Finding Visual Landmarks with E-Puck robots Loc SEGAPELLI - PowerPoint PPT Presentation

Finding Visual Landmarks with E-Puck robots Loïc SEGAPELLI

Introduction � Webots � A group of e-puck robots � Each one has a camera � They are looking for a blue light � They must use their camera only! � Collaboration

Example

Algorithm � Robot turn 360º , takes 6 photos � One robot sees the blue light: � Turns on its red leds � Waits � Goes to the light source

Algorithm � Another robot sees red leds: � Turns on its red leds -> chain reaction � Waits � Goes to the red light source

Blue light detection 1

Blue light detection 2

Blue light detection 3

Blue light detection 4

Blue light detection 5

Blue light detection 6

Blue light detection 7

Image segmentation � One picture = 1 min! � Get the most from one photo � Vertical image segmentation

Image segmentation � Position correction: robots turn where there is the most blue / red pixels

Image segmentation 1

Image segmentation 2

Image segmentation 3

Image segmentation 4

Detection The whole state machine… Exploration

Real camera pictures � 30x30 pixels, should not be a handicap

Conclusion - No real-robot experimentation + Interesting image segmentation strategy + Simple state machine