During on-orbit operation, space cameras are exposed to complex thermal environments. Non-uniform variations in the structural temperature field can induce thermoelastic deformation, leading to line-of-sight (LOS) pointing deviations and significantly degrading imaging accuracy and stability. To address the insufficient robustness of the traditional Least Squares (LS) method in analyzing LOS pointing stability of space cameras under complex thermal conditions, this paper proposes a thermal line-of-sight pointing analysis method based on the Iteratively Reweighted Least Squares (IRLS) algorithm. First, a thermo-structural coupled model of the space camera is established to analyze the mapping relationship between temperature field variations and LOS pointing deviation. Then, the IRLS algorithm is introduced to perform robust estimation of model parameters. By constructing a weighted residual function, the influence of abnormal measurement data on parameter identification is effectively suppressed, thereby improving the prediction accuracy of thermal deformation. Meanwhile, an energy-iterative window adaptive centroiding algorithm is adopted to capture the variation of spot centroid positions with temperature changes. To investigate thermally induced pointing drift of the on-orbit camera, thermal experiments are conducted. Simulation results are further validated using ground-based thermal test data, and the performance of the proposed IRLS method is compared with that of the traditional LS method in terms of pointing error prediction accuracy and convergence characteristics. The results demonstrate that the proposed IRLS-based thermal analysis method significantly improves the prediction accuracy of LOS pointing deviation in the presence of measurement noise and outliers, while enhancing the stability and robustness of the model. This approach provides an effective technical solution for on-orbit thermal deformation compensation and accuracy maintenance of high-resolution space cameras.