The reconstruction accuracy of annular temperature fields in high-temperature components such as aero-engine cores is critical for combustion efficiency evaluation and equipment safety operation. Addressing the complex characteristics of annular temperature fields, including radial gradient heterogeneity, circumferential periodic fluctuations, and local temperature inversions, as well as the limitations of traditional reconstruction methods in annular structure adaptability and physical compliance, this paper proposes an annular temperature field reconstruction method integrating adaptive Radial Basis Function (RBF) interpolation, a CNN-LSTM hybrid network, and layered physical constraints. The method first generates an initial temperature field through hybrid-kernel RBF interpolation with kernel parameters dynamically adjusted based on radial distance. Subsequently, a CNN-LSTM network extracts spatial local features to perform residual correction on the initial field. Finally, layered physical constraints are applied regionally according to heat transfer laws and observed inhomogeneities in the temperature distribution, ensuring the physical credibility of the reconstruction results. The effectiveness of the proposed method is validated by calculating error metrics using experimental data under typical operating conditions, and ablation studies are designed to further verify the specific contributions of each module.