TY - GEN
T1 - High precision road segmentation for cover level of forward view estimation via stereo camera
AU - Amit, Siti Nor Khuzaimah Binti
AU - Aoki, Yoshimitsu
N1 - Publisher Copyright:
© 2015 IEEE.
PY - 2015/9/8
Y1 - 2015/9/8
N2 - To develop a safe Intelligent Transportation System (ITS) while driving on unpredictable curves or road regions, high precision road segmentation and cover level of forward view estimation for drivers is necessary. Cover level of forward view is defined as the level of difficulty in predicting the dangerousness of road edge or incoming object near road edge especially at a curve due to the obstacles at the surrounding. We focus on road segmentation as it is one of the fundamental steps in developing ITS. According to the previous studies, road region is not segmented precisely; hence a new method of road segmentation is introduced. Input images had undergone illuminant invariant space conversion to remove shadow regions effectively. Next, the bottom part of illuminant invariant image (assumed to be road surface) is sampled to get a road model, which is then applied to obtain the road probability image. Lastly, dilation and erosion using mathematical morphology operation is applied to obtain road region. Our method of road segmentation shows precision of 0.73, recall as 0.82 and F measure as 0.81. High Precision road segmentation is very important for better cover level of forward view estimation. Segmented road region can be fully utilized for curve estimation and obstacles detection and hence lead to a better performance of cover level of forward view. Based on the results obtained, further improvement in several aspects, especially from the input image or illuminant invariant image to road probability image, has to be done in order to obtain perfectly segmented road region.
AB - To develop a safe Intelligent Transportation System (ITS) while driving on unpredictable curves or road regions, high precision road segmentation and cover level of forward view estimation for drivers is necessary. Cover level of forward view is defined as the level of difficulty in predicting the dangerousness of road edge or incoming object near road edge especially at a curve due to the obstacles at the surrounding. We focus on road segmentation as it is one of the fundamental steps in developing ITS. According to the previous studies, road region is not segmented precisely; hence a new method of road segmentation is introduced. Input images had undergone illuminant invariant space conversion to remove shadow regions effectively. Next, the bottom part of illuminant invariant image (assumed to be road surface) is sampled to get a road model, which is then applied to obtain the road probability image. Lastly, dilation and erosion using mathematical morphology operation is applied to obtain road region. Our method of road segmentation shows precision of 0.73, recall as 0.82 and F measure as 0.81. High Precision road segmentation is very important for better cover level of forward view estimation. Segmented road region can be fully utilized for curve estimation and obstacles detection and hence lead to a better performance of cover level of forward view. Based on the results obtained, further improvement in several aspects, especially from the input image or illuminant invariant image to road probability image, has to be done in order to obtain perfectly segmented road region.
KW - cover level of forward view
KW - illuminant invariant image
KW - road likelihood model
KW - road segmentation
KW - seed selection
KW - stereo camera
UR - http://www.scopus.com/inward/record.url?scp=84957685400&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84957685400&partnerID=8YFLogxK
U2 - 10.1109/ASCC.2015.7244743
DO - 10.1109/ASCC.2015.7244743
M3 - Conference contribution
AN - SCOPUS:84957685400
T3 - 2015 10th Asian Control Conference: Emerging Control Techniques for a Sustainable World, ASCC 2015
BT - 2015 10th Asian Control Conference
A2 - Selamat, Hazlina
A2 - Ramli, Hafiz Rashidi Haruna
A2 - Faudzi, Ahmad Athif Mohd
A2 - Rahman, Ribhan Zafira Abdul
A2 - Ishak, Asnor Juraiza
A2 - Soh, Azura Che
A2 - Ahmad, Siti Anom
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 10th Asian Control Conference, ASCC 2015
Y2 - 31 May 2015 through 3 June 2015
ER -