Abstract: An approach is proposed for the quality control of the surface temperature data, integrating an Improved Non-Local Means (INLM) algorithm and Particle Swarm Optimization (PSO) for parameter optimization. Referred to as the INLM-PSO method, this method incorporates the Non-Local Means (NLM) approach into surface temperature data quality control for measuring the correlation between reference and target stations using the Euclidean distance and cosine similarity. This enables the estimation of target station observations and subsequent quality control. To assess the effectiveness and adaptability of the INLM-PSO method, it is applied to the quality control analysis of the daily average surface temperature data from 2017 to 2018 at 14 representative surface observation stations across diverse regions in the country. Comparative analysis is conducted against the Inverse Distance Weighted (IDW) and Spatial Regression Test (SRT) methods. Experimental results demonstrate that the proposed method excels over the IDW and SRT methods in effectively identifying suspicious temperature values, achieving high predictive accuracy and showcasing strong stability and adaptability.