Tour Guide Robot Project by LeeGreenwood

VIEWS: 215 PAGES: 17

									Tour Guide Robot Project

Face Detection and Face Orientation
  on The Mobile Robot Robotino

        Gökhan Remzi Yavuz
          Ayşenur Bilgin

 Project Requirements
 Project Progress
 Human   Computer Interaction - HCI
 Detecting human faces
 Detecting interested people
   Face detection
   Face orientation and face rotation

 Keeping  people interested by interacting
  with them
 Getting closer when a certain distance is
   Robotino: a mobile robot platform from Festo

                                          9 infrared sensors
                                          3 omnidirectional drive units
                                          Bump sensor
                                          Color camera

                                          Control Unit: a PC with a 300 Mhz
                                          Processor and 128 MB SDRam

To satisfy project's requirements, we are planning some extensions on Robotino:
   Cylindirical body, total height: 1.60m including camera

   A laptop computer, attached to body at 1.20m

   Second color camera on top of the laptop computer dedicated to HCI
Hardware con’t
• Sketch of hardware
  modifications on the
  robot base Robotino
• 1 additional laptop
  computer, 2 additional
  color cameras, skeleton
  with steel sticks and not
  yet sketched cylindirical
  cover for better
Project Requirements
A tour guide robot should:

   be able to navigate autonomously and purposefully inside the building

   operate in a dynamic environment

   attract visitors' attention by interacting with them

   be able to monitor its charge level and move to its charging dock on its own

   be able to dynamically arrange tours by combining selected destinations

   deliver information in forms of visual, audio and text messages

 be able to synchronize with the CMPE web site, in order to get up to date
Project Progress
 What we   did?
   Architecture
   Pseudocode
   Examples

 What we did not?
 What we encountered?
 What we achieved?
What we did?
Combining   Robotino API and OpenCV
Modifying the open source code to satisfy
our needs
Running the modified algorithm on the
mobile robot
Testing the behavior
   Application              OpenCV
                                          Watchdog Signal
                                          Action Requests
                                          Camera Requests
                                          Camera Data
             RobotinoAPI                  Action Requests
                                          Camera Data
                 Wireless            OpenCV
                                          Image Data
                                          Face Information

Connect to Robotino over Wireless
While bumpsensorvalue = false do

Retrieve camera image from RobotinoAPI

Process image
      . OpenCV face detection using haarcascades, frontal and profile faces
      . Compute approximate distances of faces and their angle to the robot

Decide action; rotate, move forward, attract attention, stop


end While
Examples cont’d
What we did not?

 Pose estimation
 Tracking
 Face recognition
 Completion of hardware modifications
What we encountered?

 Major challenge : Erroneous detection of
 Processing power
 Minor problems :
       Relatively low (10 fps) performance when
        frontal and profile face detection algorithms
        are combined
       Position of the camera
What we achieved?

 Experiments   held in laboratory
 Moving towards the human face until the
  distance is reduced to 1 meter
 Turning with respect to the weighted
  average of areas of detected faces
Thank you!

To top