Vision-based forward-looking traffic scene analysis scheme

Jyh-Yeong Chang, Chien-Wen Cho

Research output: Contribution to conferencePaperpeer-review

2 Scopus citations


Vision-based driver assistant systems are very promising in Intelligent Transportation System (ITS); however, algorithms capable of describing traffic scene images are still very difficult to date. This paper proposes a system which can segment forward-looking road scene image into natural elements and detect front vehicles. First, the scene analysis system deals with scene segmentation and natural object labeling of forward-looking images. By the use of fuzzy Adaptive Resonance Theory (ART) and fuzzy inference techniques, the scene analysis task is accomplished with tolerance to uncertainty, ambiguity, irregularity, and noise existing in the traffic scene images. Secondly, the proposed system can detect the front vehicles and utilize a bounding box shape to further refine the segmentation result. Compared with conventional approaches, the proposed scheme can analyze forward-looking traffic scenes and yield reliable and efficient segmentation results. The validity of the proposed scheme in car detection was verified by field-test experiments. The traffic scene segmentation and front vehicle detection are successful.
Original languageAmerican English
Number of pages6
StatePublished - 2007
Event2007 IEEE Intelligent Vehicles Symposium, IV 2007 - Istanbul, Turkey
Duration: 13 Jun 200715 Jun 2007


Conference2007 IEEE Intelligent Vehicles Symposium, IV 2007



Fingerprint Dive into the research topics of 'Vision-based forward-looking traffic scene analysis scheme'. Together they form a unique fingerprint.

Cite this