Shadow detection and removal for illumination consistency on the road

Shadows on the road always trouble vision tasks like visual navigation and road detection. Shadows will change road characteristics and occlude road objects. It is a great challenge to effectively detect and remove the shadows on the road to maintain illumination consistency for the vehicle. To tackle the adverse effect caused by shadows on the road, this paper attempts to detect shadows with Support Vector Machine (SVM) based on color saliency space and gradient field. Shadowed areas are distinguished and recognized by nonlinear SVM classifier through reconstructing road shadow descriptor after analyzing its color saliency space and gradient information. Then adaptive variable scale regional compensation operator is adopted to remove the shadows. Extensive experiments show that the shadow detection and removal method proposed in this paper has good feasibility and adaptability, and the method performs well under a variety of road environment.