GPU implementation of Ant Colony Optimization algorithm for endmember extraction from hyperspectral image

Endmember extraction is an important procedure in spectral unmixing of hyperspectral images. A swarm intelligence algorithm named Ant Colony Optimization (ACO) was recently introduced for endmember extraction from hyperspectral images. Through theoretic and experimental analysis, this new algorithm can be utilized in various situations and can get the optimum endmembers. However, considering the high computational amount, the ACO Endmember Extraction algorithm (ACOEE) based on the random search is time-consuming for hyperspectral image analysis. In recent years, high-performance computing based on the Graphics Processing Unit (GPU) has been increasingly applied in hyperspectral image processing. In this paper, we propose a new parallel implementation of the ACOEE based on the GPU. The algorithm is implemented using the Compute Device Unified Architecture (CUDA), and tested on the NVidia Quadro 5000 architecture. The results show that the computational performance of ACOEE based on GPU has significantly improved in the analysis of real hyperspectral data.