Accelerated hyperspectral image recursive hierarchical segmentation using GPUs, multicore CPUs, and hybrid CPU/GPU cluster

摘要

Rescue missions, military target detection, hazard prevention, and other time-critical remote-sensing applications require real-time or autonomous decision making and onboard processing capabilities. Thus, lightweight, small size, and low-power-consumption hardware is essential for onboard real-time processing systems. With the increasing need for dimensionality, size, and resolution of hyperspectral sensors, additional challenges are posed upon remote-sensing processing systems, and more capable computing architectures are needed. Graphical processing units (GPUs) emerged as promising architecture for light-weight high-performance computing. In this paper, we propose accelerated parallel solutions for the well-known recursive hierarchical segmentation (RHSEG) analysis algorithm, using a GPU, hybrid multicore CPU with GPU and hybrid multicore CPU/GPU clusters. RHSEG is a method developed by the National Aeronautics and Space Administration, which is designed to provide more useful classification information with related objects and regions across the hierarchy of output levels. The proposed solutions are built using the NVidia's compute unified device architecture and Microsoft's C++ Accelerated Massive Parallelism (C++ AMP) and are tested using NVidia GeForce hardware and Amazon Elastic Compute Cluster (EC2). The achieved speedups by parallel solutions compared with CPU sequential implementations are 21x for parallel single GPU and 240x for hybrid multinode computer clusters with 16 computing nodes. The energy consumption is reduced to 74 % when using a single GPU, compared to that for the equivalent parallel CPU cluster.