Abstract: Traditional multispectral radiation thermometry typically requires assuming an emissivity model, but actual material emissivity often exhibits complex and irregular distributions. This complexity can lead to insufficient inversion accuracy. To address these issues, this study proposes a multispectral radiation thermometry method integrating a multi-strategy ISGA-NOA algorithm (Information-Sharing and GoldenSearch Augmented Nutcracker Optimization Algorithm via Good Point Set). Without requiring preset emissivity models, it converts the temperature and emissivity inversion problem into a constrained nonlinear optimization problem, simultaneously retrieving true temperature and spectral emissivity. The algorithm employs a good point set and specular reflection strategy to construct a diverse initial population. During the exploration phase, it integrates the golden section search algorithm with an information-sharing strategy, establishing a parameter adaptive adjustment mechanism to effectively enhance global exploration and local exploitation capabilities. Performance was evaluated using CEC2022 benchmark functions, confirming the algorithm's effectiveness and feasibility. Simulation experiments based on six typical emissivity models validated its robustness. Results show relative errors not exceeding 1.18% when inverting true temperatures of 1000 K, 1100 K, and 1200 K. Validation with measured temperature data from rocket engines yielded relative errors below 0.64%. Performance tests and simulations demonstrate the method's effectiveness and engineering applicability, offering an innovative approach for multispectral radiation thermometry.