presented by: doron brot, maimon vanunu, elia tzirulnick supervised by: johanan erez, ina krinsky,...

32
Presented by : Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by : Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department of Electrical Engineering,Technion

Post on 20-Dec-2015

218 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Presented by:

Doron Brot, Maimon Vanunu, Elia Tzirulnick

Supervised by:

Johanan Erez, Ina Krinsky, Dror Ouzana

Vision & Image Science Laboratory, Department of Electrical Engineering,Technion

Page 2: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Steps to achieve the goal

Aims and motivation of the project

Algorithm for Traffics Signs Recognition

Results

Conclusions

Page 3: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Control a self navigating vehicle according to traffic signs

Page 4: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department
Page 5: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

• Build a controllable vehicle.

• Attach a wireless camera to the vehicle.

• Capture pictures from camera to computer.

• Analyze the visual data and translate it into controlling commands for the vehicle

Page 6: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Build a controllable vehicle.

Mindstorms Robot Mindstorms Robot Invention System Invention System

Page 7: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Build a controllable vehicle

Page 8: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Build a controllable vehicle.

Communication to PC through Infra-red transmitter.

Page 9: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Attach a wireless camera to the vehicle.

WAT-207CD CCD Color Camera

Wireless video transmitter

Page 10: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Attach a wireless camera to the vehicle.

Wireless connection between camera and PC

Page 11: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Capture pictures from camera to computer.

VideoOCX® software can handle all kinds of ‘Video for Windows’ ® compatible devices.

Flyvideo 98 video capture card

Microsoft Visual Basic 6.0

Phantom- a set of functions for the VB 6.0 that helps us control the LEGO™ vehicle.

Page 12: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Analyze the visual data and translate it into controlling commands for the vehicle

Calculate the distance of the vehicle from traffic sign.

Capture one frame from the video camera.

Decide whether there is a traffic sign in the frame or not.

If there is, recognize the traffic sign.

NO

X

25 cm

GO RIGHT

If vehicle is close enough to the sign

send control command to RCX.

YES

Page 13: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

• How humans see colors.

• Conversion from RGB to HSV color space.

• Use Saturation in order to find colored areas in frame.

• Analyze the colored areas according to Hue.

• Recognize traffic sign.

• Sum colored pixels to calculate distance to the traffic sign.

• Send control command according to recognized sign.

Page 14: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

The human eye

Page 15: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Visible Light

Page 16: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

The Retina

שני סוגי קולטנים:

(Rodsקנים )•

(Conesמדוכים )•

Page 17: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

הקולטנים ברשתית

-http://wwwמקור: cvrl.ucsd.edu/

Page 18: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department
Page 19: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Image representation in computer file – graylevel image.

Page 20: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Image representation in computer file – color image.

Page 21: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

RGB values of traffic sign images

Not very helpful !

Page 22: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Conversion from RGB to HSV color space.

The HSV color space )hue, saturation, value( is often used by people because it corresponds better to how people experience color than the RGB color space does.

Page 23: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

As hue varies, the corresponding colors vary from red, through yellow, green, cyan, blue, and magenta, back to red.

Understanding HSV color space

Page 24: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

As saturation varies, the corresponding colors )hues( vary from unsaturated )shades of gray( to fully saturated )no white component(.

Understanding HSV color space

Page 25: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

As value, or brightness, varies, the corresponding colors become increasingly brighter.

Understanding HSV color space

Page 26: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Use Saturation in order to find colored areas in each frame.

Page 27: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Analyze the colored areas according to Hue.

Recognize traffic sign. For example:

If the hue value of any pixel is between 200 and 250 that means that the color is red so we painted the pixel pure red.

Page 28: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Sum colored pixels to calculate distance to the traffic sign.

Send control command according to recognized sign.

If number of colored pixels suits a known sign, in a sufficient distance

Example:

Blue – 434 pixels

Red – 591 pixels

Page 29: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

Graphical User Interface - GUI

Page 30: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

The navigating vehicle.

Page 31: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

• Successful recognition of traffic signs of different colors.

• White – gray background was helpful.

• For real traffic sign recognition more sophisticated algorithms have to be used )colored background, real-time processing etc(.

• The vehicle can only recognize the traffic signs we programmed it to )“Turn Right”, “No Parking” and “Stop”(.

Page 32: Presented by: Doron Brot, Maimon Vanunu, Elia Tzirulnick Supervised by: Johanan Erez, Ina Krinsky, Dror Ouzana Vision & Image Science Laboratory, Department

http://visl.technion.ac.il/projects/scitech02

We would like to thank our mentors: Johanan Erez, Ina Krinsky and Dror Ouzana.

Thanks to our counselors: Adva, Eran , May-Tal and koby.

Thanks to Ort Management.

We would also like to thank the Ollendorff Research Center for its support.