31 0

Traffic Light Detection and Recognition based on Haar-like Features

Title
Traffic Light Detection and Recognition based on Haar-like Features
Author
임준홍
Keywords
Haar-like Feature; Image processing; Object detection; Self-driving vehicles; SVM
Issue Date
2018-01
Publisher
IEEE
Citation
2018 International Conference on Electronics, Information, and Communication (ICEIC), Page. 251-254
Abstract
The problem of traffic light detection and recognition is investigated in this paper. Most algorithms used in traffic light detection and recognition are based on color detection. The color-based approach has some difficulties in that if the color of the traffic lights is changed by external factors, they will not be recognized and errors will occur. We propose an algorithm for traffic light detection and recognition based on Haar-like features in this paper. We use Haar-like features to learn about the traffic light image and detect the candidate area based on the learning data. The detected candidate image is verified by the pre-learned SVM(Support Vector Machine) classifier, and binarization and morphology operations are performed on the verified candidate image for detection of the traffic light object. The detected traffic light is divided into respective signal areas to determine the current on/off status of traffic lights. The signal signs in the respective areas are defined by regulation and the sign of traffic lights can be recognized by recognizing on/off of the signals in the respective areas. The experimental study is performed to show that it is possible to detect and recognize traffic lights irrespective of color changes.
URI
https://ieeexplore.ieee.org/document/8330598http://repository.hanyang.ac.kr/handle/20.500.11754/105319
ISBN
978-1-5386-4754-7
DOI
10.23919/ELINFOCOM.2018.8330598
Appears in Collections:
COLLEGE OF ENGINEERING SCIENCES[E](공학대학) > ELECTRICAL ENGINEERING(전자공학부) > Articles
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE