Traffic Sign Recognition

1
Traffic Sign Recognition

• Jacob Carlson Sean St. Onge
• Advisor Dr. Thomas L. Stewart

2
Traffic Sign Recognition

• Project Overview
• System Description
• Current Functionality
• Future Work

3
Traffic Sign Recognition

• Project Overview
• System Description
• Current Functionality
• Future Work

4
Project Overview

• Object identification has many applications in
  various fields. This project aims to identify a
  traffic sign from a digital image. This would be
  useful in an autonomous vehicle application.
  These ideas and methods could also be used in
  other areas.

5
Project Overview

• The overall objective of this project is to write
  a program what will identify a traffic sign from
  a digital photograph.
• Traffic signs appear in diverse background
  situations and, at times, may be partially
  obscured.

6
Traffic Sign Recognition

• Project Overview
• System Description
• Current Functionality
• Future Work

7
System Description
8
System Description

• When the program is initialized, an image,
  previously saved on the systems hard drive, is
  loaded for analysis.
• At this point, some preliminary analysis will be
  performed, and preprocessing will be performed
  manually.

9
System Description

• This portion of the program will gather and
  analyze color data, and will also perform edge
  detection.

Red Green
Blue
10
System Description

• Additional methods (dilation, opening, closing,
  erosion) may also be applied at this time.
• The sign will be classified based on color.

11
System Description

• After classification, the software will highlight
  the image or area of interest.
• The software will then write pertinent data to
  either the screen or an output file.

12
Traffic Sign Recognition

• Project Overview
• System Description
• Current Functionality
• Future Work

13
Current Functionality

• Currently our program divides the color image
  into the three color planes.
• We first look for red signs (stop sign, do not
  enter, wrong way). Our algorithm currently
  isolates most red signs effectively.
• It can also isolate yellow signs, but this still
  requires some optimization.

14
Current Functionality

• Initial Image

15
Current Functionality

• Red Plane

16
Current Functionality

• Red Plane, after Thresholding

17
Current Functionality

• Green Plane

18
Current Functionality

• Blue Plane

19
Current Functionality

• Threshold red plane after median filter.

20
Current Functionality

• Sobel Masks Used for edge detection
  (differentiation).

21
Current Functionality

• Horizontal Edge Detection using Sobel masks.

22
Current Functionality

• Vertical Edge Detection using Sobel masks.

23
Current Functionality

• Sum of horizontal and vertical edge detection.

24
Current Functionality

• Image after erosion by a line structuring element.

25
Current Functionality

• Image after closing with octagon structuring
  element.

26
Current Functionality

• Stop sign identified using blob recognition
  techniques.

27
Current Functionality

• Final image with stop sign highlighted.

28
Traffic Sign Recognition

• Project Overview
• System Description
• Current Functionality
• Future Work

29
Traffic Sign Recognition

• Current problem is having the computer recognize
  that the shape is a stop sign.

30
Traffic Sign Recognition

• Identifying a region of interest and cropping out
  the background prior to performing main
  processing would streamline calculations.
• Speed could also be increased by using C or C
  to implement the processing algorithms.

31
Questions?