This study proposed a robust estimation of thermal sensation for vehicle occupants using seat-belt-type sensors and thermal cameras. The estimation of thermal sensation of a vehicle faces challenges, including the difficulty in measuring the clothing insulation value of vehicle occupants and influence of airflow from a car air conditioner. To obtain clothing insulation, we applied a dual-heat-flux method using two heat flux sensors mounted on a seat belt. To overcome the influence of airflow, we proposed a new air flow robust model by utilizing the clothing insulation value and skin temperature of the belt-type sensor, and the clothing surface temperature obtained by a thermal camera. The clothing insulation value estimated by the belt-type sensor showed good agreement with a T-shirt (0.39 clo) and sweatshirt (0.83 clo), with errors of 0.01 or less. An experiment with 11 subjects was also conducted in an environment with variable airflow that simulated a passenger compartment. The estimated thermal sensation showed high accuracy with subjective values (R=0.74, RMSE=1.19), compared to the conventional method (R=0.63, RMSE=1.36), and a significant difference was obtained using Student's t-test (t=4.49, p<0.01). Thus, the proposed method is likely to be employed for vehicle occupants in the future.