Connecting rooms through doorways + Inspection tool

Answered

Comments

41 comments

  • Avatar
    Michal Rus

    Hi,

    are doorways open ? No doors ? If so, you can use common CPs that help connect both datasets (rooms). Just make sure the same CP is visible on at least 3 images from both datasets.

    Hope it helps.

  • Avatar
    Tom Foster

    Is this saying that placing CPs is the necessary way to link rooms through doorways?

    Is it not do-able simply by the right kind of linking photos, so the two rooms' photos 'automatically' register into a single Component, if necessary using the tweaks in the above link (which I do anyway)?

    I'm up for making two separate models of the two rooms, then 'Merge Components only' - but is Michal saying it has to be done by placing CPs?

    I really want to believe that CPs are a last resort, and should not be necessary given good photo set(s).

     

  • Avatar
    Michal Rus

    CPs will help for sure if you have problems with normal overlap workflow. It's not requirement however. Depends on your capturing workflow and situation.

    You can check also this thread https://support.capturingreality.com/hc/en-us/community/posts/360000096131-Workflow-Process-Interior-Image-Alignment-

     

  • Avatar
    Tom Foster

    I'd forgotten that one - v gd.

  • Avatar
    Götz Echtenacher

    Bottom line is you need to treat the door like any other connection of two loops. After all, the door is just a comparatively narrow part of the whole model...  :-)

  • Avatar
    Tom Foster

    Sorry Gotz, you've prob explained but I'm not sure what you mean by 'connection of two loops' - can you shortcut me please, with particular reference to this connecting doorway situation. Much obliged.

    Having 'tiled' my photos with two sets of shots more or less perp to both faces of the wall i.e. ea set in a different, adjoining room, I have to make connecting photos through the doorway, so RC can register both photo sets together, into a single Component, whether by Aligning all in a single model, or by Aligning in two separate models (with the connecting photos added to both photo sets) and then 'Merge Components only'.

    So the connecting photos have to both transition from longer to closer-range shots, and also transition through 180o in less than 30o steps. That means a lot of shots, and experience tells me that very systematic and careful shooting to a formula or pattern that RC likes, is by far quickest, with least shots, and reliably effective.

    It's just that I haven't found that pattern or formula yet, for the connecting doorway situation, either by thinking it through geometrically, or by experimentation. What I have ought to work, by the thinking-through approach - but it doesn't.

    I'm wondering if, with flash, a deep shadow is being cast on the near part of one room or the other, which makes photos that ought to be common to both, look v different to RC. About to revert to long shutter on tripod, without flash, to get rid of such shadow.

  • Avatar
    Götz Echtenacher

    Hi Tom,

    you basically answer the question by yourself. You need to give RC enough images to slowly transition. As for the best way to do that, everybody has to figure that out for themselves - it depends on so many things. Shadows from a flash is certainly a possibility and switching it on and off as needed probably the best way to handle it.

    What I realized only not too long ago is that RC can work best with surfaces that are close to parallel to the sensor plane (pointing the camera orthogonally at them), hugely angled surfaces ( more than 20-30°)are very difficult because the features change magnitudes more than with parallel surfaces.

  • Avatar
    Tom Foster

    So Gotz can I ask you directly -

    Do you agree, at a doorway, a 180o transition has to be made, between the two photo sets of the two room-faces of the same wall?

    So, after transitioning in to say one jamb (vertical side) of the door frame, photograph it at 0o (looking perpendicular to the wall face), 30o, 60o, walking yourself through the doorway, 90o (now looking along the length of the wall), 120o, 150o, 180o (now looking perpendicular to the other face of same wall)?

    And/or do same up under the head (top) of the door frame?

    Or is that much too complicated?

  • Avatar
    Götz Echtenacher

    Hi Tom,

    depends a bit on the object, but in general I would say too complicated and also not overly accurate because the whole transition will revolve around narrow surfaces that on top of everything are often painted with a glossy finish.

    I only did it once ages ago so that I could quickly document a few rooms - there was not enough time for taking measurements before it was demolished.

    You can see how I move forward through the door (I did it in the other direction as well) and also slightly turn to the left (because that was my area of interest). But the main point is that I use the background that I see through the door and the wall around the door as the faces for transition. And even though the shoot war very unsystematic (as you can see), the walls come out pretty good, as in they are all equally thick (or thin). So that seems to work.

    At the top however, it did not work as well because there is a long dark hallway with fewer features, so I had to add a few CPs to help it along.

    If the walls are too featureless, maybe the ceiling would work as well? It's the same distance to the camera all the way through the door and nothing is obstructing the view as would be the case when using the floor...

  • Avatar
    Tom Foster

    I think I understand ... so not concentrating particularly on the doorway/frame itself - instead, as you say, on what can be seen beyond, going in both directions.

    Because fragments of the far wall of Room A can be seen through the doorway while photographing Room B, and vice versa. A revelation - of course.

    That means fully photgraphing all walls of both rooms as well as the walk through the doorway, before Aligning. Plus, before walking through, taking skew views from Room A through the doorway into Room B, to left and right, to the max. Neat.

    Not sure I understand your alternative suggestion to link via the ceiling?

    Thank you very much for pulling these pics from your archive, complete with Inspect view.

    BTW, in Inspect - you still say that dark blue means less connections, light blue/green more, yellow/orange more again, cerise, then red the most connections?

    Finally, how did you get that plan view, without all the ceiling etc blocking the view? I haven't found how to get 'inside' a hollow image like a room. Ortho projection - is that a static, exported image, or something you can move around in?

  • Avatar
    Götz Echtenacher

    Hey Tom, no problem. As long as you ask concise and intriguing questions, I am happy to share my thoughts. You've tought me a few things as well!  :-)

    Yes, angled shots are also very important. I always think of spinning a web.

    The ceiling could be used as a surface to transition from one room to the other, it is in most cases much closer than the opposite walls and at a similar distance in bot rooms. I imagine pointing the camera at the ceiling and then walking through the door, paying attention that the remaining part of the wall above the lintel does not obscure too much or throws shadows.

    Yes, the color coding has been explained a few times by Devs and I thought it is also in the help - but then I am sure you would have noticed it.

    I did no trick at all, the ceiling and floor are just not very well covered and so the walls stand out since the tie point density is much higher.

  • Avatar
    Tom Foster

    "I imagine pointing the camera at the ceiling and then walking through the door, paying attention that the remaining part of the wall above the lintel does not obscure too much or throws shadows."

    Unfortunately, if using flash, while still in Room A, I've found that's exactly what does throw deep shadow onto the nearest visible bits of Room B - and as you pass under the door head into Room B, the shadow switches onto nearest visible bits of Room A. So the same features look maybe unrecognisably different.

    With a single flash unit, that's true whether the flash is on or off camera - just depends which room it's in. Unfortunately I don't at present have off-camera flash facility. Maybe with two flash units one in ea room. Or single off-camera flash unit exactly on the door threshold looking up, if the wall has little thickness.

    Still, will try again.

    If, in the first idea, relying on far wall seen through the doorway, need to well iluminate both rooms at same time? Two flash units again.

  • Avatar
    Götz Echtenacher

    Connection shots don't need to be perfect - just good enough. So maybe no flash for those?

    As for he first option, I only used my tiny on-board flash and it did the trick...

  • Avatar
    Tom Foster

    Blimey, I look forward to finding out how to make it as easy as you do!

  • Avatar
    Götz Echtenacher

    Maybe the trick was that I didn't have too much time to think?  :-)

    It took only about 20 minutes or so...

  • Avatar
    tingelbobber

    I also noticed to build a path of images for merging "components" is easier then using the CP stuff, also more stable. I learned a lot by the good old try and error method. 

  • Avatar
    Tom Foster

    "Maybe the trick was that I didn't have too much time to think?"

    It's only people who have learned the trade the hard way (like Picasso) who get to say that.

  • Avatar
    Tom Foster

    "to build a path of images for merging "components" is easier ... also more stable. I learned a lot by the good old try and error method."

    Exactly. It's a matter of getting almost muscle-memory, so you get it right first time without having to go back for excessive folder-bloating extra shots. I hope.

  • Avatar
    Götz Echtenacher

    Tom, I am flattered!  :-)

    But that was still early-ish on in my learning process.

    Also a bit of luck maybe?

    Plus nicely textured walls, the white areas are almost devoid of tie points - which is causing most of your difficulties I believe...

  • Avatar
    Tom Foster

    Wow looka that - first time I've effortlesly got a result like a veteran! Well, after doubling the no of photos by blasting the corners where the Components wouldn't join up. Though I bet if I deleted that Component and Aligned again it wd go back to 2 or 3 Components.

    That's one room done, all walls - now the other one, then the look-both-ways through the doorway.

  • Avatar
    Tom Foster

    I said to Gotz: "in Inspect - you still say that dark blue means less connections, light blue/green more, yellow/orange more again, cerise, then red the most connections?"

    Gotz answered: "Yes, the color coding has been explained a few times by Devs and I thought it is also in the help"

    which leaves me puzzled - 'blue = weak, red = best' doesn't agree with what happens when I tweak Inspect settings. So I'm collecting votes - which way it really is - and it's contradictory.

    In https://support.capturingreality.com/hc/en-us/articles/115001552532-Quality-assessment : "A web of blue lines should appear. The colour scale of the Inspection tool is blue to red. The more red, the worse. On the other hand, the closer to blue colour, the better."

    Zuzanna in https://support.capturingreality.com/hc/en-us/community/posts/360001888972-Border-between-areas-with-differing-distances-of-the-cameras-to-the-object says: "The edge color is blue-to-red (Jet color scale) indicating how strong an edge is - actually more features, more red."

    Not trying to smart-arse - just want to get things clear, so's to actually intelligently use the huge amount of precise data that RC delivers - if only they'd explain (better than in Help) what things mean.

  • Avatar
    Michal Rus

    Blue is best connection, red is worst. Anything between is average.

  • Avatar
    Tom Foster

    Thank you greatly.

  • Avatar
    Götz Echtenacher

    Sorry to disagree, but in the help it clearly states:

    "The edge color is blue-to-red indicating how strong an edge is - more features, more red."

    More features is commonly associated with better.  :-)

     

  • Avatar
    Tom Foster

    RC staff, if you're reading this, there seems to be a contradictory/inconsistency in what RC people publish. Evenly divided.

    e.g. Help contradicts https://support.capturingreality.com/hc/en-us/articles/115001552532-Quality-assessment .

    What is the double-checked true answer?

  • Avatar
    Götz Echtenacher

    Tom, I think that either the content in the you provided is wrong, or I really miss the point entirely.

    Just try it out. If you set the Maximal Matches to, say, 1000, all of a sudden you have plenty of red lines and more cyan, green and yellow as well. The value defines the saturation point for red, 1000 tie points between two images. Play around with the number and you'll see. Blue is the only colour that connects images far apart, which stronlgy indicates that those are weaker connections.

    What I haven't figured out is the difference between Matches Count and Minimal Matches. Maybe somebody is willing to shed some light on that. Is Matches count basically a filter for images that have fewer overall tie points than the threshold?

  • Avatar
    Zuzana Ďuríčková CR

    Hello guys,

    Thank you very much for bringing up this issue. It is the best way how we can find out that we need to make the help section clearer about this feature. We are sorry for the misleading information and inconsistency.

    The full and correct extent of answer is as stated in the RC App Help. „The edge color is blue-to-red indicating how strong an edge is - more features, more red.“ is correct. If the edge is red, it means that it is the strongest one.

    We will correct the information in the Quality assessment tutorial and we will also try to improve the description of this tool in the application Help.

    To answer the Gӧtz’s question about the difference between Minimal matches and Matches count, this is how it should work:

    Matches count and Component connectivity define how the cameras are connected into virtual components.

    Component connectivity –from Help section – “Every two different components would have less (component connectivity) edges in-between.” If there are at least this number edges between cameras, then those cameras are connected into one component (two components are not created).

    Matches count – two cameras are connected into one component only if there are at least this number of common features (matches) among those two cameras. Those features however have to suffice the condition defined by apical angle and Feature consistency.

    Minimal matches – edges with the number of features lower that this value are not displayed at all. Edge with this number of features is displayed with dark blue color.

    Maximal matches – if set to zero, all edges are displayed. The edge with the highest number of features is displayed with dark red color. If you set the specific value, then it is the greatest number of connections between two images. Color scale is recomputed accordingly to minimal and maximal matches.

    The most important thing to check with Inspect tool is the number of created virtual components with the component connectivity set to at least 4. This means that if the images are connected into one virtual component with the component connectivity set to at least 4 (that is the reason why the default value is 4), the alignment can be considered as stable and no cracks or misalignments should appear. If there are several virtual components created than the connection among the cameras from these separate components is not that stable and we would recommend to add images in these areas.

  • Avatar
    Götz Echtenacher

    Hi Zuzana,

    thanks a lot for the extensive answer!

    I read the help section today of course...   :-)

    I think I got it now. Matches Count defines the basis for the calculation and Minimal Matches caps only the display of the result. So if Matches Count is set to 50 but Minimal Matches remains 100, then the display won't change.

    Concerning Maximal Count: I think it's important to mention that the value will not result in fewer lines, just move the color threshold for red down to this number. That means all image pairs with more than e.g. 1000 matches will also be red.

    So Connectivity means how many "partners" each camera has?

  • Avatar
    Götz Echtenacher

    Also, is it correct to assume that if Maximal Count is left blank, red will be the connection with the most tie points in this model?

Please sign in to leave a comment.