traffic sign pattern recognition

13
Traffic Sign Pattern Recognition Pilho Kim (ECE), Zhaohua Wang, Yichang (James) Tsai (CE)

Upload: adelio

Post on 12-Feb-2016

51 views

Category:

Documents


0 download

DESCRIPTION

Traffic Sign Pattern Recognition. Pilho Kim (ECE), Zhaohua Wang, Yichang (James) Tsai (CE). CONTENTS. Introduction and Motivation Preliminary Works Build the project collaboration environment Construct the traffic sign database from MUTCD Reviews on related works using ANN - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Traffic Sign Pattern Recognition

Traffic Sign Pattern Recognition

Pilho Kim (ECE), Zhaohua Wang, Yichang (James) Tsai (CE)

Page 2: Traffic Sign Pattern Recognition

2

CONTENTS Introduction and Motivation Preliminary Works

Build the project collaboration environment Construct the traffic sign database from MUTCD Reviews on related works using ANN Search proper image abstractions for sign recognition Develop the ANN modules

Proposed Approaches Closed convex polygon detection algorithms for precise

traffic sign region extraction Color-coded line receptors as image features for ANN

Concluding Remarks

Page 3: Traffic Sign Pattern Recognition

3

Introduction and Motivation

Traffic asset management High demands on the efficient and cost-saving traffic asset

management system

Safe driving and autonomous land vehicle (ALV) Driver assistant system can reduce car accidents to save lives of

drivers ALV should have this technology to make it practical

So we need “Automatic geographical traffic sign location and type logging system using computer vision.”

Page 4: Traffic Sign Pattern Recognition

4

Preliminary Works:Build the project collaboration environment TSPR Project Wiki:

http://www.pilhokim.com/project/signpattern/signwiki/

Page 5: Traffic Sign Pattern Recognition

5

Preliminary Works:Construct the traffic sign database from MUTCD Build the traffic sign database from MUTCD (Manual

on uniform traffic control devices) Prepare two sets for the computation and the database

(MySQL)

Page 6: Traffic Sign Pattern Recognition

6

Preliminary Works:Construct the traffic sign database from MUTCD Reviews on related works using ANN

http://www.citeulike.org/user/pilho/tag/sign

Fang et al. (2004)

Page 7: Traffic Sign Pattern Recognition

7

Preliminary Works:Search proper image abstractions for sign recognition Devise the feature correlation graph (FCG)

Template matching Canny edge X-Y profile

Page 8: Traffic Sign Pattern Recognition

8

Preliminary Works:Develop the ANN modules

Choose the proper ANN engine

Page 9: Traffic Sign Pattern Recognition

9

PROPOSED PLAN

Closed convex polygon detectionfor the accurate traffic sign boundary detection

Color-coded line receptors as image features

Page 10: Traffic Sign Pattern Recognition

10

Proposed Approaches:Closed convex polygon detection algorithms for precise traffic sign region extraction

Page 11: Traffic Sign Pattern Recognition

11

Proposed Approaches:Color-coded line receptors as image features for ANN

Line Receptors Line Encoding Example

Enhance above simple image pattern recognition algorithms to : Count on the image color and line crossing features by introducing

introduce the multi-level encoding schema Improve the existing inner and outer entropy computing methods.

Page 12: Traffic Sign Pattern Recognition

12

Concluding Remarks for future investigation

Traffic sign pattern recognition in the real scene capture is very challenging

Finding the proper robust features for the ANN training is the key to solve the problem

Multi-level image processing and recursive result enhancements are required.

Understanding the context of image capturing environment will give clues for recognition

Page 13: Traffic Sign Pattern Recognition

13

Appreciate Your Attention