Semantic classification and uncertainty analysis of pedestrians and road geometry using LiDAR point clouds to assist autonomous driving
Project Principal Investigator(s): Jaya Sreevalsan Nair
A research project on determining and quantifying uncertainty in semantic classification of LiDAR point clouds acquired from moving vehicles. Semantic classification is achieved using random forest classifiers. We propose to use state-of-the-art machine learning algorithms for the semantic classification. Subsequent to the classification, we also determine the uncertainty involved in the classification in the spatial context. Objects of interest for autonomous vehicles are pedestrians and road intersections. Hence, we focus on classifying these objects of interests, along with an annotation of uncertainty in
classification. We heavily rely on our previous work on local geometric descriptors to determine the uncertainties.