SlideShare a Scribd company logo
1 of 67
Download to read offline
Introduction Design Implementation Conclusion
Correcting Projector Display for
Multi Planar Surfaces
by
Nilesh Heda
under the guidance of
Prof. Sharat Chandran
Computer Science and Engineering Department
Indian Institute of Technology, Bombay
July 7, 2005
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion
Outline
1 Introduction
Motivation
Related Issues
Problem Statement
Our Contribution
2 System Design
Projector-camera System
Display on Irregular Surface
Constraint Homography
3 Implementation and Results
System Requirements
Modules of the system
Results
4 Conclusion
Future Work
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Outline
1 Introduction
Motivation
Related Issues
Problem Statement
Our Contribution
2 System Design
Projector-camera System
Display on Irregular Surface
Constraint Homography
3 Implementation and Results
System Requirements
Modules of the system
Results
4 Conclusion
Future Work
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Conceptual Sketch of “Office of the Future”
Office of Future
Projectors creating virtual
office environment
Display anywhere
Teleconferencing
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Issues Regarding Projector Based System
Geometric Issues
Continuous Display Surface
Projector
Projector Projector
Planar Display Surface Discontinuous Display Surface
Single Surface Display Multi Surface Display
Image Intensity and color
Non-uniform Intensity
Overlapped region
Focussing of projected image
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Issues Regarding Projector Based System
Geometric Issues
Continuous Display Surface
Projector
Projector Projector
Planar Display Surface Discontinuous Display Surface
Single Surface Display Multi Surface Display
Image Intensity and color
Non-uniform Intensity
Overlapped region
Focussing of projected image
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Problem Statement
The aim of our project is to create a projector based system for
correctly displaying on irregular surfaces.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Problem Statement
The aim of our project is to create a projector based system for
correctly displaying on irregular surfaces.
Irregular Surface
Single Planar
Multi Planar
Curved
Combination of above
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Problem Statement
The aim of our project is to create a projector based system for
correctly displaying on irregular surfaces.
Problems with projected image
Focussing of projected
image
Image Intensity and color
Geometric issues
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Problem Statement
The aim of our project is to create a projector based system for
correctly displaying on irregular surfaces.
Problems with projected image
Focussing of projected
image
Image Intensity and color
Geometric issues
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Motivation Issues Problem Contribution
Overview of work done
Our Contribution
We have been successful in correctly displaying on
irregular surfaces.
Our algorithm doesn’t require any knowledge regarding
surface geometry.
We have proposed an iterative algorithm for satisfying the
constraint between homographies.
We have also sketched out a conceptual design for
shooting range simulator which is based on our method.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Outline
1 Introduction
Motivation
Related Issues
Problem Statement
Our Contribution
2 System Design
Projector-camera System
Display on Irregular Surface
Constraint Homography
3 Implementation and Results
System Requirements
Modules of the system
Results
4 Conclusion
Future Work
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Projector-camera Based System
β
Projector(s)
X
Z
Y
Screen
xz
y
α
Projectors
Display surface
Stationary cameras
Projector-screen-camera
calibration.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Projector-camera Based System
Camera
Projector(s)
X
Z
Y
Screen
xz
y
α
β
Projectors
Display surface
Stationary cameras
Projector-screen-camera
calibration.
Camera makes system active by
adding visual feedback.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Projector-camera Based System
Projector(s)
X
Z
Y
xz
y
α
β
pHc
pHs
cHs
Screen
Camera
Projectors
Display surface
Stationary cameras
Projector-screen-camera
calibration.
Projector as a dual of camera.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Projector-camera Homography
Camera and projector optics can be modeled as
projectivities.
We need to find mapping between P(xi, yi) in projector
coordinate system and C(Xi, Yi) in camera coordinate
system.
Homography is a mapping between two perspective
images of scene from two different viewpoints.
Mapping can be expressed in terms of a projective
transform P =p HcC.


xiw
yiw
w

 =


h1 h2 h3
h4 h5 h6
h7 h8 h9




Xi
Yi
1


Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Finding Projector-camera Homography
Steps for recovering pHc
1 Display a calibration pattern (4 points (P1, P2, P3, P4)).
2 Capture this calibration pattern by camera.
3 Detect and measure Feature points (C1, C2, C3, C4) from
captured camera image using standard computer vision
techniques.
4 Determine the point correspondence between Pi ↔ Ci.
5 Using these four point correspondences, recover pHc.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Camera-Screen Homography
Camera Screen Homography CHS
Mapping between pixels in projected display on the screen
and pixels in the camera image
Can be computed by point correspondence of known point
on screen with those in camera image
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Projector-Screen Homography
Projector-Screen Homography pHs
Mapping between pixels in projector input and pixels on
projected output on the screen
pHs = cHs.pHc
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Applying Pre-warp to input image
Compute the inverse homography pH−1
s = sHp
pre-warp the input signal to the projector
After Pre−warping
1
P
1
P
2 P
2
P
3P
4
P
4
p
1
p
1
p
2
p
2
P
3
p
4 p
3
p
3
p
4
Projector Image Projector ImageCamera Image Camera Image
Before Pre−warping
P
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Irregular Display Surface
Assumptions
A small quadrilateral can be considered as basic unit.
Surface can be reconstructed by tiling these small
quadrilaterals.
Need to determine point correspondence for all grid points
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Irregular Display Surface
Assumptions
A small quadrilateral can be considered as basic unit.
Surface can be reconstructed by tiling these small
quadrilaterals.
Need to determine point correspondence for all grid points
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Grid Point Detection Algorithm
Previous Approaches
Use different color for each grid points
Not a robust method.
Vertical/horizontal scan
Structured light
O(n) steps for n × n grid points.
Question ?
Can it be done in more faster way i.e. in O(log n) steps?
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Binary Encoding based Algorithm
Algorithm
Assign Unique binary code to all Projector feature Points
At each step i, the display of particular projector feature
point depends on ith bit.
Observe them by camera and detect feature points.
Set ith bit of code of these camera feature points.
After completion of logn2 steps, we will have set of n2
camera feature points each having unique code
Remove code with all bits ones as noise.
Find n2 point correspondences between projector and
camera feature points by simply matching the code.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Binary Encoding based Algorithm
Example
For computing point correspondence of 1024 × 1024 grid, we
need steps 20 steps.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Computation of Individual Homography
Compute n2 feature point correspondences.
For each quadrilateral i, Compute homography i
pHs from
the projector to the screen independently.
Problem
These homographies may not be consistent along the
boundaries between quadrilaterals.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Computation of Individual Homography
Compute n2 feature point correspondences.
For each quadrilateral i, Compute homography i
pHs from
the projector to the screen independently.
Problem
These homographies may not be consistent along the
boundaries between quadrilaterals.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
2 × 2 Grid Setup
Assumption
Unit quadrilaterals are extremely small.
Instead of considering all points at the
boundaries between adjacent
quadrilaterals, we consider only the
vertexes of the (tiny) quadrilaterals.
X ↔ xi ≡i
s HpX
x
p s
1
H 4
p
2
3
p s
pHs
Hs
X
H
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
2 × 2 Grid Setup
We would like to get
x1 = x2 = x3 = x4.
Example
This may not be the case due
to image processing errors in
the detection of grid points. x
p s
1
H 4
p
2
1
s
2
s
4
s
3
p s
pHs
p
3
s H
Hs
Hp pH
Hp
X
X1
X3 X4
X2
x2x1
x3 x4
H
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach I - Median
Median Approach
Choosing x as the median (instead of the average) of
x1, x2, x3 and x4.
Gives an acceptable solution even if two homographies are
incorrectly computed.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach II - Iterative algorithm
2
P1 P2 P3
P4 P5 P6
C1 C2 C3
C4 C5 C6
Projector Image Camera Image
pHs pHs
1
1
PHS and 2
pHS are homographies mapping between two
surfaces to the projector.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach II - Iterative algorithm
p1 p2 p3
p5
c1 c2 c3
c6c5c4p4 p6
Projector Image Camera Image
pHs pHs
1 −1 2 −1
1
PH−1
Sx =2
P H−1
Sx
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach II - Iterative algorithm
C1
P6
C2 C3
C4 C5 C6
Projector Image Camera Image
pHs
1
P3
P5
P1 P2
P4
Compute 1
PHS using four point correspondence (P1 ↔ C1,
P2 ↔ C2,P4 ↔ C4,P5 ↔ C5).
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach II - Iterative algorithm
−1
P2
P4 P6
C2 C3
C4 C5 C6
Projector Image Camera Image
pHs
1
P1 P3
P5
C1
c5
c3c1
c4
c2
c6
p1
p4 p5
p2
Compute p1, p2, p4, p5 from c1, c2, c4, c5 using equation
1
SHP
−1
c = p.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach II - Iterative algorithm
C6P4 C4
Projector Image Camera Image
P1 C1
c1
c4 c6
p1
p4
2
pHs
P2
p2
p5
P5 P6
P3 C2 C3
c2
c3
c5
C5
Compute 2PHS using six point correspondence (P2 ↔ C2,
P3 ↔ C3, P5 ↔ C5, P6 ↔ C6, 1
SHP
−1
c2 ↔ c2, 1
SHP
−1
c5 ↔ c5)
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach II - Iterative algorithm
p3
P2
P4 P6
C2 C3
C4 C5 C6
Projector Image Camera Image
P1 P3
P5
C1
c5
c3c1
c4
c2
c6
p1
p4
pHs
−12
p2
p5 p6
Update value of p2 and p5 and compute values of p3, p6 from
c2, c5, c3, c6 using equation 2SHP
−1
c = p.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Proj-cam Irregular constraint
Approach II - Iterative algorithm (contd.)
As we have used point correspondence 1
SHP
−1
c2 ↔ c2, we
can say that
1
PHS
−1
c2 =2
P HS
−1
c2 (1)
But, as computation of 2
PHS involves least square
approximation, above equation may not hold true.
Recompute the value of 1
PHS using constraint from 2
PHS.
After some iteration, the solution will converge and above
equation will be satisfied.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Outline
1 Introduction
Motivation
Related Issues
Problem Statement
Our Contribution
2 System Design
Projector-camera System
Display on Irregular Surface
Constraint Homography
3 Implementation and Results
System Requirements
Modules of the system
Results
4 Conclusion
Future Work
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Hardware/Software Requirements
Projector - Sharp XR-1X Multimedia Projector, 1100 ANSI
lumen, maximum native resolution of 1024 × 768.
Camera - D-Link DSC 350F camera, USB based,
640 × 480, 320 × 240 and 160 × 120 pixel resolution.
Platform - Windows 9x/NT/XP.
Programming Language and Libraries
VC++ for GUI handelling
DirectShow for video capture and filter graph generation
Direct3D for generation of virtual world.
Intel OpenCV for Computer Vision algorithms
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Input/Output of the System
Input
Texture Data
Image files or .x files or Video files
Output
Self adjusted projected image according to
projector-screen-camera position and orientation.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
System Diagram
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Module details
Capture Module
Captures frame from camera using Microsoft DirectShow at
640 × 480 resolution with 15 fps.
Uses callback functions for further computation.
Camera Calibration Module
Computes lens distortion using calibrated checker board
patterns
Radial Distortion
Tangential distortion
One time process for a particular camera
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Modules details
Mapping of Camera Space to Global Space
User specifies 4 fiducial in camera image using mouse
Computes Camera-screen homography CHS
Structured Light Registration
Uses gaussian blob as our feature point.
Determines center of blob with sub-pixel accuracy.
At End, the system has n2
features mapping from projector
pixels to the global coordinate system.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Modules details
Homography Computation
Computes individual homography i
PHS using least square
error SVD method.
Applies iterative recursive algorithm for constraint
homography solution.
Rendering Module
Uses Microsoft Direct3D for rendering.
Uses texture rendering techniques for rendering image data
on surface.
We were able to get 45fps at 1024 × 768 resolution with
5000 polygons on standard PC without any High end
graphics card.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
GUI
GUI
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
2 × 2 grid for a planar surface
Grid patterns
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
2 × 2 grid for a planar surface
Camera Point correspondence Planar Image after applying
Homograhy
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Display on 2-planar surface
Display Surface Distorted image
Corrected Image
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Display on 3-planar surface
Display Surface Distorted image
Corrected Image
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Keystone Correction
Keystone correction - before and after
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Constraint Homography
Constraint homography - Without and with
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Video on irregular Surface
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Requirements Modules Result
Video on irregular Surface
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Outline
1 Introduction
Motivation
Related Issues
Problem Statement
Our Contribution
2 System Design
Projector-camera System
Display on Irregular Surface
Constraint Homography
3 Implementation and Results
System Requirements
Modules of the system
Results
4 Conclusion
Future Work
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Conclusion
Successfully accomplished task of correcting projector
display on irregular surface.
Proposed an iterative refinement algorithm for satisfying
constrained between homography.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Future work
Improvement in current system
Deciding optimum grid size
Accurate detection of blurred feature points
Improving Iterative algorithm
Implementation of Shooting Range Simulator
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Demo
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Question and Answers
Thank You.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Shooting Range
Practicing and testing
shooting skills of shooter.
Pre-defined target
Accuracy of shooting
depends on distance
between target and actual
hit.
Important issues
Accurate detection.
Real-time response.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Our Approach
Screena(x,y) a’(x’,y’)
Camera Projector
Soldier
Projector creates
simulated environment.
Shooter(s) are sitting in a
simulated environment.
Shoots with the laser-gun
on a screen.
The camera(s) captures
the hit screen.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Important Components
Virtual World
Created by system.
Rendered as First Person View(FPV).
Shooter
Positioned in front of screen.
Shoots with Laser-pointer gun.
Projector-camera system
Projector projects virtual world.
Camera detects the hit.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Procedure for detection I
1 Compute H and H−1.
2 Create the Virtual world and render the scene from a first
person view (FPV).
3 Pre-warp the render scene by pre-computed H−1 and
project it.
4 Find the background intensity from first frame of each shot.
5 Do the Background Subtraction(Ideally zero).
6 Position the soldier (shooter) in front of display surface,
facing the screen.
7 Ask shooter to shoot the specified target on screen by
laser-gun ( Gun with laser pointer attached).
8 If laser pointer is displayed on screen, we will get change
in the color value in camera image.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Procedure for detection II
9 Find the region with high value after background
subtraction and mark its centroid as laser pointer position.
10 After detecting the laser pointer position, compute the
distance between actual target coordinate and detected hit
coordinate for finding accuracy of shooting.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Extension for invisible laser
The hit on screen provided visible feedback for next shot.
Use infra-red laser and additional infra-red camera.
Given a infra-red laser dot hit in infra-red camera image,
we need to determine where it lies in projector world.
Compute infra
c Hs by observing four fiducial points in
infra-red camera.
infra
p Hc =infra
c H−1
s.pHs (2)
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Extension for Multiple Shooter Environment
Factors under consideration
Accurate real time detection of individual laser pointers
Each soldier may have different view point for looking at
screen
We might be required to use multiple projector for
displaying on large area (10 × 3 )
Network based solution
Multiple instances of single shooter system connected to a
centralized rendering system through network.
The centralized rendering system will provide the rendered
view to all single shooter systems from their view point.
Single View solution
n different infra-red cameras configured for n different
frequencies.
Each infra-red camera will capture laser of particular
frequency.
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Multi-Planar Display
Before correction After correction
Nilesh Heda MTP Final Presentation
Introduction Design Implementation Conclusion Future Work
Multi-projector Display System
Images from individual projector
Image before and after brightness correction
Nilesh Heda MTP Final Presentation

More Related Content

Viewers also liked

Boletim de notícias RIET Nº 13 – novembro de 2016
Boletim de notícias RIET Nº 13 – novembro de 2016Boletim de notícias RIET Nº 13 – novembro de 2016
Boletim de notícias RIET Nº 13 – novembro de 2016RIET_INEW
 
Trabajo ruedas baloncesto
Trabajo ruedas baloncestoTrabajo ruedas baloncesto
Trabajo ruedas baloncestoJuanitoaragon
 
Boletín de noticias riet n° 11 – agosto 2016
Boletín de noticias riet n° 11 – agosto 2016Boletín de noticias riet n° 11 – agosto 2016
Boletín de noticias riet n° 11 – agosto 2016RIET_INEW
 
Paises hispanoamericanos
Paises hispanoamericanosPaises hispanoamericanos
Paises hispanoamericanosDaniela Picoli
 
Современные DDoS-атаки: тенденции, опасность, подходы к защите
Современные DDoS-атаки: тенденции, опасность, подходы к защитеСовременные DDoS-атаки: тенденции, опасность, подходы к защите
Современные DDoS-атаки: тенденции, опасность, подходы к защитеPositive Hack Days
 
CAD -- CUSUM Anomaly Detection
CAD -- CUSUM Anomaly DetectionCAD -- CUSUM Anomaly Detection
CAD -- CUSUM Anomaly DetectionFarkas Kinga
 
Literature review paper
Literature review paperLiterature review paper
Literature review papersepadge
 
National physical therapy month
National physical therapy monthNational physical therapy month
National physical therapy monthsepadge
 

Viewers also liked (9)

Boletim de notícias RIET Nº 13 – novembro de 2016
Boletim de notícias RIET Nº 13 – novembro de 2016Boletim de notícias RIET Nº 13 – novembro de 2016
Boletim de notícias RIET Nº 13 – novembro de 2016
 
Trabajo ruedas baloncesto
Trabajo ruedas baloncestoTrabajo ruedas baloncesto
Trabajo ruedas baloncesto
 
Boletín de noticias riet n° 11 – agosto 2016
Boletín de noticias riet n° 11 – agosto 2016Boletín de noticias riet n° 11 – agosto 2016
Boletín de noticias riet n° 11 – agosto 2016
 
Paises hispanoamericanos
Paises hispanoamericanosPaises hispanoamericanos
Paises hispanoamericanos
 
Современные DDoS-атаки: тенденции, опасность, подходы к защите
Современные DDoS-атаки: тенденции, опасность, подходы к защитеСовременные DDoS-атаки: тенденции, опасность, подходы к защите
Современные DDoS-атаки: тенденции, опасность, подходы к защите
 
CAD -- CUSUM Anomaly Detection
CAD -- CUSUM Anomaly DetectionCAD -- CUSUM Anomaly Detection
CAD -- CUSUM Anomaly Detection
 
Ventas cruzadas y aumenta tus ventas
Ventas cruzadas y aumenta tus ventasVentas cruzadas y aumenta tus ventas
Ventas cruzadas y aumenta tus ventas
 
Literature review paper
Literature review paperLiterature review paper
Literature review paper
 
National physical therapy month
National physical therapy monthNational physical therapy month
National physical therapy month
 

Similar to nilesh-Mtech-presentation

FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDSFACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDSIRJET Journal
 
Build Your Own 3D Scanner: 3D Scanning with Structured Lighting
Build Your Own 3D Scanner: 3D Scanning with Structured LightingBuild Your Own 3D Scanner: 3D Scanning with Structured Lighting
Build Your Own 3D Scanner: 3D Scanning with Structured LightingDouglas Lanman
 
Build Your Own 3D Scanner: 3D Scanning with Swept-Planes
Build Your Own 3D Scanner: 3D Scanning with Swept-PlanesBuild Your Own 3D Scanner: 3D Scanning with Swept-Planes
Build Your Own 3D Scanner: 3D Scanning with Swept-PlanesDouglas Lanman
 
B Eng Final Year Project Presentation
B Eng Final Year Project PresentationB Eng Final Year Project Presentation
B Eng Final Year Project Presentationjesujoseph
 
10cs65-cgnotes
10cs65-cgnotes10cs65-cgnotes
10cs65-cgnotesmushtaqdm
 
Non-Planar Projections (GRAPP 2008)
Non-Planar Projections (GRAPP 2008)Non-Planar Projections (GRAPP 2008)
Non-Planar Projections (GRAPP 2008)Matthias Trapp
 
Master defence 2020 - Roman Riazantsev - 3D Reconstruction of Video Sign Lan...
Master defence 2020 -  Roman Riazantsev - 3D Reconstruction of Video Sign Lan...Master defence 2020 -  Roman Riazantsev - 3D Reconstruction of Video Sign Lan...
Master defence 2020 - Roman Riazantsev - 3D Reconstruction of Video Sign Lan...Lviv Data Science Summer School
 
Video Stitching using Improved RANSAC and SIFT
Video Stitching using Improved RANSAC and SIFTVideo Stitching using Improved RANSAC and SIFT
Video Stitching using Improved RANSAC and SIFTIRJET Journal
 
Evaluation of conditional images synthesis: generating a photorealistic image...
Evaluation of conditional images synthesis: generating a photorealistic image...Evaluation of conditional images synthesis: generating a photorealistic image...
Evaluation of conditional images synthesis: generating a photorealistic image...SamanthaGallone
 
Caleb Vanderpleog Resume and Academic Summary
Caleb Vanderpleog Resume and Academic SummaryCaleb Vanderpleog Resume and Academic Summary
Caleb Vanderpleog Resume and Academic SummaryCaleb VanderPloeg
 
Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...
Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...
Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...CSCJournals
 
Computer graphics 9068
Computer graphics  9068Computer graphics  9068
Computer graphics 90681061992
 
[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...
[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...
[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...Owen Wu
 
Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...
Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...
Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...IJERA Editor
 
IEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargeting
IEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargetingIEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargeting
IEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargetingIEEEBEBTECHSTUDENTPROJECTS
 
Color Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible CloakColor Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible CloakAviral Chaurasia
 
Imaging automotive 2015 addfor v002
Imaging automotive 2015   addfor v002Imaging automotive 2015   addfor v002
Imaging automotive 2015 addfor v002Enrico Busto
 

Similar to nilesh-Mtech-presentation (20)

FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDSFACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
FACE COUNTING USING OPEN CV & PYTHON FOR ANALYZING UNUSUAL EVENTS IN CROWDS
 
Build Your Own 3D Scanner: 3D Scanning with Structured Lighting
Build Your Own 3D Scanner: 3D Scanning with Structured LightingBuild Your Own 3D Scanner: 3D Scanning with Structured Lighting
Build Your Own 3D Scanner: 3D Scanning with Structured Lighting
 
Build Your Own 3D Scanner: 3D Scanning with Swept-Planes
Build Your Own 3D Scanner: 3D Scanning with Swept-PlanesBuild Your Own 3D Scanner: 3D Scanning with Swept-Planes
Build Your Own 3D Scanner: 3D Scanning with Swept-Planes
 
B Eng Final Year Project Presentation
B Eng Final Year Project PresentationB Eng Final Year Project Presentation
B Eng Final Year Project Presentation
 
Narendra achari.s
Narendra achari.sNarendra achari.s
Narendra achari.s
 
10cs65-cgnotes
10cs65-cgnotes10cs65-cgnotes
10cs65-cgnotes
 
Non-Planar Projections (GRAPP 2008)
Non-Planar Projections (GRAPP 2008)Non-Planar Projections (GRAPP 2008)
Non-Planar Projections (GRAPP 2008)
 
Master defence 2020 - Roman Riazantsev - 3D Reconstruction of Video Sign Lan...
Master defence 2020 -  Roman Riazantsev - 3D Reconstruction of Video Sign Lan...Master defence 2020 -  Roman Riazantsev - 3D Reconstruction of Video Sign Lan...
Master defence 2020 - Roman Riazantsev - 3D Reconstruction of Video Sign Lan...
 
Praseed Pai
Praseed PaiPraseed Pai
Praseed Pai
 
Video Stitching using Improved RANSAC and SIFT
Video Stitching using Improved RANSAC and SIFTVideo Stitching using Improved RANSAC and SIFT
Video Stitching using Improved RANSAC and SIFT
 
Evaluation of conditional images synthesis: generating a photorealistic image...
Evaluation of conditional images synthesis: generating a photorealistic image...Evaluation of conditional images synthesis: generating a photorealistic image...
Evaluation of conditional images synthesis: generating a photorealistic image...
 
Caleb Vanderpleog Resume and Academic Summary
Caleb Vanderpleog Resume and Academic SummaryCaleb Vanderpleog Resume and Academic Summary
Caleb Vanderpleog Resume and Academic Summary
 
Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...
Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...
Tracking Chessboard Corners Using Projective Transformation for Augmented Rea...
 
Computer graphics 9068
Computer graphics  9068Computer graphics  9068
Computer graphics 9068
 
[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...
[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...
[GDC 2012] Enhancing Graphics in Unreal Engine 3 Titles Using AMD Code Submis...
 
Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...
Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...
Deblurring, Localization and Geometry Correction of 2D QR Bar Codes Using Ric...
 
IEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargeting
IEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargetingIEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargeting
IEEE 2014 MATLAB IMAGE PROCESSING PROJECTS Compressed domain video retargeting
 
Color Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible CloakColor Detection & Segmentation based Invisible Cloak
Color Detection & Segmentation based Invisible Cloak
 
Ppts21
Ppts21Ppts21
Ppts21
 
Imaging automotive 2015 addfor v002
Imaging automotive 2015   addfor v002Imaging automotive 2015   addfor v002
Imaging automotive 2015 addfor v002
 

nilesh-Mtech-presentation

  • 1. Introduction Design Implementation Conclusion Correcting Projector Display for Multi Planar Surfaces by Nilesh Heda under the guidance of Prof. Sharat Chandran Computer Science and Engineering Department Indian Institute of Technology, Bombay July 7, 2005 Nilesh Heda MTP Final Presentation
  • 2. Introduction Design Implementation Conclusion Outline 1 Introduction Motivation Related Issues Problem Statement Our Contribution 2 System Design Projector-camera System Display on Irregular Surface Constraint Homography 3 Implementation and Results System Requirements Modules of the system Results 4 Conclusion Future Work Nilesh Heda MTP Final Presentation
  • 3. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Outline 1 Introduction Motivation Related Issues Problem Statement Our Contribution 2 System Design Projector-camera System Display on Irregular Surface Constraint Homography 3 Implementation and Results System Requirements Modules of the system Results 4 Conclusion Future Work Nilesh Heda MTP Final Presentation
  • 4. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Conceptual Sketch of “Office of the Future” Office of Future Projectors creating virtual office environment Display anywhere Teleconferencing Nilesh Heda MTP Final Presentation
  • 5. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Issues Regarding Projector Based System Geometric Issues Continuous Display Surface Projector Projector Projector Planar Display Surface Discontinuous Display Surface Single Surface Display Multi Surface Display Image Intensity and color Non-uniform Intensity Overlapped region Focussing of projected image Nilesh Heda MTP Final Presentation
  • 6. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Issues Regarding Projector Based System Geometric Issues Continuous Display Surface Projector Projector Projector Planar Display Surface Discontinuous Display Surface Single Surface Display Multi Surface Display Image Intensity and color Non-uniform Intensity Overlapped region Focussing of projected image Nilesh Heda MTP Final Presentation
  • 7. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Problem Statement The aim of our project is to create a projector based system for correctly displaying on irregular surfaces. Nilesh Heda MTP Final Presentation
  • 8. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Problem Statement The aim of our project is to create a projector based system for correctly displaying on irregular surfaces. Irregular Surface Single Planar Multi Planar Curved Combination of above Nilesh Heda MTP Final Presentation
  • 9. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Problem Statement The aim of our project is to create a projector based system for correctly displaying on irregular surfaces. Problems with projected image Focussing of projected image Image Intensity and color Geometric issues Nilesh Heda MTP Final Presentation
  • 10. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Problem Statement The aim of our project is to create a projector based system for correctly displaying on irregular surfaces. Problems with projected image Focussing of projected image Image Intensity and color Geometric issues Nilesh Heda MTP Final Presentation
  • 11. Introduction Design Implementation Conclusion Motivation Issues Problem Contribution Overview of work done Our Contribution We have been successful in correctly displaying on irregular surfaces. Our algorithm doesn’t require any knowledge regarding surface geometry. We have proposed an iterative algorithm for satisfying the constraint between homographies. We have also sketched out a conceptual design for shooting range simulator which is based on our method. Nilesh Heda MTP Final Presentation
  • 12. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Outline 1 Introduction Motivation Related Issues Problem Statement Our Contribution 2 System Design Projector-camera System Display on Irregular Surface Constraint Homography 3 Implementation and Results System Requirements Modules of the system Results 4 Conclusion Future Work Nilesh Heda MTP Final Presentation
  • 13. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Projector-camera Based System β Projector(s) X Z Y Screen xz y α Projectors Display surface Stationary cameras Projector-screen-camera calibration. Nilesh Heda MTP Final Presentation
  • 14. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Projector-camera Based System Camera Projector(s) X Z Y Screen xz y α β Projectors Display surface Stationary cameras Projector-screen-camera calibration. Camera makes system active by adding visual feedback. Nilesh Heda MTP Final Presentation
  • 15. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Projector-camera Based System Projector(s) X Z Y xz y α β pHc pHs cHs Screen Camera Projectors Display surface Stationary cameras Projector-screen-camera calibration. Projector as a dual of camera. Nilesh Heda MTP Final Presentation
  • 16. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Projector-camera Homography Camera and projector optics can be modeled as projectivities. We need to find mapping between P(xi, yi) in projector coordinate system and C(Xi, Yi) in camera coordinate system. Homography is a mapping between two perspective images of scene from two different viewpoints. Mapping can be expressed in terms of a projective transform P =p HcC.   xiw yiw w   =   h1 h2 h3 h4 h5 h6 h7 h8 h9     Xi Yi 1   Nilesh Heda MTP Final Presentation
  • 17. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Finding Projector-camera Homography Steps for recovering pHc 1 Display a calibration pattern (4 points (P1, P2, P3, P4)). 2 Capture this calibration pattern by camera. 3 Detect and measure Feature points (C1, C2, C3, C4) from captured camera image using standard computer vision techniques. 4 Determine the point correspondence between Pi ↔ Ci. 5 Using these four point correspondences, recover pHc. Nilesh Heda MTP Final Presentation
  • 18. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Camera-Screen Homography Camera Screen Homography CHS Mapping between pixels in projected display on the screen and pixels in the camera image Can be computed by point correspondence of known point on screen with those in camera image Nilesh Heda MTP Final Presentation
  • 19. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Projector-Screen Homography Projector-Screen Homography pHs Mapping between pixels in projector input and pixels on projected output on the screen pHs = cHs.pHc Nilesh Heda MTP Final Presentation
  • 20. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Applying Pre-warp to input image Compute the inverse homography pH−1 s = sHp pre-warp the input signal to the projector After Pre−warping 1 P 1 P 2 P 2 P 3P 4 P 4 p 1 p 1 p 2 p 2 P 3 p 4 p 3 p 3 p 4 Projector Image Projector ImageCamera Image Camera Image Before Pre−warping P Nilesh Heda MTP Final Presentation
  • 21. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Irregular Display Surface Assumptions A small quadrilateral can be considered as basic unit. Surface can be reconstructed by tiling these small quadrilaterals. Need to determine point correspondence for all grid points Nilesh Heda MTP Final Presentation
  • 22. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Irregular Display Surface Assumptions A small quadrilateral can be considered as basic unit. Surface can be reconstructed by tiling these small quadrilaterals. Need to determine point correspondence for all grid points Nilesh Heda MTP Final Presentation
  • 23. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Grid Point Detection Algorithm Previous Approaches Use different color for each grid points Not a robust method. Vertical/horizontal scan Structured light O(n) steps for n × n grid points. Question ? Can it be done in more faster way i.e. in O(log n) steps? Nilesh Heda MTP Final Presentation
  • 24. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Binary Encoding based Algorithm Algorithm Assign Unique binary code to all Projector feature Points At each step i, the display of particular projector feature point depends on ith bit. Observe them by camera and detect feature points. Set ith bit of code of these camera feature points. After completion of logn2 steps, we will have set of n2 camera feature points each having unique code Remove code with all bits ones as noise. Find n2 point correspondences between projector and camera feature points by simply matching the code. Nilesh Heda MTP Final Presentation
  • 25. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Binary Encoding based Algorithm Example For computing point correspondence of 1024 × 1024 grid, we need steps 20 steps. Nilesh Heda MTP Final Presentation
  • 26. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Computation of Individual Homography Compute n2 feature point correspondences. For each quadrilateral i, Compute homography i pHs from the projector to the screen independently. Problem These homographies may not be consistent along the boundaries between quadrilaterals. Nilesh Heda MTP Final Presentation
  • 27. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Computation of Individual Homography Compute n2 feature point correspondences. For each quadrilateral i, Compute homography i pHs from the projector to the screen independently. Problem These homographies may not be consistent along the boundaries between quadrilaterals. Nilesh Heda MTP Final Presentation
  • 28. Introduction Design Implementation Conclusion Proj-cam Irregular constraint 2 × 2 Grid Setup Assumption Unit quadrilaterals are extremely small. Instead of considering all points at the boundaries between adjacent quadrilaterals, we consider only the vertexes of the (tiny) quadrilaterals. X ↔ xi ≡i s HpX x p s 1 H 4 p 2 3 p s pHs Hs X H Nilesh Heda MTP Final Presentation
  • 29. Introduction Design Implementation Conclusion Proj-cam Irregular constraint 2 × 2 Grid Setup We would like to get x1 = x2 = x3 = x4. Example This may not be the case due to image processing errors in the detection of grid points. x p s 1 H 4 p 2 1 s 2 s 4 s 3 p s pHs p 3 s H Hs Hp pH Hp X X1 X3 X4 X2 x2x1 x3 x4 H Nilesh Heda MTP Final Presentation
  • 30. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach I - Median Median Approach Choosing x as the median (instead of the average) of x1, x2, x3 and x4. Gives an acceptable solution even if two homographies are incorrectly computed. Nilesh Heda MTP Final Presentation
  • 31. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach II - Iterative algorithm 2 P1 P2 P3 P4 P5 P6 C1 C2 C3 C4 C5 C6 Projector Image Camera Image pHs pHs 1 1 PHS and 2 pHS are homographies mapping between two surfaces to the projector. Nilesh Heda MTP Final Presentation
  • 32. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach II - Iterative algorithm p1 p2 p3 p5 c1 c2 c3 c6c5c4p4 p6 Projector Image Camera Image pHs pHs 1 −1 2 −1 1 PH−1 Sx =2 P H−1 Sx Nilesh Heda MTP Final Presentation
  • 33. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach II - Iterative algorithm C1 P6 C2 C3 C4 C5 C6 Projector Image Camera Image pHs 1 P3 P5 P1 P2 P4 Compute 1 PHS using four point correspondence (P1 ↔ C1, P2 ↔ C2,P4 ↔ C4,P5 ↔ C5). Nilesh Heda MTP Final Presentation
  • 34. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach II - Iterative algorithm −1 P2 P4 P6 C2 C3 C4 C5 C6 Projector Image Camera Image pHs 1 P1 P3 P5 C1 c5 c3c1 c4 c2 c6 p1 p4 p5 p2 Compute p1, p2, p4, p5 from c1, c2, c4, c5 using equation 1 SHP −1 c = p. Nilesh Heda MTP Final Presentation
  • 35. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach II - Iterative algorithm C6P4 C4 Projector Image Camera Image P1 C1 c1 c4 c6 p1 p4 2 pHs P2 p2 p5 P5 P6 P3 C2 C3 c2 c3 c5 C5 Compute 2PHS using six point correspondence (P2 ↔ C2, P3 ↔ C3, P5 ↔ C5, P6 ↔ C6, 1 SHP −1 c2 ↔ c2, 1 SHP −1 c5 ↔ c5) Nilesh Heda MTP Final Presentation
  • 36. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach II - Iterative algorithm p3 P2 P4 P6 C2 C3 C4 C5 C6 Projector Image Camera Image P1 P3 P5 C1 c5 c3c1 c4 c2 c6 p1 p4 pHs −12 p2 p5 p6 Update value of p2 and p5 and compute values of p3, p6 from c2, c5, c3, c6 using equation 2SHP −1 c = p. Nilesh Heda MTP Final Presentation
  • 37. Introduction Design Implementation Conclusion Proj-cam Irregular constraint Approach II - Iterative algorithm (contd.) As we have used point correspondence 1 SHP −1 c2 ↔ c2, we can say that 1 PHS −1 c2 =2 P HS −1 c2 (1) But, as computation of 2 PHS involves least square approximation, above equation may not hold true. Recompute the value of 1 PHS using constraint from 2 PHS. After some iteration, the solution will converge and above equation will be satisfied. Nilesh Heda MTP Final Presentation
  • 38. Introduction Design Implementation Conclusion Requirements Modules Result Outline 1 Introduction Motivation Related Issues Problem Statement Our Contribution 2 System Design Projector-camera System Display on Irregular Surface Constraint Homography 3 Implementation and Results System Requirements Modules of the system Results 4 Conclusion Future Work Nilesh Heda MTP Final Presentation
  • 39. Introduction Design Implementation Conclusion Requirements Modules Result Hardware/Software Requirements Projector - Sharp XR-1X Multimedia Projector, 1100 ANSI lumen, maximum native resolution of 1024 × 768. Camera - D-Link DSC 350F camera, USB based, 640 × 480, 320 × 240 and 160 × 120 pixel resolution. Platform - Windows 9x/NT/XP. Programming Language and Libraries VC++ for GUI handelling DirectShow for video capture and filter graph generation Direct3D for generation of virtual world. Intel OpenCV for Computer Vision algorithms Nilesh Heda MTP Final Presentation
  • 40. Introduction Design Implementation Conclusion Requirements Modules Result Input/Output of the System Input Texture Data Image files or .x files or Video files Output Self adjusted projected image according to projector-screen-camera position and orientation. Nilesh Heda MTP Final Presentation
  • 41. Introduction Design Implementation Conclusion Requirements Modules Result System Diagram Nilesh Heda MTP Final Presentation
  • 42. Introduction Design Implementation Conclusion Requirements Modules Result Module details Capture Module Captures frame from camera using Microsoft DirectShow at 640 × 480 resolution with 15 fps. Uses callback functions for further computation. Camera Calibration Module Computes lens distortion using calibrated checker board patterns Radial Distortion Tangential distortion One time process for a particular camera Nilesh Heda MTP Final Presentation
  • 43. Introduction Design Implementation Conclusion Requirements Modules Result Modules details Mapping of Camera Space to Global Space User specifies 4 fiducial in camera image using mouse Computes Camera-screen homography CHS Structured Light Registration Uses gaussian blob as our feature point. Determines center of blob with sub-pixel accuracy. At End, the system has n2 features mapping from projector pixels to the global coordinate system. Nilesh Heda MTP Final Presentation
  • 44. Introduction Design Implementation Conclusion Requirements Modules Result Modules details Homography Computation Computes individual homography i PHS using least square error SVD method. Applies iterative recursive algorithm for constraint homography solution. Rendering Module Uses Microsoft Direct3D for rendering. Uses texture rendering techniques for rendering image data on surface. We were able to get 45fps at 1024 × 768 resolution with 5000 polygons on standard PC without any High end graphics card. Nilesh Heda MTP Final Presentation
  • 45. Introduction Design Implementation Conclusion Requirements Modules Result GUI GUI Nilesh Heda MTP Final Presentation
  • 46. Introduction Design Implementation Conclusion Requirements Modules Result 2 × 2 grid for a planar surface Grid patterns Nilesh Heda MTP Final Presentation
  • 47. Introduction Design Implementation Conclusion Requirements Modules Result 2 × 2 grid for a planar surface Camera Point correspondence Planar Image after applying Homograhy Nilesh Heda MTP Final Presentation
  • 48. Introduction Design Implementation Conclusion Requirements Modules Result Display on 2-planar surface Display Surface Distorted image Corrected Image Nilesh Heda MTP Final Presentation
  • 49. Introduction Design Implementation Conclusion Requirements Modules Result Display on 3-planar surface Display Surface Distorted image Corrected Image Nilesh Heda MTP Final Presentation
  • 50. Introduction Design Implementation Conclusion Requirements Modules Result Keystone Correction Keystone correction - before and after Nilesh Heda MTP Final Presentation
  • 51. Introduction Design Implementation Conclusion Requirements Modules Result Constraint Homography Constraint homography - Without and with Nilesh Heda MTP Final Presentation
  • 52. Introduction Design Implementation Conclusion Requirements Modules Result Video on irregular Surface Nilesh Heda MTP Final Presentation
  • 53. Introduction Design Implementation Conclusion Requirements Modules Result Video on irregular Surface Nilesh Heda MTP Final Presentation
  • 54. Introduction Design Implementation Conclusion Future Work Outline 1 Introduction Motivation Related Issues Problem Statement Our Contribution 2 System Design Projector-camera System Display on Irregular Surface Constraint Homography 3 Implementation and Results System Requirements Modules of the system Results 4 Conclusion Future Work Nilesh Heda MTP Final Presentation
  • 55. Introduction Design Implementation Conclusion Future Work Conclusion Successfully accomplished task of correcting projector display on irregular surface. Proposed an iterative refinement algorithm for satisfying constrained between homography. Nilesh Heda MTP Final Presentation
  • 56. Introduction Design Implementation Conclusion Future Work Future work Improvement in current system Deciding optimum grid size Accurate detection of blurred feature points Improving Iterative algorithm Implementation of Shooting Range Simulator Nilesh Heda MTP Final Presentation
  • 57. Introduction Design Implementation Conclusion Future Work Demo Nilesh Heda MTP Final Presentation
  • 58. Introduction Design Implementation Conclusion Future Work Question and Answers Thank You. Nilesh Heda MTP Final Presentation
  • 59. Introduction Design Implementation Conclusion Future Work Shooting Range Practicing and testing shooting skills of shooter. Pre-defined target Accuracy of shooting depends on distance between target and actual hit. Important issues Accurate detection. Real-time response. Nilesh Heda MTP Final Presentation
  • 60. Introduction Design Implementation Conclusion Future Work Our Approach Screena(x,y) a’(x’,y’) Camera Projector Soldier Projector creates simulated environment. Shooter(s) are sitting in a simulated environment. Shoots with the laser-gun on a screen. The camera(s) captures the hit screen. Nilesh Heda MTP Final Presentation
  • 61. Introduction Design Implementation Conclusion Future Work Important Components Virtual World Created by system. Rendered as First Person View(FPV). Shooter Positioned in front of screen. Shoots with Laser-pointer gun. Projector-camera system Projector projects virtual world. Camera detects the hit. Nilesh Heda MTP Final Presentation
  • 62. Introduction Design Implementation Conclusion Future Work Procedure for detection I 1 Compute H and H−1. 2 Create the Virtual world and render the scene from a first person view (FPV). 3 Pre-warp the render scene by pre-computed H−1 and project it. 4 Find the background intensity from first frame of each shot. 5 Do the Background Subtraction(Ideally zero). 6 Position the soldier (shooter) in front of display surface, facing the screen. 7 Ask shooter to shoot the specified target on screen by laser-gun ( Gun with laser pointer attached). 8 If laser pointer is displayed on screen, we will get change in the color value in camera image. Nilesh Heda MTP Final Presentation
  • 63. Introduction Design Implementation Conclusion Future Work Procedure for detection II 9 Find the region with high value after background subtraction and mark its centroid as laser pointer position. 10 After detecting the laser pointer position, compute the distance between actual target coordinate and detected hit coordinate for finding accuracy of shooting. Nilesh Heda MTP Final Presentation
  • 64. Introduction Design Implementation Conclusion Future Work Extension for invisible laser The hit on screen provided visible feedback for next shot. Use infra-red laser and additional infra-red camera. Given a infra-red laser dot hit in infra-red camera image, we need to determine where it lies in projector world. Compute infra c Hs by observing four fiducial points in infra-red camera. infra p Hc =infra c H−1 s.pHs (2) Nilesh Heda MTP Final Presentation
  • 65. Introduction Design Implementation Conclusion Future Work Extension for Multiple Shooter Environment Factors under consideration Accurate real time detection of individual laser pointers Each soldier may have different view point for looking at screen We might be required to use multiple projector for displaying on large area (10 × 3 ) Network based solution Multiple instances of single shooter system connected to a centralized rendering system through network. The centralized rendering system will provide the rendered view to all single shooter systems from their view point. Single View solution n different infra-red cameras configured for n different frequencies. Each infra-red camera will capture laser of particular frequency. Nilesh Heda MTP Final Presentation
  • 66. Introduction Design Implementation Conclusion Future Work Multi-Planar Display Before correction After correction Nilesh Heda MTP Final Presentation
  • 67. Introduction Design Implementation Conclusion Future Work Multi-projector Display System Images from individual projector Image before and after brightness correction Nilesh Heda MTP Final Presentation