A Distributed System for Multiscale Feature Extraction and Semantic Classification of Large-Scale Lidar Point Clouds

Managing and processing large-scale point clouds are much needed for the exploration and contextual understanding of the data. Hence, we explore the use of a widely used big data analytics framework, Apache Spark, in distributed systems for large-scale point cloud processing. To effectively use Spark, we propose to use its integration with Cassandra for persistent storage, and to appropriately partition the point cloud across the nodes in the distributed system. We use this integrated framework for multiscale feature extraction and semantic classification using random forest classifier. We have shown the efficacy of our proposed application through our results in the DALES aerial LiDAR point cloud.