Performance evaluation for curb detection problem

Curbs are important cues identifying the boundary of a roadway. Their detection is required by many automotive features. Currently there is no an agreed benchmark to report and compare curb detection results. This paper presents annotation and performance evaluation toolboxes (PET) developed by us for measuring the performance of curb detection algorithms. The PET is independent of particular sensor systems and is operating on planar curves or projection of 3D curb curves on any plane. We introduce two new criteria measuring directly deviation between the found and annotated curb curves. Usage of the distance transform to measure curb detection performance leads to very fast computations. This allows us to compare different sensor systems and analyze sensitivity of curb detection algorithms to different parameters and surrounding conditions.