Efficient Classification of Hyperspectral Images on Commodity GPUs using ELM-based Techniques

TítuloEfficient Classification of Hyperspectral Images on Commodity GPUs using ELM-based Techniques
AutoresJavier López-Fandiño, Dora B. Heras, Francisco Argüello
TipoComunicación para congreso
Fonte The 2014 International Conference on Parallel and Distributed Processing Techniques and Applications, Las Vegas, USA, 2014.
AbstractHyperspectral image processing algorithms are computationally very costly, which makes them good candidates for parallel and, specifically, GPU processing. Extreme Learning Machine (ELM) is a recently proposed classification algorithm very suitable for its implementation on GPU platforms. In this paper we propose an eficient GPU implementation of an ELM-based classification strategy for hyperspectral images. ELM can be expressed in terms of matrix operations that can take maximum advantage of the GPU architecture. Regarding the classification accuracy, the proposed algorithm achieves competitive results as compared to a traditional SVM strategy with significantly lower running times. Additionally, the use of a voting mechanism to improve the accuracy results is also considered.
Palabras chaveHyperspectral images, ELM, SVM, GPU, CUDA.