Vision for Automation...Lane departure warning systems. Head tracking systems for drowsiness detection. Driver assistance systems. Reading automobile license plates, and trafﬁc management

Vision for Automation

Ranga RodrigoDepartment of Electronic and Telecommunication Engineering

University of MoratuwaSri Lanka

ICIAfS 2008 Vision for Automation Workshop

December 10, 2008

ICIAfS 2008 Robotics and Automation Workshop Vision for Automation 1/43

Outline

1 Introduction

2 Applications

3 Vision in Automation

4 Software Tools

5 Examples of State-of-the-Art

6 Summary


Introduction

Outline

1 Introduction

2 Applications


4 Software Tools


6 Summary


Introduction

What Is Computer Vision?

The goal is the emulation of the visual capabilityof human beings using computers.

In other words, computer vision is making themachine see as we do!It is challenging.Steps:

1 Image acquisition2 Image manipulation3 Image understanding4 Decision making


Introduction


The goal is the emulation of the visual capabilityof human beings using computers.In other words, computer vision is making themachine see as we do!

It is challenging.Steps:



Introduction


The goal is the emulation of the visual capabilityof human beings using computers.In other words, computer vision is making themachine see as we do!It is challenging.

Steps:1 Image acquisition2 Image manipulation3 Image understanding4 Decision making


Introduction


The goal is the emulation of the visual capabilityof human beings using computers.In other words, computer vision is making themachine see as we do!It is challenging.Steps:



Introduction

Main Driving Technologies

Signal processing.Multiple view geometry [2].Optimization.Machine learning.Hardware and algorithms.


Applications

Outline

1 Introduction

2 Applications


4 Software Tools


6 Summary


Applications

Applications 1

Automotive:Lane departure warning systems.Head tracking systems for drowsiness detection.Driver assistance systems.Reading automobile license plates, and trafficmanagement.

Photography:In camera face detection [6], red eye removal, andother functions.Automatic panorama stitching [1].

1(From http://www.cs.ubc.ca/spider/lowe/vision.html)ICIAfS 2008 Robotics and Automation Workshop Vision for Automation 7/43

Applications

ApplicationsMovie and video (a very big industry):

Augmented reality.Tracking objects in video or film and solving for 3-Dmotion to allow for precise augmentation with 3-Dcomputer graphics.Multiple cameras to precisely track tennis and cricketballs.Human expression recognition.Software for 3-D visualization for sportsbroadcasting and analysis.Tracking consistent regions in video and insertvirtual advertising.Tracking for character animation.Motion capture, camera tracking, panoramastitching, and building 3D models for movies.


Applications

Camera Tracking

Source: http://www.2d3.com/capability

Show 2d3 video.


http://www.2d3.com/capability

Applications

ApplicationsGames:

Tracking human gestures for playing games orinteracting with computers.Tracking the hand and body motions of players (tocontrol the Sony Playstation).Image-based rendering, vision for graphics.

General purpose:Inspection and localization tasks, people counting,biomedical, and security. etc.Object recognition and navigation for mobilerobotics, grocery retail, and recognition from cellphone cameras.Laser-based 3D vision systems for use on the spaceshuttles and other applications.Image retrieval based on content.


Applications

ApplicationsIndustrial automation (a very big industry):

Vision-guided robotics in the automotive industry.Electronics inspection systems for componentassembly.

Medical and biomedical (maturing):Vision to detect and track the pose of markers forsurgical applications, needle insertion, and seedplanting.Teleoperations.Quantitative analysis of medical imaging, includingdiagnosis such as cancer.

Security and biometrics (thriving):Intelligent video surveillance.Biometric face, fingerprint, and iris recognition.Behavior detection.


Applications

Minimal Invasive Surgery

Source: http://www.davincisurgery.com/surgery/system/index.aspx


http://www.davincisurgery.com/surgery/system/index.aspx

Applications

Areas of Advancement

Hardware.Image segmentation.3-D reconstruction.Object detection.Navigation.

Scene understanding.


Applications

Areas of Advancement

Hardware.Image segmentation.3-D reconstruction.Object detection.Navigation.Scene understanding.


Vision in Automation

Outline

1 Introduction

2 Applications


4 Software Tools


6 Summary



What’sneeded?

cameras

software

actuators



Cameras

Camera, and a frame grabber.IEEE 1394 or USB cameras.Ethernet cameras.



Source: http://www.ptgrey.com/products/chameleon/index.asp

Source: http://www.matrox.com/imaging/products/vio/home.cfm


http://www.ptgrey.com/products/chameleon/index.asp

http://www.matrox.com/imaging/products/vio/home.cfm


Source: http://www3.elphel.com/sites/default/files/images/nc353 io geo.jpeg


http://www3.elphel.com/sites/default/files/images/nc353_io_geo.jpeg


Inspection

ColorBarcode scanningCharacter recognition



Inspection: Examples

Defects in parts, measurement of size.Robotic bin picking.If each slot is filled in a carton of pills.Character recognition.



Visual Servoing

Uses vision in the servo loop [3].Dynamic look and move needs the accuracy of thevision sensor and robot end-effector.Having visual feedback in the control loop increasesthe overall accuracy of the control loop.

Visual ServoingMachine vision can provide closed-loop positioncontrol for a robot end-effector—this is referred to asvisual servoing.



Visual Servoing

Uses vision in the servo loop [3].Dynamic look and move needs the accuracy of thevision sensor and robot end-effector.Having visual feedback in the control loop increasesthe overall accuracy of the control loop.

Visual ServoingMachine vision can provide closed-loop positioncontrol for a robot end-effector—this is referred to asvisual servoing.



Visual Servoing—Camera Configuration

End-effector mounted Fixed configuration



Servoing Architectures

Is the control structure hierarchical, with thevision system providing set-points as input to therobot’s joint level controller, or does the visualcontroller directly compute the joint-level inputs?Is the error signal defined in 3-D (task space)coordinates, or directly in terms of imagefeatures?



Servoing Architectures

Dynamic position-based look-and-movestructure.Dynamic image-based look-and-move structure.Position-based (direct) visual servo structure.Image-based (direct) visual servo structure.



Dynamic Position-BasedLook-and-Move Structure

cxd

Cartesian

control law−

+

Joint controllers and power amps

Image feature

extraction

Pose

estimation

videofc x̂



Dynamic Image-BasedLook-and-Move Structure

fdFeature space

control law−

+

Joint controllers and power amps

Image feature

extraction

videof



Position-Based (Direct) Visual ServoStructure.

cxd

Cartesian

control law−

+

Power amps

Image feature

extraction

Pose

estimation

videofc x̂



Image-Based (Direct) Visual ServoStructure.

fdFeature space

control law−

+

Power amps

Image feature

extraction

videof


Software Tools

Outline

1 Introduction

2 Applications


4 Software Tools


6 Summary


Software Tools

Software Tools

Octave or Matlab.C or C++ with a library such as OpenCV.


Software Tools

Image Processing using Octave orMatlab

Simple and quick.A lot of library functions.Interpreted.


Software Tools

Octave Examples

Image reading and writing.Histograms.Filtering.


Software Tools

Image Processing using OpenCV

Power of C++.Well coded.


Software Tools

OpenCV Examples

1 Image reading and writing.2 Edge detection.3 Template matching.4 Capturing video.


Examples of State-of-the-Art

Outline

1 Introduction

2 Applications


4 Software Tools


6 Summary



Segmentation Using Graph Cuts [5]












3-D Reconstruction

Can we obtain a 3-D view of a scene, given onlya set of (2-D) images?

Yes. Using multiple view geometry, we canreconstruct a scene.Show Leibe et al. video [4].



3-D Reconstruction

Can we obtain a 3-D view of a scene, given onlya set of (2-D) images?Yes. Using multiple view geometry, we canreconstruct a scene.Show Leibe et al. video [4].



Object Detection: Face Detection

Show OpenCV sample.ICIAfS 2008 Robotics and Automation Workshop Vision for Automation 38/43


Navigation: Sanford’s Robot Stanley

Show video.


Summary

Outline

1 Introduction

2 Applications


4 Software Tools


6 Summary


Summary

Conclusion

Vision-based automation is promising.

Solutions are simple in a controlledenvironment.State-of-the-art is very interesting.


Summary

Conclusion

Vision-based automation is promising.Solutions are simple in a controlledenvironment.

State-of-the-art is very interesting.


Summary

Conclusion

Vision-based automation is promising.Solutions are simple in a controlledenvironment.State-of-the-art is very interesting.


Summary

Thank you.

OpenCV examples, and Octave examples are here:http://www.ent.mrt.ac.lk/ ranga/publications.html


http://www.ent.mrt.ac.lk/~ranga/publications.html

Summary

Matthew Brown and David Lowe.Recognising panoramas.In Proceedings of the 9th International Conference on Computer Vision, pages1218–1225, Nice, France, October 2003.

Richard Hartley and Andrew Zisserman.Multiple View Geometry in Computer Vision.Cambridge University Press, 2nd edition, 2003.

Seth Hutchinson, Gregory D. Hager, and Peter I. Corke.A tutorial on visual servo control.IEEE Transactions Robotics and Automation, 12 No. 5:651–670, 1996.

Bastian Leibe, Nico Cornelis, Kurt Cornelis, and Luc Van Gool.Dynamic 3D scene analysis from a moving vehicle.In Proceedings of the IEEE Computer Society Conference on Computer Vision andPattern Recognition, pages 1–8, Minneapolis, MN, June 2007.

Carsten Rother, Vladimir Kolmogorov, and Andrew Blake.“GrabCut”: Interactive foreground extraction using iterated graph cuts.ACM Transactions on Graphics: Proceedings of the 2004 SIGGRAPH Conference,23(3):309–314, August 2004.

Paul Viola and Michael Jones.Rapid object selection using a boosted cascade of simple features.In Proceedings of the IEEE Computer Society Conference on Computer Vision andPattern Recognition, pages 511–518, Hawaii, December 2001.


Documents

Vision for Automation...Lane departure warning systems. Head tracking systems for drowsiness detection. Driver assistance systems. Reading automobile license plates, and trafﬁc management