Three-dimensional flame temperature reconstruction through adaptive segmentation-weighted non-negative least squares and light field imaging

Existing flame temperature reconstruction algorithms experience significant performance degradation when subjected to radiation intensity noise interference, resulting in limited accuracy in low-temperature regions of flames with broad temperature distributions. We propose a three-dimensional (3D) flame temperature reconstruction algorithm by integrating adaptive segmentation-weighted non-negative least squares with light field imaging. Building on the non-negative least squares framework, the proposed algorithm introduces an adaptive strategy to improve the temperature reconstruction accuracy of low-temperature regions of flames. It also incorporates adaptive weight factors to reduce measurement errors caused by the extensive temperature range, enabling precise 3D temperature reconstruction. To validate the algorithm, numerical simulations of a bimodal asymmetric flame were performed to evaluate its noise tolerance and compare its performance with other existing algorithms. The simulation results indicated that the proposed algorithm demonstrates strong noise resistance, achieving ∼70% higher reconstruction accuracy than the least-square QR decomposition algorithm. Experiments were carried out on bimodal flames to reconstruct the temperature under various combustion conditions. The reconstructed temperature showed good agreement with trends reported in the literature. Our results demonstrate the viability and robustness of the proposed algorithm for reconstructing broader temperature distributions of flames.