Accelerating Image Fusion Algorithms Using CUDA on Embedded Industrial Platforms Dedicated to UAV and UGV

Abstract

The objective of image fusion is to combine information from multiple images of the same scene, where the image result of fusion is more suitable for human and machine perception. This process is usually time consuming on most standard computing platforms that can be used in autonomous and semi-autonomous platforms, such as an unmanned ground vehicle (UGV) and an unmanned aerial vehicle (UAV). This creates a conflict between satisfying the hardware requirements and the platform’s current demand. In this article, we propose a solution to this problem, being an image fusion algorithm utilizing CUDA technology, dedicated for an embedded platform (~5 W), allowing for image processing in 25 fps, which should be satisfactory for the needs of the mentioned platforms. During the development, we researched multimodal image fusion algorithms and implemented the chosen methods. The chosen testing environment and chosen measures of the quality of the fusion are presented as well.