Kinect 3D-scanner - Karelia-amk

Download Report

Transcript Kinect 3D-scanner - Karelia-amk

1
Miklas & Marjolein
Zsolt
2




Introduction
Kinect?
SA3D-scanner
You want to try at home?
3


Introduction
Kinect?
◦ What is it?
◦ Possibilities


SA3D-scanner
You want to try at home?
4

What is Kinect?
◦ How do normal people see it.
◦ How do IT people see it.

Capabilities
◦ Natural User Interface
 Color, depth, skeleton, player index
◦ Audio API

System Requirements






The Idea of 3D scanning
Getting a point cloud
Coloring the point cloud
Registering point clouds
Creating a 3D model
Why is the Kinect the best method to scan in
3D?



Kinect is a motion sensing input device by
Microsoft for the Xbox 360 game console.
It’s a new generation game controller like Wii
Remote Plus or Playstation Move.
Enables users to controll and interact with the
Xbox 360 using gestures and spoken
commands.

Kinect is a physical device that contains
cameras and microphone array.
RGB Camera
Multi-array
microphones
three-axis
acceleromet
er
3D Depth
Sensors
Motorized tilt


Viewing angle: 43° vertical by 57° horizontal
field of view.
Mechanized tilt range (vertical) +/- 27°.



Resolution: VGA (640x480)
Frame rate: 30 frames per second
RGB or YUV color
The Kinect uses a USB connection to pass
data, and that connection provides a limited
amount of bandwidth.





IR light source – CMOS IR sensor
Each pixel contains the Cartesian distance
(in millimeters) from the camera plane to
the nearest object at that particular x and y
coordinate.
Pixel value 0 means that no object
found.
Does not work properly on
reflecting surfaces.
640x480, 320x240, 80x60



Audio format: 16-kHz, 16-bit mono pulse
code modulation
24 bit analog-to-digital converter
An array of four microphones arranged in an
L-shaped pattern.
◦ Improved audio quality
 Noise suppression
 Acoustic echo cancellation
◦ Beamforming and source localization



The Kinect can indentify up to six human
figures and create a player segmentation
map.
The player data comes in the depth stream.
Up to two players, the Kinect can return a
complete skeletal information.

Hardware requirements:
◦ Computer with the following capabilities:





32-bit (x86) or 64-bit (x64) processors
Dual-core, 2.66-GHz or faster processor
USB 2.0 bus dedicated to the Kinect Sensor
2 GB of RAM
Graphics card that supports DirectX 9.0c
◦ A Kinect Sensor

Software Requirements
◦ Supported operating systems:
 Windows 7
 Windows Embedded Standard 7
◦ MS Visual studio 2010
◦ Kinect SDK
Other third party SDK-s or development
systems could also be used (for example:
OpenNI)
3D Scanning with the Kinect


The goal is to create a colorfull 3D model of
the reality.
Later this modell can be used in
◦ Entertaining softwares (computer games)
◦ Industrial applications




With the Kinect’s depth stream, you can get
the distance of each pixel from the cameras’s
plane.
With the Kinect SDK, it’s possibily to convert
this data to actual XYZ points.
The resolution is 320x240.
The depth sensor is IR, it volt work on
◦ too close or too far objects.
◦ glass or reflecting surface.
◦ direct sunlight.



We can get the colors for each pixel from the
color stream.
The SDK is capable of matching the color and
the depth stream, even if they are not at the
same resolution.
Once we get the color data, we can color each
point in our point cloud.



With the use of PCL (Point cloud library) we
should be able to register point clouds
This way, we could put together scans taken
from different directions and aim for a full 3D
scan.
This feature has not been made yet.


There are numerus ways to create a 3D model
from a point cloud.
Our goal is to scan the object and save it.
Since almost every 3D modelling program can
handle point clouds, we save our scan that
way.



It’s not.
Other industrual 3D scanners may work with
better resolution, less nosie and all in all,
better performance.
The Kinect is easy to use, cheap and therefore
available for everyone.



Introduction
Kinect?
SA3D-scanner
◦ What and how?
◦ Demo
◦ What’s next?

You want to try at home?
35
SA3D-scanner
36

Kinect (xbox.com)

OpenGL (OpenGL.org)

MeshLab (meshlab.sourceforge.net)

PointCloudLibrary (pointclouds.org)
37
What’s next?
38




Printing stuff
Use in games
Use in simulators
Use in architecture applications
39
NextEngine 3D Scanner
(NextEngine)
Super Awesome 3D Scanner
Hardware €2 995
Kinect €99,99
Software €995-2 995
Software Free
Closed source
Open source
Kinect Fusion (Youtube)
Super Awesome 3D Scanner
Supercomputer
Regular laptop
In development
In development
Professionals
Students
German Research Center for AI
(Youtube)
Super Awesome 3D Scanner
Not available
Available ‘when it’s done…’
Goal: The goal of this project is to create models of real world in
frequently used data formats for low budget level applications &
education.
40

User input on events, webbrowsing,
powerpoint, windows…
Games
Medical sector
Other 3D-scanners

Coding 4 Fun Example base



41




Introduction
Kinect?
SA3D-scanner
You want to try at home?
42
43
XBox 360
Windows
Price
€100
€200
Licence
Home Use
Commercial
Distance
1,8 meters
recommended
Up to 40/50 cm
44
45
46

Create your own
◦ Basic (no overhead)

Use existing one
◦ (List of game engines on Wikipedia)
47

Mostly outdated or in development (new SDK)
◦ FAAST: Skeleton tracking (obsolete by new SDK)
◦ OpenKinect: Kinect on Mac/Linux
◦ OpenNI: Created for Asus Xtion but compatible with
Kinect.
◦ …
48





Kinectforwindows.org SDK, resellers …
Msdn.microsoft.com NUI documentation
channel9.msdn.com/coding4fun/kinect/
Demo’s, howto’s & examples
Amazon.com (& other stores): books
(Oudated with new SDK?)
And of course:
49
50
51