Convolutional neural network based radio frequency tomographic reconstruction is explored in this study. Due to the limited amount of sensors used in RF tomographic imaging system, analytical reconstruction such as filtered back-projection usually generates strong artifacts in the reconstructed images. The state of art algebraic reconstruction methods use prior knowledge to regularize reconstruction and reduce artifacts, but suffer from high computational complexity. Our study shows reconstruction based on convolutional neural network, a learning based approach, is effective in removing artifacts caused by limited number of sensors, and has low computational cost, which makes it suitable for real-time applications.
Radio Frequency Tomographic Reconstruction Based on Convolutional Neural Networks
2018-07-01
1413610 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Springer Verlag | 2017
|Efficient Binary Tomographic Reconstruction
British Library Online Contents | 2014
|Auroral Tomographic Reconstruction Technique
British Library Online Contents | 1995
|Multiresolution Tomographic Reconstruction using Wavelets
British Library Conference Proceedings | 1994
|GPS-based tomographic reconstruction of the ionosphere
British Library Conference Proceedings | 1998
|