ROAD MARKER CLASSIFICATION USING GEOMETRICAL FEATURES WITH SLOPE CONTOUR

ZAMANI BIN MD SANI, ZAMANI (2019) ROAD MARKER CLASSIFICATION USING GEOMETRICAL FEATURES WITH SLOPE CONTOUR. Doctoral thesis, MULTIMEDIA UNIVERSITY.

Text ROAD MARKER CLASSIFICATION USING GEOMETRICAL FEATURES WITH SLOPE CONTOUR.pdf Restricted to Registered users only Download (22MB)

Abstract

Driving down and up the roads requires safe manoeuvre to avoid collisions or congestions, subject to the traffic and road conditions, which are regulated by the road markers painted along the roads. As the research in the road lane departures for intelligent transport systems gains huge interest in the research community, the road markers classification becomes essential. One of the challenges in road marker classification is the varying locations of the road markers between different areas and countries. When classification is carried out using the existing methods, the road marker type transition causes relatively significant errors and delays as compared with the classification over the road stretch without road marker type transitions. Furthermore, the road marker types which look alike such as the dashed-solid (DS) and solid-dashed (SD), are observed to be easily confused, causing more detection failures than the other road marker types. In order to address these problems, an improved camera positioning method to identify a customised area for the Region of Interest (ROI) has been proposed by using the reference of the Field of View (FOV). In addition to the existing five road marker types, namely, Single Solid (SS), Double Solid (DD), Dashed (D), Solid-Dashed (SD) and Dashed-Solid (DS), three novel twolayer classification algorithms were proposed. This was done by virtue of relatively small ROI identified from the road images which were captured by vehicle-mounted cameras. The proposed algorithms extract several key geometrical features such as the total number of contours, Nc(t), the angle, �, between the centroid of two contours, binary transition percentage, Nc p and the slope of the contours, ϕ , for classifying the road markers. By using the two-layer classification algorithms, it is observed that the achievable accuracy value, when measured against the ground truth, is achieved at 98.64%, which is higher than that of the existing method. The transition delay has been further improved by 50% faster than the existing method when using the standard 30 fps camera. The algorithms have also produced better accuracy results when classifying the two look-alike classes, namely SD and DS, achieving accuracy improvement between 2% and 12% in comparison with the existing method.

Item Type:	Thesis (Doctoral)
Subjects:	T Technology > TA Engineering (General). Civil engineering (General)
Depositing User:	Encik Mohd Zulkarnain Hassan bin Mohd Zainudin
Date Deposited:	16 Jan 2026 09:59
Last Modified:	16 Jan 2026 09:59
URI:	https://repositori.mohe.gov.my/id/eprint/145

Actions (login required)

View Item