Abstract: Multi-spectral Light-field (MSLF) thermometry captures multi-band radiation data using a single light-field camera, eliminating the need for complex spectral separation systems. It enables surface temperature measurements in confined high-temperature environments. This study employs a reference temperature mathematical model and the Chameleon Swarm Algorithm (CSA) for temperature inversion, avoiding emissivity assumptions and reducing dependence on precise emissivity values. However, this advantage lacks experimental validation. To address this, plates with different emissivities were heated by a high-temperature flame. A multi-spectral light-field imager measured their surface temperatures, and inversion results were compared with thermocouple data. The test plates included alumina oxide (Al₂O₃, emissivity 0.20～0.25) and zirconia (ZrO₂, emissivity 0.4～0.5). Experimental results confirm that MSLF maintains stable accuracy across varying emissivities, demonstrating robustness against emissivity uncertainty.