Off-Road Lane Detection Using Superpixel Clustering And RANSAC Curve Fitting

Lane detection is the most important issue to be resolved for successful locomotion of Intelligent Ground Vehicles (IGV). Problems in lane detection often occur in an external setting mainly due to glare or shadow defects. A robust and real-time approach to off-road lane marker detection for IGVs is being presented here. A novel model fitting based lane detection algorithm has been developed. Linear combination of image planes is used which removes the background and uncovers the white lanes. Simple Linear Iterative Clustering is applied to the processed frame and essential thresholding is performed for noise reduction. Two operations namely a novel approach for lane model identification and estimation of chosen lane mode using RANSAC are followed in sequence on the obtained image. The proposed image processing pipeline has been successfully validated in outdoor field conditions.