Hybrid-Convolution-Based Reconstruction for Limited-View Emission Spectrum Tomography

A hybrid neural network model based on 3D-2D convolution tandem is proposed as the spatial feature extractor to overcome the problem of low accuracy of conventional iteration reconstruction algorithm in the case of limited optical windows and projection views in practical flame reconstruction. In this model, 3D convolution is utilized to extract spatial features from multi-view projections simultaneously, and 2D convolution is used to further accelerate the training speed and reduce computational consumption. Compared with conventional iteration reconstruction algorithm and reconstruction algorithms based on residual networks, the proposed model has the advantages of high reconstruction accuracy and low time consumption. It shows potential in flame on-line monitoring and rapid reconstruction.

Automated summary

A hybrid neural network model based on 3D-2D convolution tandem is proposed to overcome the problem of low accuracy in practical flame reconstruction. The model uses 3D convolution to extract spatial features from multiple views and 2D convolution to accelerate training speed. Compared to traditional algorithms, the proposed model achieves higher accuracy with lower time consumption.