Sensitivity Evaluation of Embedded Code Detection in Imperceptible - - PowerPoint PPT Presentation

sensitivity evaluation of embedded code detection in
SMART_READER_LITE
LIVE PREVIEW

Sensitivity Evaluation of Embedded Code Detection in Imperceptible - - PowerPoint PPT Presentation

Sensitivity Evaluation of Embedded Code Detection in Imperceptible Structured Light Sensing Jingwen Dai Ronald Chung Computer Vision Laboratory Dept. of Mech. and Automation Engineering The Chinese University of Hong Kong WoRV2013, Clearwater,


slide-1
SLIDE 1

Sensitivity Evaluation of Embedded Code Detection in Imperceptible Structured Light Sensing

Jingwen Dai Ronald Chung Computer Vision Laboratory

  • Dept. of Mech. and Automation Engineering

The Chinese University of Hong Kong WoRV2013, Clearwater, FL, USA, 16 Jan 2013

slide-2
SLIDE 2

Introduction & Motivation

01/16/2013 WoRV2013 2

 Projector‐Camera Systems in Robotics

 Augmented Reality  Human‐Robot Interaction  Some Other Applications

 Imperceptible Structured Light Sensing (ISL)

Show Video Content Derive the 3D information Display Device 3D Sensor

slide-3
SLIDE 3

Introduction & Motivation

01/16/2013 WoRV2013 3

  • J. Dai and R. Chung, On Making Projector both a Display Device and a 3D Sensor, In Proc.
  • f ISVC12, pages 654‐664, 2012.

 specifically designed shapes  large hamming distance  Pre‐trained shape detector

Coding Decoding

slide-4
SLIDE 4

Robotic Applications –

Sensing Surrounding Environment

01/16/2013 WoRV2013 4

  • J. Dai and R. Chung, Embedding Imperceptible Codes into Video Projection and Applications in

Robotics, In Proc. of IROS12, pages 4399‐4404, 2012.

slide-5
SLIDE 5

Robotic Applications –

Natural Human‐Robot Interaction

01/16/2013 WoRV2013 5

  • J. Dai and R. Chung, Embedding Imperceptible Codes into Video Projection and Applications in

Robotics, In Proc. of IROS12, pages 4399‐4404, 2012.

slide-6
SLIDE 6

Sensitivity Evaluation

01/16/2013 WoRV2013

Training Stage Operation Stage

Scenarios Differences Sensor‐Object Localization Projection Surfaces Surrounding Environment Hardware Platforms

6

slide-7
SLIDE 7

System Setup

01/16/2013 WoRV2013

PROCAMS‐A PROCAMS‐B

7

slide-8
SLIDE 8

Hardware Configuration

01/16/2013 WoRV2013 8

PROCAMS‐A PROCAMS‐B

Projector

Mitsubishi EX240U Projector 1024 * 768 TI DLP Pico Projector Development Kit 2 640 * 480

Camera

Point Grey Flea3 FL3‐U3‐13S2C 1328 * 1048@120fps Point Grey Flea3 FL3‐FW‐03S1C 648 * 488@120fps

Lens

Myutron FV1520 f15mm Myutron FV0622 f6mm lens

Pro‐PC

VGA HDMI

Cam‐PC

IEEE‐1394 USB 3.0

slide-9
SLIDE 9

Benchmark

01/16/2013 WoRV2013 9

Distance

800mm 500mm 1200mm 1600mm

Orientation

0° 10° 20° 30° 40° 50°

Shape

Planar Convex Concave Free‐Form

Texture

White Green Cork Poster

PROCAMS

A B Training Sample Collection Scenario

When the scenarios of training stage and operation stage are almost the same, about 95% primitive shapes can be detected and identified correctly.

slide-10
SLIDE 10

Sensitivity Evaluation: Working Distance

01/16/2013 WoRV2013

500mm 1200mm 1600mm

In training data collection: Working Distance: 800mm

10

slide-11
SLIDE 11

Sensitivity Evaluation: Projection Surface Orientation

01/16/2013 WoRV2013

° ° ° ° °

In training data collection: Surface Orientation: θ 0°

11

slide-12
SLIDE 12

Sensitivity Evaluation: Projection Surface Shape

01/16/2013 WoRV2013

Convex Surface Plaster Statue Concave Surface

In training data collection: Projection Surface: Planar

12

slide-13
SLIDE 13

Sensitivity Evaluation: Projection Surface Texture

01/16/2013 WoRV2013

Green Paper Cork Board Poster

In training data collection: Surface Texture: White Paper

13

slide-14
SLIDE 14

Sensitivity Evaluation: PROCAMS

01/16/2013 WoRV2013

Captured Image Cropped Patt. Resized Patt.

In training data collection: PROCAMS: PROCAMS‐A

14

slide-15
SLIDE 15

Sensitivity Evaluation: Conclusion

Condition Hits (%) Missed (%) False (%) Ed (pixel) Benchmark 94.53 3.95 1.52 1.632 Distance (500mm) 86.21 11.63 2.16 1.814 Orientation (50 degree) 85.91 12.03 2.06 2.728 Surface (Plaster Statue) 84.81 13.33 1.86 2.028 Texture (Poster) 91.74 6.63 1.63 2.024 PROCAMS (Cropped Pattern) 80.23 14.43 5.34 3.028

01/16/2013 WoRV2013

For more detailed sensitivity evaluation results, please refer to the paper

15

slide-16
SLIDE 16

Conclusion and Future Works

Sensitivity evaluation of embedded code detection in imperceptible structured light sensing. Future Works

 Extending imperceptible structured light sensing to a

variety of robotic applications.

01/16/2013 WoRV2013 16

Distance Surface Orientation Surface Shape Surface Texture PROCAMS

slide-17
SLIDE 17

THANK YOU!!

If you have any questions, please contact

  • Dr. Jingwen Dai

dai@cs.unc.edu

01/16/2013 WoRV2013 17