GoBack - SourceForgemiarn.sourceforge.net/pdf/Apresentacao.pdfFuture Work The End Fast Line,...

Preview:

Citation preview

GoBack

Fast Line, Arc/Circle and Leg Detectionfrom Laser Scan Data in a Player Driver

João Xavier⋆, Marco Pacheco⊚,Daniel Castro⊚, António Ruano⊚ and Urbano Nunes⋆

⊚ Centre for Intelligent Systems - CSIUniversity of Algarve, Portugal

⋆ Institute of Systems and Robotics - ISRUniversity of Coimbra, Portugal

Outline

● Presentation Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 2/21

Presentation Outline

1. Introduction2. Feature detection3. Visualization4. Results5. Conclusion and future work

Outline

● Presentation Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 2/21

Presentation Outline

1. Introduction■ Goals■ Platform used

2. Feature detection3. Visualization4. Results5. Conclusion and future work

Outline

● Presentation Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 2/21

Presentation Outline

1. Introduction2. Feature detection

■ Line■ Arc/circle■ Legs of people

3. Visualization4. Results5. Conclusion and future work

Outline

● Presentation Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 2/21

Presentation Outline

1. Introduction2. Feature detection3. Visualization4. Results5. Conclusion and future work

Outline

● Presentation Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 2/21

Presentation Outline

1. Introduction2. Feature detection3. Visualization4. Results5. Conclusion and future work

Outline

● Presentation Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 2/21

Presentation Outline

1. Introduction2. Feature detection3. Visualization4. Results5. Conclusion and future work

Outline

Introduction

● Project goals

● Hardware configuration used

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 3/21

Project goals

■ Conceive and implement algorithms for fast feature detection■ Identify architetonic structures by their geometric shape in the laser

data■ Recognize legs of persons regardless their dynamics■ View the results online in 3 Dimensions

Outline

Introduction

● Project goals

● Hardware configuration used

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 3/21

Project goals

■ Conceive and implement algorithms for fast feature detection■ Identify architetonic structures by their geometric shape in the laser

data■ Recognize legs of persons regardless their dynamics■ View the results online in 3 Dimensions

Outline

Introduction

● Project goals

● Hardware configuration used

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 3/21

Project goals

■ Conceive and implement algorithms for fast feature detection■ Identify architetonic structures by their geometric shape in the laser

data■ Recognize legs of persons regardless their dynamics■ View the results online in 3 Dimensions

Outline

Introduction

● Project goals

● Hardware configuration used

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 3/21

Project goals

■ Conceive and implement algorithms for fast feature detection■ Identify architetonic structures by their geometric shape in the laser

data■ Recognize legs of persons regardless their dynamics■ View the results online in 3 Dimensions

Outline

Introduction

● Project goals

● Hardware configuration used

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 4/21

Hardware configuration used

■ Laser SickLMS200◆ 5Hz

◆ 180◦ , with 0.5

◦ angular◆ Error ±15mm

◆ range 8m

■ Pioneer 2 DX robot

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 5/21

Range segmentation

Range segmentation produces clusters of consecutive scan points,which due to their proximity probably belong to the same object.

■ The laser ranges (Rk) areparsed in order to isolateobjects in segments

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 5/21

Range segmentation

Range segmentation produces clusters of consecutive scan points,which due to their proximity probably belong to the same object.

■ The laser ranges (Rk) areparsed in order to isolateobjects in segments

■ A new segment is started if‖Rk − Rk+1‖ ≥ Threshold

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 5/21

Range segmentation

Range segmentation produces clusters of consecutive scan points,which due to their proximity probably belong to the same object.

■ The laser ranges (Rk) areparsed in order to isolateobjects in segments

■ A new segment is started if‖Rk − Rk+1‖ ≥ Threshold

■ The following algorithms willbe applied to these segmentsto identify the desiredfeatures

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

1. Obtain the line passing bythe two extreme points bythe method of the"Orthogonal Regression of aLine"

2. The fitting error is the pointmost distant to the line

3. If the fitting error is above agiven threshold, split (wherethe greatest error occurred)and repeat 1-3 with the leftand right sub-scan

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 6/21

Line detection: Recursive line fitting

To avoid creation of smallpolygons, three rules werestate:1. more than 5 points in each

analyzed segment2. distance between extremes

greater than 0.1 meters3. maximum fitting error 0.02

meters

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 7/21

Arc/circle prerequisites, people legs

How to quickly tell if a segment can be an arc ?

■ The middle point of thesegment must be inside anarea delimited by the twolines parallel to the extremesof the same segment

P1

Maximal distance

P3

Maximal distance with noise

Minimal distance

Laser

0.1 x P1P3

0.7 x P1P3

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 7/21

Arc/circle prerequisites, people legs

How to quickly tell if a segment can be an arc ?

■ The middle point of thesegment must be inside anarea delimited by the twolines parallel to the extremesof the same segment

■ Leg detection is the samerule but with a diameter P1P3

limited from 0.1m to 0.25m

P1

Maximal distance

P3

Maximal distance with noise

Minimal distance

Laser

0.1 x P1P3

0.7 x P1P3

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 8/21

Arc/circle detection with IAV

Inscribed Angle Variance (IAV) makes use of one simple geometricpostulate from the Euclid book of elements: every point in an arc hascongruent angles (angles with equal values) in respect to the extremes.∠P1P2P4 = ∠P1P3P4 =

∠P1OP4

2

P1

P3

P4

P2 95°

95°

O

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 8/21

Arc/circle detection with IAV

Inscribed Angle Variance (IAV) makes use of one simple geometricpostulate from the Euclid book of elements: every point in an arc hascongruent angles (angles with equal values) in respect to the extremes.∠P1P2P4 = ∠P1P3P4 =

∠P1OP4

2

1. put the inscribed angles in avector

2. calculate the average andthe standard deviation of thevector

P1

P3

P4

P2 95°

95°

O

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 8/21

Arc/circle detection with IAV

Inscribed Angle Variance (IAV) makes use of one simple geometricpostulate from the Euclid book of elements: every point in an arc hascongruent angles (angles with equal values) in respect to the extremes.∠P1P2P4 = ∠P1P3P4 =

∠P1OP4

2

■ the average tells us the kindof shape

■ the standard deviationmeasures how regular it is

■ line has average of 180◦

■ arcs normally stand between90

◦ and 135◦

P1

P3

P4

P2 95°

95°

O

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 8/21

Arc/circle detection with IAV

Inscribed Angle Variance (IAV) makes use of one simple geometricpostulate from the Euclid book of elements: every point in an arc hascongruent angles (angles with equal values) in respect to the extremes.∠P1P2P4 = ∠P1P3P4 =

∠P1OP4

2

■ the average tells us the kindof shape

■ the standard deviationmeasures how regular it is

■ line has average of 180◦

■ arcs normally stand between90

◦ and 135◦

P1 P2 P3

180° 180° 180°

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 9/21

How can we be sure it was an arc ?

The confidence of the result depends on:■ the number of in-between points■ low standard deviation■ the average inscribed angle value near 90

◦. For an inscribed angleaverage of 90

◦ half of the circle is visible

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 10/21

Other IAV Possibilities

■ Detecting lines (average of180

◦)■ Detecting arcs embedded in

lines■ Detecting "arc-like" shapes

(round tree logs)■ Apply it to blocks of five

points (divide and conquerstrategy)

P1 P2 P3

180° 180° 180°

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 10/21

Other IAV Possibilities

■ Detecting lines (average of180

◦)■ Detecting arcs embedded in

lines1. Apply recursive line fitting

to the segment to identifylines

2. with the subsegments notidentified as lines applyIAV

■ Detecting "arc-like" shapes(round tree logs)

■ Apply it to blocks of fivepoints (divide and conquerstrategy)

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 10/21

Other IAV Possibilities

■ Detecting lines (average of180

◦)■ Detecting arcs embedded in

lines■ Detecting "arc-like" shapes

(round tree logs)■ Apply it to blocks of five

points (divide and conquerstrategy)

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 10/21

Other IAV Possibilities

■ Detecting lines (average of180

◦)■ Detecting arcs embedded in

lines■ Detecting "arc-like" shapes

(round tree logs)■ Apply it to blocks of five

points (divide and conquerstrategy)

Outline

Introduction

Feature detection

● Range segmentation

● Line detection: Recursive line

fitting

● Arc/circle prerequisites,

people legs

● Arc/circle detection with IAV

● How can we be sure it was an

arc ?● Other IAV Possibilities

● Cases where IAV cannot be

applied

Visualization

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 11/21

Cases where IAV cannot be applied

Embedded arcs or multiple arcs cannot be identified

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 12/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 12/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 13/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment■ Loading of 3d models from files

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 13/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment■ Loading of 3d models from files■ Plugins using libtool

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 13/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment■ Loading of 3d models from files■ Plugins using libtool■ Cross platform thanks to FLTK(Fast Light Toolkit)

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 13/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment■ Loading of 3d models from files■ Plugins using libtool■ Cross platform thanks to FLTK(Fast Light Toolkit)■ 2D(FLTK) or 3D(OpenGL)

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 13/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment■ Loading of 3d models from files■ Plugins using libtool■ Cross platform thanks to FLTK(Fast Light Toolkit)■ 2D(FLTK) or 3D(OpenGL)■ Graphical User Interface

Outline

Introduction

Feature detection

Visualization

● PlayerGL

● PlayerGL

Tests

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 13/21

PlayerGL

A visualization tool with focus on robot perception, modularity and fastdevelopment■ Loading of 3d models from files■ Plugins using libtool■ Cross platform thanks to FLTK(Fast Light Toolkit)■ 2D(FLTK) or 3D(OpenGL)■ Graphical User Interface■ Exports screenshots to ps or pdf

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 14/21

Dynamic test

1. Robot facing dead end2. Robot turns 180

3. Moves forward

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 15/21

Static test

A student pulls a cylindric robot around two collumns.

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 16/21

Results of the Static test

Outline

Introduction

Feature detection

Visualization

Tests

● Dynamic test

● Static test

● Results of the Static test

● Timings

Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 17/21

Timings

■ line detection times oscilateswith the number of subscandivisions, i.e. corners;

■ circle pre-requisites savesprecious processing time;

■ both algorithms are fast,specially when followingcorridors without corners;

■ the worst time for lines was1.940ms and for circles was1.082ms.

Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

● Conclusion

Future Work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 18/21

Conclusion

■ Algorithms for fast detection of lines, circles/arcs and legsimplemented

■ Visualization tool was started■ Simple scene interpretation was achieved

Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

● Future work

● Future work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 19/21

Future work

The Modules for Intelligent Autonomous Robot Navigation framework

Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

● Future work

● Future work

The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 20/21

Future work

On features:■ Adaptative segmentation■ scene interpretation■ correct odometry using laser data■ SLAM■ automatic object recognitionOn visualization: Release PlayerGL 0.1

Outline

Introduction

Feature detection

Visualization

Tests

Conclusion

Future Work

The End

● The End

Fast Line, Arc/Circle and Leg Detection from Laser Scan Data in a Player Driver 21/21

The End

The endDownloads at http://miarn.cjb.net

Questions ?

Recommended