics at Reg 360 import (if you're not using Field 360)-without then importing the actual High res imagery into the Reg 360/Cyclone DB to reduce the size of the additional files. All the hi res data could still be left in the RAW files IMHO.

[steves01x]

Slightly off topic - but why such a need for capturing imagery? The intensity data is up there with the P series.

I would rather avoid it but the jobs that do need it, i-star (assuming its not crashing) allows for quick capture and fairly easy import to Cyclone , avoiding any overly large files, not slowing up in the field too much or clogging the scanner up.

If you are needing very high resolution imagery then an external camera kit or a separate photogrammetry job is surely a better option both with full control over the camera etc? Or with the new LGS being able to insert tagged images where a photo is required?

Anyway, although not used the RTC yet if the size of the HDR imagery is the only issues with the work flow then its not too bad an issue until you hit a 500+ scan job in full colour.

Given some of the above tasks can easily take 24 hours + it is not realistic to use Register 360 - we can't afford to tie licenses up that long.

It would be rather handy if Cyclone allowed you to import Leica scan data with out a license and then just be able to open/view each scan that has been imported. Export is a pain as until you can batch export individual scans with their original names then its a fairly manual task and not an automated over nighter.

[jamesworrell]

Slightly off topic - but why such a need for capturing imagery? The intensity data is up there with the P series.

Clients certainly drive it .. coloured point clouds that is .. for architectural renders etc. For modelling it is handy .. eg undercover car parks - red pipes = fire - sometimes I can't work out what something is from intensity - flicking to colour will give it away .. we are also starting to get asked to model material and colours .. so need to capture.

In the not too distant future I hope Leica embrace meshes more too .. would love to see Jetstream/LGS/Cloudworx handle LOD meshes - so meshing a colour point cloud would produce a fairly awesome output given how many scans you can do and how clean the data is from the rtc. I haven't tried to mesh one yet - probably should through a few at Context and see how they come up.

Also to a point made from a previous member - the tone of the HDR imagery isn't quite there yet - I would kind of prefer myself that any colour is coming from a normally exposed image and not necessarily a blend .. I also find the tone map a bit dark in Cyclone. Not sure there is an automatic way to adjust? I am not going to tweak hundreds of scans independently.

[jamesworrell]

After a few days of considering this .. may I suggest a good first step might be to introduce an option for cloud-matched/imagery down-sampling?

I suspsect this option might be easier on your software stack guys who are probably expecting full HDR .. only the resolution is changing?

So an extra switch for cloud-matched versus full-bore imagery would be welcomed .. gets me coloured point cloud and drastically helps storage at least and perhaps processing.

I am sitting here watching 93 scans going in .. it is up to scan 45 and it has been going a good few hours now - I have another 200 scans to go for this job with more coming in from the field tonight .. another ~ 100 the day after .. at 432 mpixels per station to deal with .. it is kind of adding up - although the last 100 are the roof - so no imagery there - not worth it.

Rough numbers:

~ 169 million points high res = 432 mpix HDR .. manual says 180 million pixels to colour .. 2.4 pixels per point
~ 42 million points med res = ~ 100 mpix HDR
~ 10 million points low res = ~ 24 mpix HDR

Typical will be 1/4 of initial data size and even up to 6% of initial data size ... pretty big change in bytes ..

And ideally on-board as well .. helps USB usage .. helps download times .. helps archival .. the FPGA(s) isn't that busy anyway is it??? ;-p

[topogeo]

Look at the images taken from an RTC360 (lgs file in Jeatstream viewer) .. (scan 6mm / 10 mt .. but it seems to me that the image quality is always the same ... regardless of the resolution of the points ..)
Are you sure you can accept a lower resolution?
Already now the images are of fairly poor quality ... (a BLK360 makes them perhaps better).

I think the color is very useful and ,, which is very popular with customers ..
We got used to the Google Maps Stret View photos .. we now have to accept a lower quality ..
The quality (resolution) of the images can not be reduced

The photos are not just for coloring points .. but above all to see the panoramic image of the area (with the viewer of Jeat stream ..)

I think the best way is the HDR option yes or no ...

Piercarlo

[jamesworrell]

I think those images are a little unfair .. you are a LONG way to the surfaces .. any camera with a wide angle lens system would struggle with that .. try your iphone for example and zoom in. If you are after window frame details from 50m you are buying a Phase One and doing gigapixels per station.

If you don't have the resolution - find a scan/station that is closer ..

What would be the average station-to-surface distance you think would be fair? And how big is the object you want to see? Something 10m away and maybe 5mm in size per pixel? Read a sign from 10m? Even streetview has an image every few metres.

5mm pixels at 10m is not far off what I am proposing for the low res capture at 2.4 pixels per point ..

If you want details on a wall 5m away - go the cloud-matched imagery .. if you can't get closer - go full bore.

I do agree though .. cloud matched vs full bore AND HDR vs non-HDR would be good .. but imagine there is a bit of re-engineering in that.

I haven't quite worked out what the spec sheet/manual mean either .. is it 432 mpix per station plus 4 more brackets? Or is it 432 mpix with the 5 brackets .. so 432/5? And it uses 180 mpix to colour the 3mm cloud? IS the full 432mpix coming through to jetstream et al? It talks of 4k cube maps - so perhaps not full res going all the way through? In which case - why bother taking it in the first place ..