“When high-sensitivity infrared detectors operate in orbit, their background radiation is modulated by complex time-varying temperature fields, which severely limits the accuracy of radiation calibration. Researchers propose a high-precision real-time estimation method for detector background radiation, and use the random forest algorithm to establish a nonlinear mapping model between temperature field and background radiation. The validation accuracy R ² reaches 0.9854, providing a solution to improve the accuracy of infrared detector in orbit radiation calibration.”