Īlthough the existing hyperspectral target recognition algorithms implemented on GPU have a good performance in reducing the operation time, there are few methods to consider synchronizing imaging, data transmission and target recognition for real-time push-broom hyperspectral imagery target search (RT-PBHSI-TS). Benefits from GPU’s ability to support Device Overlap, in this paper, we propose a new RT-PBHSI-TS method, which is implemented on GPU utilizing CUDA streams. The proposed method involves two main phases: (1) real-time hyperspectral target recognition (2) synchronizing imaging, data transmission and target recognition. Spectral angle mapping (SAM) and Euclidean distance (ED) are used for spectral matching implemented on GPU to realize real-time target recognition, and then, Device Overlap utilizing CUDA streams is applied to synchronize imaging, data transmission and target recognition. the experimental results show that the real-time target recognition algorithm speedups over 20x using GPU architecture by NVIDIA GeForce GTX745 compared to CPU implementation at same recognition accuracy that when the false alarm rate is 10-4, the recognition accuracy can reach 93.46%. Finally, the execution efficiency of the real-time target recognition algorithm is analyzed, and the process of RT-PBHSI-TS method is introduced in detail, which provides a reference for further application of target search in the fields of civilian search and rescue, dangerous substances investigation and so on. In Section 2, the real-time hyperspectral target recognition algorithm is introduced in detail.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |