Road detection based on scene structural knowledge and CRF

Yanzi Deng,Zhaoyang Lu,Jing Li
DOI: https://doi.org/10.13245/j.hust.160905
2016-01-01
Abstract:Existing approaches classify the road area by using appearance-based features,which is vul-nerable to the complicated imaging conditions such as extreme shadows,illumination and occlusion.A new road detection method was proposed to overcome this problem brought about by these factors, which combined the advantages of the structural knowledge and fully connected conditional random fields (CRFs).Firstly,a confidence map of road was generated based on the detection of the vanis-hing point and road boundaries.Secondly,a scene layout map was estimated by training a regression model using superpixel features.Two maps and appearance features were used to calculate the energy function of the CRF.Finally,road pixels could be labeled using an efficient inference method.Experi-mental results prove the effectiveness of the usage of structure information and a fully connected CRF, and the proposed method is robust to the shadows and occlusion in real road scenes.
What problem does this paper attempt to address?