Lin, Jianfan, Joseph Casillo, and Mark Paulik
A multi-stage image processing algorithm is introduced to assist with automatic identification of road lanes and obstacles. This task is critical for the development of safe and effective self-driving or autonomous vehicles. Monocular color images captured with a video camera provide system input. These are initially processed to improve contrast and illumination characteristics. Next, color-aware edge and region enhancement algorithms prepare the raw frames for subsequent segmentation. Local and global methods utilizing spatial derivatives, adaptive thresholding and binary morphology are combined with knowledge-based heuristics to segment and score image features to facilitate classification. Algorithm performance is demonstrated on representative image sets.