OpenCL Portability Study of an Algorithm for Automatically Detecting Targets for Hyperspectral Image Analysis

Abstract

In the last decades, the issue of target detection has received considerable attention on remote sensing applications, where the use of high performance computing (HPC) has been linked. One of the most popular algorithm in target detection and identification is the automatic target detection and classification algorithm (ATDCA) widely used in the hyperspectral image analysis community. Previous research has already investigated the mapping of ATDCA on multicore processors, graphics processing units (GPUs) and accelerators like as field programmable gate arrays (FPGAs), showing impressive speedup factors that allow its exploitation in time-critical scenarios. Based on these studies, this paper explores a portability study of a tuned OpenCL implementation based on performance, energy consumption and code quality parameters compared to hand-tuned versions previously investigated. This approach differs from previous papers, which focused on achieving the optimal performance on each platform. Our study includes the analysis of different tuning techniques that expose data parallelism as well as enable an efficient exploitation of the complex memory hierarchies found in these new heterogeneous devices, as well as measuring the energy consumption on each platform and metrics to analyze the quality of our code. Experiments results demonstrate the importance of performance, energy consumption and code quality parameters applied on synthetic and real hyperspectral data sets collected by the Hyperspectral Digital Imagery Collection Experiment (HYDICE) and NASA’s Airborne Visible Infrared Imaging Spectrometer (AVIRIS) sensors, in order to really calibrate the possibility of using heterogeneous platforms for efficient hyperspectral imaging real-time processing in remote sensing missions.