Shop OBEX P1 Docs P2 Docs Learn Events
What are applications for wide range / dynamic 3D measurement? — Parallax Forums

What are applications for wide range / dynamic 3D measurement?

LindaLinda Posts: 1
edited 2013-07-16 05:14 in General Discussion
What are applications for wide range / dynamic 3D measurement?




We are currently involved in a new technology that captures 3D environments in the course of a university project. In particular, the technology offers the following benefits:

- Generates 3D models
- Captures large areas of up to 32x32ft (or 10x10m)
- High precision of 0.03" (or 1mm)
- Motion detection in real-time with up to 1000fps (Response time: 1ms)
- Independent of light (works in both bright and dark environments)
- Quite portable (0.8"x0.8") with low energy consumption
- Captures an angle of 180°; when combining two modules 360° would be possible.

What application fields / problems could be relevant that would require operating in such large areas, at given precision and time resolution? We would be highly interested in applications or problems with focus on a rather large group of users.

Thanks in advance for your support!

Comments

  • DeathbatDeathbat Posts: 4
    edited 2013-06-27 14:48
    We use similar equipment call LiDAR all the time to scan rooms or roadways. Basically before we do a civil or architectural design we need to know where things are. The scans give us point clouds which we take into CAD to produce working models. While that type of application does not require real-time data, I know there are many telepresence systems that could benefit. Even for the more static need like ours we are always looking for lighter, faster, cheaper systems.
  • prof_brainoprof_braino Posts: 4,313
    edited 2013-06-29 11:33
    We use the equivalent of a professional surveyor's laser. This is accurate to a few centimeters up to a 350 meters. We are working on an openCV (open computer vision) module to capture scene input, and use these as textures in a 3d video game engine. Combining the two and some others might allow us to dynamically generate accurate virtual maps. We have not attempted to identify and track transient features (people cats) yet.

    We also use multiple units with simple ultrasonic range finders (SR04). These are very short range and tend to get noisy readings, but the solid readings can be identified and saved, adn the noisy measurements discarded.

    In both cases, there can be interference when using multiple units. One task is to take the interference and use it as an improved complex sensor. That is, using a pulse from one unit and detecting it on multiple units, and combining the data.
  • vanmunchvanmunch Posts: 568
    edited 2013-06-29 21:51
    Am I reading this correctly? You're saying that the entire sensor is less that a square inch!?!? You mention 1000 fps, is that for a single point or the entire field of view? Regardless, I think it would be a real winner in anything flying, especially autonomous. How much does it cost and how do I connect it to my propeller chip?
  • Heater.Heater. Posts: 21,230
    edited 2013-07-15 06:19
    Is this spam or what?

    I mean, a 3D scanner that generates 3D models, over a 100 square meters, at 1mm resolution and 1000 times per second, all in a 1 inch cube sounds totally outrageous.

    Consider moving a flat panel of 1m by 1m a distance of 1m over one second in the scanned area.

    To represent the panel at 1mm resolution we have a million data points. Every ms we are going to have a million new data points. Do that for one second and we have a billion (10 to the 9) data points. Each on of those is at least 3 coordinates of 16 bits. So that 6 Giga Bytes per second output from this device. Which seems pretty unlikely.

    What am I missing here?

    If it exists and is practical to use there are a billion uses for it. The company I work for would certainly be interested.

    Given that the university in question is not named and their are no links to the technology in question I don't know what to make of this.
  • rjo__rjo__ Posts: 2,114
    edited 2013-07-15 17:47
    Given the fact that Linda has 1 post and hasn't responded. I'm not sure what to think.

    It seems that today, specs seem to require a context, which is somehow missing.
    if you look at the specs claimed for Kinect 3D, it is clear that specs don't always say what we think they
    do.

    Let's assume Linda is real... then let's look at the spec and consider them... not as a group but individually.

    Generates 3D models. OK... Lot's of things do this. Linda's thing can do this too.

    Captures areas up to 32ft x 32 ft... OK, why argue?

    High Precision... 1mm. This is where the devil is in the details... does it mean that 1 mm is the greatest resolution?... probably.
    If so, does this apply to 32ft X 32 ft... probably not.

    Motion detection in real-time up to 1000 fps... this is possible. And if we add adaptive optics, we get 3D... but not at 1000 FPS... 1000FPS is for motion detection, good. How long does 3D take... probably depends on the resolution you want.

    Portable with low energy consumption... ok ok... it is a camera head with adaptive optics...with or without prismatic distortion.
    Captures 180degrees... really? with a single lens or using adaptive optics? Pretty hard, but not impossible.

    So, it really isn't impossible. And if it is coming from Vienna, it is probably this place https://maps.google.com/maps?client=safari&oe=UTF-8&ie=UTF-8&q=Vienna+University+of+Technology&fb=1&gl=us&hq=Vienna+University+of+Technology&hnear=Vienna+University+of+Technology&cid=0,0,2204850258396665874&ei=FZbkUbPPCoTFyAHUmoGwCg&ved=0CJQBEPwSMAI

    Linda, the applications depend upon the exact specifications. We all want one. Please come back:)

    Rich
  • frank freedmanfrank freedman Posts: 1,983
    edited 2013-07-15 22:05
    Personally, I would like to see the compression algorithm that can make this practical. even at a 256 greyscale 1byte per pixel has anyone considered the storage required for a cube 32' on a side in esentially 1mm cubes? Uhm, dats alota data......?
  • Heater.Heater. Posts: 21,230
    edited 2013-07-15 22:39
    Yeah, I did, see post #6.

    One is not going to be storing one byte per 1mm pixel over a volume of 10 by 10 by 10 meters. Actually I think such a 3D pixel is called a voxel.

    After all when looking at a 3D scene one can only see the surfaces of things. That is, we are not doing a deep scan like an MRI scanner. Hence my talking about that 1 meter square plate moving within the volume. Still that's a lot of data to collect and crunch especially at 1000 frame per second.

    Of course, it does say "builds 3D models" so when it come to storage you are storing a model not the raw voxel data. That 1 meter square plate might become only four points and four vertices in space.

    We don't have enough info on this thing to make any headway.
  • rjo__rjo__ Posts: 2,114
    edited 2013-07-16 05:14
    http://en.wikipedia.org/wiki/Kinect

    In "public" announcements, the Kinect resolution wasn't stated completely either. And it is made by a fairly large company.

    The claimed 1.3 mm resolution of the Kinect is stated for a minimum object distance. Obviously, at the maximum imaging distance the resolution is far less,
    In public announcements the part about "minimum object distance" was simply not mentioned.

    I think this is a smart camera with smart optics, using large pixels and a hardware point spread function for 3D.

    Linda please come back:)

    Rich
Sign In or Register to comment.