Development of an edge computing device by integrating deep learning model for flame tomographic reconstruction

由于环境法规的日益严格，用于燃烧诊断的光学成像技术在工业环境中变得越来越普遍，以提高燃烧效率和减少污染物排放。体积层析成像(VT)是其中一种有效的燃烧监测和诊断方法，它为燃烧诊断提供了真实的三维信息。VT的发展有助于理解复杂的燃烧过程，如火焰点火和燃烧不稳定性。此外，VT的独特特性包括非侵入性和易于实现，这使得这种光学成像技术适合于燃烧系统的时空监测和诊断。它们包括基于数学逆的解析重建、基于数值逆的迭代重建和基于深度学习的图像重建。

Due to increasingly tight environment legislation, optical imaging techniques for combustion diagnostics are becoming prevalent in industrial environments for improved combustion efficiency and reduced pollutant emissions.Volumetric tomography (VT) which is one of these techniques is an effective methodused for combustion monitoring and diagnostics as it provides authentic 3D information for combustion diagnostics. The development of VT has helped in understanding of complex combustion processes such as flame ignition and combustion instability. Also, the unique features of VT include non-intrusiveness and easy implementation, which has made such an optical imaging technique suitable for the spatial and temporal monitoring and diagnostics of combustion systems.There are various reconstruction approaches suitable for volumetric image reconstruction (Fig 1). They include analytic reconstruction which is based on the mathematical inversion, iterative reconstruction based on numerical inversion and deep learning-based image reconstruction.