A Robust Detection Method for Multilane Lines in Complex Traffic Scenes

Table 2

The various scene information of the CULane dataset.

Scene	Scene description	Image proportion (%)
Normal	The driving vision is good, and the lane markings are clearly visible	27.7
Crowd	A large number of vehicles are moving slowly on the road, and the lane line is blocked by vehicles	23.4
Night	Visibility is low at night, lane lines are blurred, and different lights interfere with each other	20.3
No line	No lane markings, usually narrow roads, and many vehicles parked on the side of the road	11.7
Shadow	Under the sky bridge or viaduct, the light is dark, the shade of the trees or the shadow of the building	2.7
Strong light	Various strong lights cause roadside reflections, and the lane lines on the road are not clear	1.4
Curve	Curved lane line ahead	1.2
Crossroad	Interference of zebra crossing markings at intersections, no lane markings in the middle of the intersection	9.0
Arrow	There are interference marks between lane lines such as going straight, turning, and turning around	2.6