Abstract: A method for reconstructing the three-dimensional temperature field of an aircraft model surface is proposed to address the issues of low resolution of infrared images and limited testing angles in wind tunnel tests, which result in insufficient accuracy in reconstructing the surface temperature field of the aircraft. Firstly, a feature point recognition algorithm based on multi-frame infrared images is adopted to identify and locate the feature points; Secondly, the temperature distortion area of the aircraft model is filtered through the panel selection method; Finally, the multi view temperature data obtained from wind tunnel tests are assigned using temperature assignment. The experimental results show that compared to conventional single-frame recognition methods, the multi-frame processing algorithm improves feature point positioning accuracy by 25%. Under low-resolution infrared imaging conditions (≤ 640 pixel ×512 pixel), the proposed method achieves a feature point recognition rate of 96.875% for 3D temperature field reconstruction on aircraft surfaces, surpassing both normalized cross correlation and convolutional neural network-based algorithms (93.75%). This advancement significantly enhances the precision of the reconstructed 3D temperature field.