Post on 16-Jul-2015
transcript
CONTENTS
• INTRODUCTION
• 3D SENSING DEVICES
• 3D SENSING WITH KINECT
• 3D MODEL RECONSTRUCTION
• KINECT FUSION
• 3D PRINTER
• SYSTEM ARCHITECTURE
• 3D PRINTING
• CONCLUSION & FUTURE SCOPE2
INTRODUCTION
• From sense to print system:
• System that generates ready-to-print 3D models of
objects.
• It doesn’t require manual intervention the processing.
• These 3D models generated resemble and emulate the
real world.
3
• Steps in creating a 3D model :
• Segmentation
• Triangulation
• Scaling of the object to the dimension of the 3D printer
• Main components of system:
• 3D sensing device connected to the cloud
• Model reconstruction device to generate 3D CAD models
• 3D printers5
INTRODUCTION
3D SENSING DEVICES
• Generates 3D representations of the object using a
sensor.
• Generally in the form of 3D point clouds.
• 3D point clouds are generated from depth images.
• In depth image each pixel has a depth value assigned to
it.
5
• Depth values: Distances of surfaces in the world to the
origin of camera
• Common approaches for acquisition of depth images are:
Time of flight(TOF) systems
Triangulation based systems
Kinect sensor based systems
6
3D SENSING DEVICES
• TOF systems :
• The systems that emit signals like light.
• It then measure the time taken for the signal to
bounce back.
• Examples : LIDAR , radars etc
• Triangulation based systems:
• Measures distances to surfaces by matching
correspondences viewed by two sensors.
• Example: multi camera stereo vision system 7
3D SENSING DEVICES
• Kinect sensor based systems are:
• Monochrome CMOS sensors combined with laser
projector.
• The technology is called light coding.
• It is a variant of image based 3D reconstruction.
8
3D SENSING DEVICES
3D SENSING WITH KINECT
• This technique make use of the Kinect sensor.
• Kinect is a cheap and accessible 3D sensor.
• Uses light coding technique.
• Developed by Microsoft
9
3D SENSING WITH KINECT
• It has one camera and one IR projector
• Principle of light coding:
• Project a known IR pattern on to the screen
• Depth information is acquired from this pattern.
10
IR ProjectorCamera IR Camera
3D SENSING WITH KINECT
• IR Patterns from a Kinect Sensor (can only be viewed
only in a night vision camera)
11
3D MODEL RECONSTRUCTION
• Model Reconstruction is done by taking multiple views.
• These are obtained by moving the 3D sensing device
around the object.
• A full set of 3600 views are taken.
• Registering these multiple views will develop the full
3D model.
12
3D MODEL RECONSTRUCTION
• Registration : Estimation of rigid motion of a set of
points w.r.t another set of points.
• A set of points are transformed into a surface using mesh
algorithms
• Popular mesh algorithms are:
Greedy triangulation
Marching cubes
Poisson reconstruction13
RECONSTRUCTION – MARCHING CUBES
14
• The 3D points from the sensor is divied into number
of cubes.
• Each cube will intersect the surface in different
manners.
• This algorithm determines how the surface intersects
each cubes.
• There are 15 possible ways of intersecting the cube.
RECONSTRUCTION – MARCHING CUBES
15
• Creates a triangular mesh from the intersections in one of
the several cubes.
• Calculates the normals to the surface at each vertex of the
triangle.
• Then it marches to the next cube and repeat the above
process.
3D PRINTERS
• 3D objects are created using layering techniques.
• 3D printing technologies include:
Stereo lithography
Fused deposition modeling (FDM)
• FDM is popular in commercial 3D printers.
16
3D PRINTING
• Here printing is based on the following assumptions
• Scanned object, an inanimate object or human is
standing, sitting or lying on a plane
• 3D sensing device must close a loop around this
object
18
• Two ways of printing a 3D human are:
Full body statue
Bust
• To print a bust we concentrate on:
Face and head
There is no ground plane
• Camera poses tells us whether the human is sitting or
standing upright
19
3D PRINTING
• After applying segmentation and scaling, a polygon
mesh is generated.
• The resulting file is automatically sent to the machine
on the network.
• The machine is connected to the 3D printer.
20
3D PRINTING
CONCLUSION & FUTURE WORKS
• From Sense to Print is a system that can automatically generate
ready-to-print 3D CAD models.
• It uses low-cost Kinect sensor.
• It automatically scales the reconstruction to the size of the 3D
printer.
• The 3D prints currently generated are full models.
• Works are progressing on an automatic approach to generate the 3D
shells.
• They create replicas in more expensive materials such as copper or
other metals. 22
REFERENCES
• R. A. Newcombe, S. Izadi, O. Hilliges, D. Molyneaux, D.
Kim, A. J. Davison, P. Kohli, J. Shotton, S. Hodges, and A. W.
Fitzgibbon, “Kinect- fusion: Real-time dense surface mapping
and tracking.”
• R. Rusu and S. Cousins, “3d is here: Point cloud library
(PCL)”
23