Fusion of Lidar, Hyperspectral and RGB Data for Urban Land Use and Land Cover Classification

Abstract

In this paper, we present an ensemble-based classification approach for urban land use and land cover classification based on multispectral LiDAR, hyperspectral and very high resolution RGB data. The approach has been evaluated on the data set provided for the IEEE GRSS 2018 Data Fusion Contest organized by the GRSS IADF technical committee and has been proven to have a high operational performance, being able to distinguish between different grass-, building- and street-types among other classes like water, railways and parking lots as well as other non-typical classes like cars, trains, stadium seats, etc.