The idea is mainly focused on a very low-cost pattern recognition, distance measurement and coords. streaming to an existing Revit plugin in order to have them collected easily with decent precision.
Nice to hear from you! I will be much more detailed in future contacts, if you wish, as soon as I'll be back to my University from some conferences duties. Thank you for your interest.
Nice talking to you! In my institute summer vacation starts after 10 days. I will be considerably free! We can work on the issue raised by you.
I'm interested but I can't, So I leave a few recommendation for you.
If you're using windows platform, I think OpenNi driver is easy to use compared to amount of classes. And Kinfu can be used almost directly for your purpose. You see, I have seen that this problem is partially solved so you don't need to work on it much, more likely you have to use some constrains to simplify matter but using Kinfu for registration, gives you pretty most of work.
If you want to develop program from scratch, then using OpenCV and feature matching for registration is better or more precise. Pattern recognition for registration, I'm not sure. But to stream it to revit, is a pretty good idea but might render all or part of my previous comment useless.
This sounds very interesting. I have been working on methods to validate as-built conditions against Revit/Bentley BIM models. I think Kinect would definitely be the tool of choice for that project and there are many examples in the DIY community of generating 3D geometry from Kinect data.
The idea is cool. I have been working on project to integrate BIM models and Virtual Environment. The use of kinect camera would be interesting. Please keep me posted on the up date. Thanks
I'm not sure the level of accuracy that Kinect delivers is adequate for the kind of validation I would like to accomplish though. The Kinect models I have seen are blocky representations of the original scanned object without much detail. That's why in my project I have been working on my own methodologies towards that end and have been using the Arduino integrated with commercially available CMOS cameras. I would be very interested to know if you're able to achieve the accuracy you require from the Kinect.
@Scott Dennis; If you mean precision by accuracy, then you're probably aware of the fact that its value is a function of distance to object. and x,y dimensions have more precision than z. Precision itself can be considered a polynomial of second degree by variable of distance. (2m distance, 3mm (in x,y) & 10mm (in z))
But if you talk about accuracy itself; then that's not a problem. If it shows that distance is 2m then it's 2m+-10mm.
And You can stay for release of new Kinect in which precision has dramatically increased by development of new technologies.
@Mahmood; any cameras, even HD, have limitations given the increase in area covered by a single pixel as distance increases. My goal is to generalize photographed volumes in such a way as to relate them to features of the BIM model with a level of accuracy sufficient to validate conditions quickly and identify areas where risks to later phases of construction exist. That's the one downfall of all BIM modeling, construction does not always provide adequate feedback to identify risks to later construction phases. In some cases the BIM model is rendered almost useless for facilities management purposes once construction is completed.
@ Scott . Actually for your purpose I'm sure that even next Kinect is useless.
And decrease of precision by increase in distance can differ in value but as you said is always there. (the decrease in Kinect range data is so much that in 6 meters distance it will be less than half a meter). Risk determination in construction seems interesting!
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Fixatives