A Flexible Camera Calibration Tool for 3D Capture

1
A Flexible Camera Calibration Tool for 3D Capture

• Lei Wang
• Media and Machine Lab
• Advisor Cindy Grimm

2
3D Capture

• Data Acquisition
• Camera Calibration
• Shape Integration
• Texture Synthesis
• Shape Texture Integration

3
Camera Calibration

• Camera Model
• Global Approach
• Requirement
• Previous Work
• Our Approach
• Conclusion

4
Pinhole Camera Model
5
Denotation

• 2D Point m u, v, 1T
• 3D Point M X, Y, ZT
• 2D 3D s m A R T M
• A Intrinsic Parameters
• R T Extrinsic Parameters

6
Equation
7
Global Approach

• Take pictures
• Detect the pattern
• Extract the feature
• Solve for intrinsic parameters (one for all)
• Solve for extrinsic parameters (one for each)

8
Requirement

• Automated
• Hemisphere Visibility
• Not occlude the object
• Robust on lighting conditions
• Easy to build
• Easy to detect

9
Planar Pattern

• Checkerboard
• Tsais Algorithm
• Zhangs Algorithm
• Geometry Pattern
• Concentric Circle
• Jun-Sik Kim, Ho-Won Kim and In-So Kweons
  Algorithm

10
Improved Planar Pattern

• Move the pattern by hand
• Place two or more patterns on different plane
• Place mirrors and use the additional reflected
  pattern

11
Our Approach Outline

• Design Calibrating Pattern
• Build Calibrating Pattern
• Feature Extraction
• Feature Mapping Gradient Search
• Test

12
Cone with basic 3D shapes

• Automated
• Hemisphere visibility
• Not occlude the object part of the shape is
  enough for relying on a group of points
• Robust for lighting conditions color ratio
• Easy to detect color boundary
• Easy to build build from a planar pattern

13
Basic Shapes

• Circles
• Ellipses
• Lines
• Points

14
Stable Color Ratio
15
Parameterized Representation

• Line
• Circle
• Ellipse

16
The Printable Planar Pattern

• 3D equation, 2D print translation
• Physical tips

17
Boundary Detection

• Color ratio
• Grouping

18
Ellipses

• Ellipse detection
• Ellipse fitting

19
Lines

• Line detection
• Line fitting

20
Points

• Direct Point detection
• Intersection of two shapes
• - Line-line intersection
• - Line-circle intersection
• - Line-ellipse intersection
• - Ellipse-ellipse intersection(not recommend)

21
Solve for Intrinsic Parameters

• Checkerboard
• OpenCV
• Flagged Checkerboard

22
Extrinsic by Points

• Use Points
• Linear
• Limitation
• - Hard to label
• - Inaccurate
• - Six points rules not guaranteed

23
Extrinsic by Shapes

• Other Shape
• Non-Linear
• Gradient-decent Search
• - Cost Function
• - Initial Guess
• - Step Choose
• - Stop Condition

24
Extrinsic by Lines

• Cost Function
• Initial Guess
• Step Choose
• Stop Condition

25
Extrinsic by Conics

• Cost Function
• Initial Guess
• Step Choose
• Stop Condition

26
General Extrinsic

• General Case
• Use Combination Pattern

27
Test
28
Conclusion

• Easiness
• Efficiency

29
Conclusion

• Accuracy

30
References
31
Questions?
32
Thank You