Enhanced Fog Detectiom And Free Space Segmentation For Car Navigation

Abstract

Free- space detection is a primary task for car 
navigation. Unfortunately, classical approach 
have difficulties in adverse weather conditions, in 
particular in day time fog. In this paper, a solution 
is proposed thanks to a contrast restoration 
approach on images grabbed by an in- vehicle 
camera. The proposed method improves the state
of-the-art in several ways. First, the segment of the 
fog region of interest is better segmented thanks to 
the computation of the shortest route maps. Second, 
the fog density as well as the position of the horizon 
lines is jointly computed. Then, the method restores 
the contrast of the road by only assuming that the 
road is flat and, at the same time, detects the 
vertical objects. Finally. A segment of the 
connected components in font of the vehicle gives 
the free-space area. An experimental validation 
was carries out to foresee the effectiveness of the 
method. different results are shown on sample 
images extracted from video sequences acquired 
from an in-vehicle camera. The proposed method 
is complementary to existing free-space area 
detection methods replying on colour segmentation 
and stereovision