Research on a computational spectral splitting-based method for reconstructing mid-wave infrared hyperspectral radiant brightness

Mid-wave infrared hyperspectral radiant brightness (MWIR-HRB) is a key parameter in radiometric characterization measurements. Traditional hardware-based MWIR hyperspectral systems often face a trade-off between temporal and spatial resolution, resulting in limited detection capability for dim infrared targets. To address the aforementioned issues, this paper proposes a computational spectral splitting-based method for reconstructing MWIR-HRB. Firstly, according to Planck’s law and compressed sensing theory, a MWIR-HRB reconstruction model based on computational spectral splitting is proposed. Secondly, in order to achieve sparse representation of hyperspectral radiant brightness, an overcomplete dictionary design method based on Online Dictionary Learning (ODL) is proposed. At the same time, in order to reconstruct hyperspectral radiant brightness from five-channel broadband radiant brightness measurements, a hyperspectral radiant brightness reconstruction method based on Batch Orthogonal Matching Pursuit (Batch-OMP) is proposed. Then, in order to establish the connection between the hyperspectral overcomplete dictionary and the measurement system, a set of solutions for the design of the multispectral measurement matrix and the reconstruction of the hyperspectral overcomplete dictionary are proposed. Finally, validation experiments are conducted to verify the feasibility of the proposed method. The experimental results show that the proposed method can effectively reconstruct the MWIR-HRB of different materials, and the maximum relative error of the reconstructed results is consistently below 2 %. In addition, the proposed method does not require emissivity as an explicit prior input and can still accurately reconstruct MWIR-HRB under unknown-emissivity conditions.