Thermal offset in Enhanced Reference System (ERS) points under non-uniform temperature fields is one of the primary factors causing reduced coordinate registration accuracy between the tooling measurement coordinate system (MCS) and the global coordinate system (GCS). Addressing this issue is critical for ensuring precision in large-scale metrology and aircraft assembly. This study aims to address the challenges posed by sparse measurement points, complex temperature distributions, and finite element model simplifications, which collectively limit the accuracy of temperature field modeling and thermal offset prediction. A novel method integrating global temperature field reconstruction, virtual measurement point dynamic optimization, and coordinate registration error compensation was developed to improve the accuracy and reliability of thermal offset prediction and registration.A novel methodology integrating virtual measurement point dynamic optimization and precise temperature field reconstruction was proposed to address the challenges of thermal offset prediction and coordinate registration under non-uniform temperature fields. First, a global temperature field reconstruction algorithm (GPR-GTFR) based on Gaussian Process Regression was developed to accurately capture the spatial distribution characteristics of complex non-uniform temperature fields, providing high-quality input data for subsequent thermal offset prediction. Next, a thermal offset prediction model was established using the finite element method (FEM) to quantitatively analyze the thermal displacement of ERS points under varying temperature conditions. To overcome the limitations of sparse measurement point distributions and inaccuracies caused by FEM model simplifications, a virtual measurement point dynamic optimization strategy was proposed. This strategy incorporated an improved Adaptive Dynamic Local Search Prairie Dog Optimization algorithm (ADLSPDO) to dynamically optimize the spatial layout of the temperature field, enhancing the spatial coverage and accuracy of temperature field reconstruction. Finally, the predicted thermal offset values were used to correct the nominal positions of ERS points, effectively reducing the coordinate registration errors between the tooling measurement coordinate system (MCS) and the global coordinate system (GCS). The proposed methodology was validated through simulations and experiments to verify its effectiveness in addressing thermal offset challenges.The proposed method was evaluated through a series of experiments and simulations, and its effectiveness was verified at multiple levels. The GPR-GTFR algorithm demonstrated superior accuracy and robustness compared to traditional Kriging methods in reconstructing temperature fields, particularly under sparse measurement conditions (Fig.6). The introduction of dynamically optimized virtual measurement points significantly improved the spatial coverage and accuracy of temperature field modeling, reducing errors in thermal offset predictions (Tab.2). Furthermore, the coordinate registration compensation results showed that the Root Mean Square Errors (RMSE) in the X, Y, and Z directions after compensation were reduced by 84.5%, 77.9%, and 78.8%, respectively (Tab.3, Fig.14). These findings highlight the proposed method’s ability to mitigate thermal offset effects and enhance the accuracy of coordinate registration between the tooling MCS and the GCS.This study presents an integrated methodology combining precise temperature field reconstruction, dynamic optimization of virtual measurement points, and coordinate registration compensation to address thermal offset challenges in tooling ERS points under non-uniform temperature fields. Compared to existing methods, the proposed approach demonstrates significant improvements in thermal offset prediction accuracy and registration precision, particularly under sparse measurement conditions and modeling simplifications. The results validate the robustness and reliability of the method for maintaining ERS point precision in challenging thermal environments. However, further research is required to improve computational efficiency and adapt the methodology to real-time applications and more complex industrial scenarios. The findings of this study provide valuable insights and practical solutions for enhancing precision in large-scale metrology and aircraft assembly.